DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Response to Amendment
(Submitted 2/26/2026)
In regard to Objections to the Drawings
On Page 18, the applicant has amended [0028] per examiner’s suggestions. The storage 230 is amended to the spec[0028] (see in “specification objections “to be consistent with the drawing.
Upon review, as a result, the examiner hereby WITHDRWAS the objections
In regard to Specifications Objections
On Page 17, the applicant gas amended the paragraphs [0001], [0012], [0028] and [0029] per examiner’s suggestions.
Upon review, as a result, the examiner hereby WITHDRWAS the objections
In regard to Claims Objections
On Page 17, the applicant has stated that the claim 15 is amended accordingly.
Upon review, as a result, the examiner hereby WITHDRWAS the objections on claims 15-20.
In regard to 101 Rejection
From Pages 18-22, the applicant argues on 101 rejections with respect to the amended claims 1, 8 and 15. On Page 19, the applicant argues that ensemble ML is a practical application. The applicant has amended “ computing a cost associated with each of the selected base forecast models”
Examiner’s Response
The “ computing a cost associated with each of the selected base forecast models” is still a mathematical framework in which they abstract the data into parameters states or variables. It is recognized as a mathematical function without any measurable technical improvement. In regard to ensemble ML itself is not a practical application, It can be a practical strategy that need to show the technological improvement integrated into a practical application with invention as solving a concrete technical problem with measurable benefits, supported by detailed implementation details. In general, s Highlight non-obviousness by demonstrating a specific technical advantage or efficiency gain from your approach. Merely stating a resource need for wired or wireless tower is not sufficient. The applicant need to provide detailed examples of how your cost model improves decision-making compared to prior art, or compare the financial impact of each scenario or sensitivity analysis to see how changes in key variables affect total cost
In CONCLUSION, the examiner maintains the 101 rejections for claims 1, 3-8, and 10-20.
In regard to 103 Rejection
The applicant has amended the independent claims 1, 8 and 15. The applicant has argued on the prior art on Page 24 with regard to the amendments, such arguments include [original claim 3 with respect to “Nara” and “Wu”, argument with regard to “Ames” and “Cheng” on Page 25].
Examiner’s Response
The examiner submits that the applicant’s arguments are MOOT as a result of new grounds of rejections based on new reference “Jennings” in view of “Nara”, in view of “Xu”.
In CONCLUSION, the examiner rejects claims 1, 3-8, and 10-20 under 103 and MOVE the application as FINAL REJECTION.
Claim Rejections - 35 USC § 101
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claim 1, 3-8 and 10-20 are rejected under 35 U.S.C. 101 because the claims are directed to an abstract idea without significantly more.
Step 1: The claim is directed to a method, corresponding to a process, which is one of the
statutory categories.
In regard to claim 1: (Currently Amended):
Step 2A, Prong 1:
“ by computing a cost associated with each of the selected forecast models” encompasses the mathematical concept of algebraic functions associated with the cost values of the base forecast models (i.e., mathematical relationships, mathematical formulas or equations, and mathematical calculations).
Additional Elements
Step 2A, Prong 2:
“A method comprising” recited in the preamble does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ determining whether characteristics of historic time series data continuously collected from a wires or wireless tower includes seasonality” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ based on the determination, selecting from available forecast models from available forecast models, either both of linear and non-linear forecast mode as a plurality of candidate base forecast models”
does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ generating the ensemble forecast model adjusted overtime on-the-fly by the continuously collected historic time series data” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ selecting, from available forecast models, either both of linear and non-linear forecast models or non-linear forecast models as a plurality of candidate base forecast” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ generating an ensemble forecast model adjusted over time on-the-fly by the continuously collected historic time series data,” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ determining a set of parameters to be used for generating the ensemble forecast models based on the costs associated respectively with the selected
and creating the ensemble forecast model based on the selected base forecast models in accordance with the set of parameters; does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ forecasting a resource need associated with the wired or wireless tower using the ensemble model, wherein the forecasted resource need is for allocating a resource to the wired or wireless” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ collecting resource usage data associated with the wired or wireless tower” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ and adding the resource usage data to the historic time series data to form a dynamic self- adaptation loop with respect to the wired or wireless tower” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
Step 2B:
“ A method comprising” recited in the preamble does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ determining whether characteristics of historic time series data continuously collected from a wires or wireless tower includes seasonality” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ based on the determination, selecting from available forecast models from available forecast models, either both of linear and non-linear forecast mode as a plurality of candidate base forecast models”
does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ generating the ensemble forecast model adjusted overtime on-the-fly by the continuously collected historic time series data” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ selecting, from available forecast models, either both of linear and non-linear forecast models or non-linear forecast models as a plurality of candidate base forecast” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ generating an ensemble forecast model adjusted over time on-the-fly by the continuously collected historic time series data,” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ determining a set of parameters to be used for generating the ensemble forecast models based on the costs associated respectively with the selected
“ and creating the ensemble forecast model based on the selected base forecast models in accordance with the set of parameters” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ forecasting a resource need associated with the wired or wireless tower using the ensemble model, wherein the forecasted resource need is for allocating a resource to the wired or wireless” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ collecting resource usage data associated with the wired or wireless tower” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ and adding the resource usage data to the historic time series data to form a dynamic self- adaptation loop with respect to the wired or wireless tower” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
In regard to claim 3: (Currently Amended):
Step 2A, Prong 2:
“ wherein the selecting the plurality of candidate base forecast model comprises:” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ if the historic time series data exhibits seasonality, selecting one or more non-linear forecast base models from the available forecast models as the plurality of candidate base forecast models;” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ and if the historic time series data does not exhibit seasonality, selecting one or more non- linear forecast base models and one or more linear forecast base models from the available forecast models as the plurality of candidate base forecast models” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
Step 2B:
“ wherein the selecting the plurality of candidate base forecast model comprises:” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ if the historic time series data exhibits seasonality, selecting one or more non-linear forecast base models from the available forecast models as the plurality of candidate base forecast models;” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ and if the historic time series data does not exhibit seasonality, selecting one or more non- linear forecast base models and one or more linear forecast base models from the available forecast models as the plurality of candidate base forecast models” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
In regard to claim 4: (Currently Amended):
Step 2A, Prong 1:
“performing linear regression on smoothed historic time series data” is a mathematical concept of calculus/linear algebra functions (i.e., mathematical relationships, mathematical formulas or equations, and mathematical calculations).
“ and determining whether the historic time series data exhibits seasonality based on the auto-correlation results” is associated with the mental process of judgement/opinion of time series data.
Step 2A, Prong 2:
“ generating detrended historic time series data based on the smoothed historic time series data and the linear regression result” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ performing auto-correlation on the smoothed time series data and the detrended historic time series data to generate auto-correlation results” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
Step 2B:
“ generating detrended historic time series data based on the smoothed historic time series data and the linear regression result” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ performing auto-correlation on the smoothed time series data and the detrended historic time series data to generate auto-correlation results” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
In regard to claim 5: (Original)
Step 2A, Prong 2:
“ wherein the auto-correlation results include: a first auto-correlation metric obtained via auto-correlation on the smoothed historic time series data” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“and a second auto-correlation metric obtained via auto-correlation on the detrended historic time series data” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
Step 2B:
“ wherein the auto-correlation results include: a first auto-correlation metric obtained via auto-correlation on the smoothed historic time series data” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“and a second auto-correlation metric obtained via auto-correlation on the detrended historic time series data” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
In regard to claim 6: (Currently Amended)
Step 2A, Prong 1:
“ computing a measure indicative of the performance of the candidate base forecast model based on the forecast result” encompasses the mathematical concept of calculus/statistical functions (i.e., mathematical relationships, mathematical formulas or equations, and mathematical calculations).
“ selecting the plurality of “ can be performed mentally or with a pen and paper.
Step 2A, Prong 2:
“ wherein the selecting the plurality of candidate base forecast models comprises: designating multiple candidate base forecast models from the available forecast models based on whether the historic time series data exhibits seasonality” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ and identifying the plurality of candidate base forecast models from the multiple candidate base forecast models based on forecast performance of each of the multiple candidate base forecast models, by with respect to each of the multiple candidate base forecast models” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
Step 2B:
“ wherein the selecting the plurality of candidate base forecast models comprises: designating multiple candidate base forecast models from the available forecast models based on whether the historic time series data exhibits seasonality” ” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ and identifying the plurality of candidate base forecast models from the multiple candidate base forecast models based on forecast performance of each of the multiple candidate base forecast models, by with respect to each of the multiple candidate base forecast models” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
In regard to claim 7: (Currently Amended)
Step 2A, Prong 2:
“ wherein the ensemble candidate forecast model corresponds to a weighted sum of the plurality of base forecast models” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ wherein the set of parameters correspond to weights to be applied to the respective base forecast models” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
Step 2B:
“ wherein the ensemble candidate forecast model corresponds to a weighted sum of the plurality of base forecast models” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ wherein the set of parameters correspond to weights to be applied to the respective base forecast models” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
In regard to claim 8: (Currently Amended):
Step 2A, Prong 1:
“ by computing a cost associated with each of the selected forecast models” encompasses the mathematical concept of algebraic functions associated with the cost values of the base forecast models (i.e., mathematical relationships, mathematical formulas or equations, and mathematical calculations).
Additional Elements
Step 2A, Prong 2:
“A machine readable and non-transitory medium having information recorded thereon, wherein the information, when read by the machine, causes the machine to perform the following steps:”
recited in the preamble does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ determining whether characteristics of historic time series data continuously collected from a wires or wireless tower includes seasonality” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ based on the determination, selecting from available forecast models from available forecast models, either both of linear and non-linear forecast mode as a plurality of candidate base forecast models”
does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ generating the ensemble forecast model adjusted overtime on-the-fly by the continuously collected historic time series data” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ selecting, from available forecast models, either both of linear and non-linear forecast models or non-linear forecast models as a plurality of candidate base forecast” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ generating an ensemble forecast model adjusted over time on-the-fly by the continuously collected historic time series data,” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ determining a set of parameters to be used for generating the ensemble forecast models based on the costs associated respectively with the selected
and creating the ensemble forecast model based on the selected base forecast models in accordance with the set of parameters; does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ forecasting a resource need associated with the wired or wireless tower using the ensemble model, wherein the forecasted resource need is for allocating a resource to the wired or wireless” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ collecting resource usage data associated with the wired or wireless tower” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ and adding the resource usage data to the historic time series data to form a dynamic self- adaptation loop with respect to the wired or wireless tower” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
Step 2B:
““A machine readable and non-transitory medium having information recorded thereon, wherein the information, when read by the machine, causes the machine to perform the following steps:” recited in the preamble does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ determining whether characteristics of historic time series data continuously collected from a wires or wireless tower includes seasonality” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ based on the determination, selecting from available forecast models from available forecast models, either both of linear and non-linear forecast mode as a plurality of candidate base forecast models”
does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ generating the ensemble forecast model adjusted overtime on-the-fly by the continuously collected historic time series data” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ selecting, from available forecast models, either both of linear and non-linear forecast models or non-linear forecast models as a plurality of candidate base forecast” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ generating an ensemble forecast model adjusted over time on-the-fly by the continuously collected historic time series data,” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ determining a set of parameters to be used for generating the ensemble forecast models based on the costs associated respectively with the selected
“ and creating the ensemble forecast model based on the selected base forecast models in accordance with the set of parameters” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ forecasting a resource need associated with the wired or wireless tower using the ensemble model, wherein the forecasted resource need is for allocating a resource to the wired or wireless” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ collecting resource usage data associated with the wired or wireless tower” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ and adding the resource usage data to the historic time series data to form a dynamic self- adaptation loop with respect to the wired or wireless tower” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
In regard to claim 10: (Currently Amended):
Step 2A, Prong 2:
“ wherein the selecting the plurality of candidate base forecast model comprises:” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ if the historic time series data exhibits seasonality, selecting one or more non-linear forecast base models from the available forecast models as the plurality of candidate base forecast models;” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ and if the historic time series data does not exhibit seasonality, selecting one or more non- linear forecast base models and one or more linear forecast base models from the available forecast models as the plurality of candidate base forecast models” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
Step 2B:
“ wherein the selecting the plurality of candidate base forecast model comprises:” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ if the historic time series data exhibits seasonality, selecting one or more non-linear forecast base models from the available forecast models as the plurality of candidate base forecast models;” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ and if the historic time series data does not exhibit seasonality, selecting one or more non- linear forecast base models and one or more linear forecast base models from the available forecast models as the plurality of candidate base forecast models” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
In regard to claim 11: (Currently Amended):
Step 2A, Prong 1:
“performing linear regression on smoothed historic time series data” is a mathematical concept of calculus/linear algebra functions (i.e., mathematical relationships, mathematical formulas or equations, and mathematical calculations).
“ and determining whether the historic time series data exhibits seasonality based on the auto-correlation results” is associated with the mental process of judgement/opinion of time series data.
Step 2A, Prong 2:
“ generating detrended historic time series data based on the smoothed historic time series data and the linear regression result” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ performing auto-correlation on the smoothed time series data and the detrended historic time series data to generate auto-correlation results” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
Step 2B:
“ generating detrended historic time series data based on the smoothed historic time series data and the linear regression result” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ performing auto-correlation on the smoothed time series data and the detrended historic time series data to generate auto-correlation results” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
In regard to claim 12: (Original)
Step 2A, Prong 2:
“ wherein the auto-correlation results include: a first auto-correlation metric obtained via auto-correlation on the smoothed historic time series data” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“and a second auto-correlation metric obtained via auto-correlation on the detrended historic time series data” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
Step 2B:
“ wherein the auto-correlation results include: a first auto-correlation metric obtained via auto-correlation on the smoothed historic time series data” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“and a second auto-correlation metric obtained via auto-correlation on the detrended historic time series data” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
In regard to claim 13: (Currently Amended)
Step 2A, Prong 1:
“ computing a measure indicative of the performance of the candidate base forecast model based on the forecast result” encompasses the mathematical concept of calculus/statistical functions (i.e., mathematical relationships, mathematical formulas or equations, and mathematical calculations).
“ selecting the plurality of “ can be performed mentally or with a pen and paper.
Step 2A, Prong 2:
“ wherein the selecting the plurality of candidate base forecast models comprises: designating multiple candidate base forecast models from the available forecast models based on whether the historic time series data exhibits seasonality” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ and identifying the plurality of candidate base forecast models from the multiple candidate base forecast models based on forecast performance of each of the multiple candidate base forecast models, by with respect to each of the multiple candidate base forecast models” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
Step 2B:
“ wherein the selecting the plurality of candidate base forecast models comprises: designating multiple candidate base forecast models from the available forecast models based on whether the historic time series data exhibits seasonality” ” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ and identifying the plurality of candidate base forecast models from the multiple candidate base forecast models based on forecast performance of each of the multiple candidate base forecast models, by with respect to each of the multiple candidate base forecast models” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
In regard to claim 14: (Currently Amended)
Step 2A, Prong 2:
“ wherein the ensemble candidate forecast model corresponds to a weighted sum of the plurality of base forecast models” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ wherein the set of parameters correspond to weights to be applied to the respective base forecast models” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
Step 2B:
“ wherein the ensemble candidate forecast model corresponds to a weighted sum of the plurality of base forecast models” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ wherein the set of parameters correspond to weights to be applied to the respective base forecast models” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
In regard to claim 15: (Currently Amended):
Step 2A, Prong 1:
“ by computing a cost associated with each of the selected forecast models” encompasses the mathematical concept of algebraic functions associated with the cost values of the base forecast models (i.e., mathematical relationships, mathematical formulas or equations, and mathematical calculations).
Additional Elements
Step 2A, Prong 2:
“ A system, comprising: a data preprocessor implemented by a processor and configured for “recited in the preamble does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ determining whether characteristics of historic time series data continuously collected from a wires or wireless tower includes seasonality” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ a performance based model selector implemented by a processor and configured for based on the determination, selecting, from available forecast models, either both of linear and non-linear forecast models or non-linear forecast models as a plurality of candidate base forecast” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ an integrated model ensemble unit implemented by a processor and configured for generating an ensemble forecast model adjusted over time on-the-fly by the continuously collected historic time series data” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“determining a set of parameters to be used for generating the ensemble forecast models based on the costs associated respectively with the selected base forecast models” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ and creating the ensemble forecast model based on the selected base forecast models in accordance with the set of parameters” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ an ensemble model based forecaster implemented by a processor and configured for forecasting a resource need associated with the wired or wireless tower using the ensemble model, wherein the forecasted resource need is for allocating a resource to the wired or wireless “does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ and a resource use data collector implemented by a processor and configured for collecting resource usage data associated with the wired or wireless tower” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ and adding the resource usage data to the historic time series to form a self-adaptation loop with respect to the wired or wireless tower” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
Step 2B:
“ A system, comprising: a data preprocessor implemented by a processor and configured for “recited in the preamble does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ determining whether characteristics of historic time series data continuously collected from a wires or wireless tower includes seasonality” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ a performance based model selector implemented by a processor and configured for based on the determination, selecting, from available forecast models, either both of linear and non-linear forecast models or non-linear forecast models as a plurality of candidate base forecast” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ an integrated model ensemble unit implemented by a processor and configured for generating an ensemble forecast model adjusted over time on-the-fly by the continuously collected historic time series data” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“determining a set of parameters to be used for generating the ensemble forecast models based on the costs associated respectively with the selected base forecast models” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ and creating the ensemble forecast model based on the selected base forecast models in accordance with the set of parameters” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h)).
“ an ensemble model based forecaster implemented by a processor and configured for forecasting a resource need associated with the wired or wireless tower using the ensemble model, wherein the forecasted resource need is for allocating a resource to the wired or wireless “does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ and a resource use data collector implemented by a processor and configured for collecting resource usage data associated with the wired or wireless tower” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ and adding the resource usage data to the historic time series to form a self-adaptation loop with respect to the wired or wireless tower” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
In regard to 16: (Currently Amended):
Step 2A, Prong 2:
“ wherein the selecting the plurality of candidate base forecast model comprises:” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ if the historic time series data exhibits seasonality, selecting one or more non-linear forecast base models from the available forecast models as the plurality of candidate base forecast models;” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ and if the historic time series data does not exhibit seasonality, selecting one or more non- linear forecast base models and one or more linear forecast base models from the available forecast models as the plurality of candidate base forecast models” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
Step 2B:
“ wherein the selecting the plurality of candidate base forecast model comprises:” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ if the historic time series data exhibits seasonality, selecting one or more non-linear forecast base models from the available forecast models as the plurality of candidate base forecast models;” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ and if the historic time series data does not exhibit seasonality, selecting one or more non- linear forecast base models and one or more linear forecast base models from the available forecast models as the plurality of candidate base forecast models” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
In regard to claim 17: (Currently Amended):
Step 2A, Prong 1:
“performing linear regression on smoothed historic time series data” is a mathematical concept of calculus/linear algebra functions (i.e., mathematical relationships, mathematical formulas or equations, and mathematical calculations).
“ and determining whether the historic time series data exhibits seasonality based on the auto-correlation results” is associated with the mental process of judgement/opinion of time series data.
Step 2A, Prong 2:
“ generating detrended historic time series data based on the smoothed historic time series data and the linear regression result” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ performing auto-correlation on the smoothed time series data and the detrended historic time series data to generate auto-correlation results” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
Step 2B:
“ generating detrended historic time series data based on the smoothed historic time series data and the linear regression result” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ performing auto-correlation on the smoothed time series data and the detrended historic time series data to generate auto-correlation results” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
In regard to claim 18: (Original)
Step 2A, Prong 2:
“ wherein the auto-correlation results include: a first auto-correlation metric obtained via auto-correlation on the smoothed historic time series data” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“and a second auto-correlation metric obtained via auto-correlation on the detrended historic time series data” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
Step 2B:
“ wherein the auto-correlation results include: a first auto-correlation metric obtained via auto-correlation on the smoothed historic time series data” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“and a second auto-correlation metric obtained via auto-correlation on the detrended historic time series data” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
In regard to claim 19: (Currently Amended)
Step 2A, Prong 1:
“ computing a measure indicative of the performance of the candidate base forecast model based on the forecast result” encompasses the mathematical concept of calculus/statistical functions (i.e., mathematical relationships, mathematical formulas or equations, and mathematical calculations).
“ selecting the plurality of “ can be performed mentally or with a pen and paper.
Step 2A, Prong 2:
“ wherein the selecting the plurality of candidate base forecast models comprises: designating multiple candidate base forecast models from the available forecast models based on whether the historic time series data exhibits seasonality” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ and identifying the plurality of candidate base forecast models from the multiple candidate base forecast models based on forecast performance of each of the multiple candidate base forecast models, by with respect to each of the multiple candidate base forecast models” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
Step 2B:
“ wherein the selecting the plurality of candidate base forecast models comprises: designating multiple candidate base forecast models from the available forecast models based on whether the historic time series data exhibits seasonality” ” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ and identifying the plurality of candidate base forecast models from the multiple candidate base forecast models based on forecast performance of each of the multiple candidate base forecast models, by with respect to each of the multiple candidate base forecast models” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
In regard to claim 20: (Currently Amended)
Step 2A, Prong 2:
“ wherein the ensemble candidate forecast model corresponds to a weighted sum of the plurality of base forecast models” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ wherein the set of parameters correspond to weights to be applied to the respective base forecast models” does not integrate the judicial exception into a practical application. This additional element is merely using a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
Step 2B:
“ wherein the ensemble candidate forecast model corresponds to a weighted sum of the plurality of base forecast models” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
“ wherein the set of parameters correspond to weights to be applied to the respective base forecast models” does not amount to significantly more than the judicial exception in the claim. The additional element merely uses a computer as a tool to perform an abstract idea (see MPEP 2106.05(h).
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1, 3, 7-8, 10, 14-16 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over
Jiansheng Lin, et al (hereinafter Lin) A data-driven base station sleeping strategy based on traffic prediction." IEEE Transactions on Network Science and Engineering 11.6 (2021): 5627-5643.
In view of Balakrishnan Narayanaswamy (hereinafter Nara ) US 2015/0253463 A1;
In regard to claim 1: (Amended):
Lin discloses:
- A method, comprising: determining whether historic time series continuously collected from a wired or wireless tower includes seasonality;
In [1, Page 1]: “ In [4], the BS sleeping mode was designed according to a deterministic traffic changes over time. A similar idea was proposed in [5], where some BSs are shut down during the night time. All of these works utilize the periodic traffic fluctuation to save energy by switching some BSs to sleep mode. In Fig. 1, we plot the wireless traffic of Milan city at 10am and 10pm. From the figure, we can observe a significant traffic disparity between different regions for a given time instance. Meanwhile, comparing the traffic at 10am and that at 10pm, we can find that some regions tend to have traffic peaks in the day time, while some regions behave in a opposite way. Hence, in [6–9], the authors propose to selectively shut down BS resources in the network during off-peak hours based on the traffic monitor”
PNG
media_image1.png
298
526
media_image1.png
Greyscale
in [1, Page 1]: “ the accurate capacity model of BSs is another key factor of sleeping strategy. It is known that macro-cell BSs (MBSs) aim at providing an umbrella coverage while small-cell BSs (SBSs) target at throughput enhancement in traffic hotspots.
in [1, Page 1]: “ Network operators are deploying a large number of base stations (BSs) to meet the traffic requirements in peak hours. Hence, network densification has become an irresistible trend since 3G network. With the development of dense and ultra dense small-cell networks, the excessive power consumption has become a major issue of network operation. For example, BSs account for two-thirds of the total energy consumption in a wireless network [1,2]. Even if there is no or little traffic load, a BS can still consume more than 90% of their peak energy. The ultra-dense deployment of small-cell BSs (SBSs) leads to a low energy utilization in most time [3].
In [1, Page 2]: “ We propose a novel cellular traffic prediction method, where a multi-graph convolutional network (MGCN) is used to extract the spatial features, and a multi-channel long short-term memory (LSTM) network is adopted to efficiently capture the major frequency components in the traffic data. In addition, an attention scheme is designed to distinguish the significance of traffic sequences at different time”
- based on the determination, selecting, from available forecast models, either both of linear and non-linear forecast models or non-linear forecast models as a plurality of candidate base forecast models
In [1, Page 2]: “Based on the predicted traffic demand and the learned MBS and SBS capacity models, the optimal numbers of active MBSs and SBSs are obtained to minimize the power consumption in a given area. Due to the intractable expression of the capability functions obtained by the neural network, we have used both linear and non-linear fitting models to derive the optimal strategies without the high complexity of exhaustive searching”,
In [1, Page 2]: “ a novel cellular traffic prediction method, where a multi-graph convolutional network (MGCN) is used to extract the spatial features, and a multi-channel long short-term memory (LSTM) network is adopted to efficiently capture the major frequency components in the traffic data. In addition, an attention scheme is designed to distinguish the significance of traffic sequences at different time”,
In [II, Page 2]: “Many works on traffic forecasting are based on neural networks”,
In [II, Page 2]: “the differential evolution algorithm was combined with the back propagation algorithm to optimize the fuzzy neural network forecasting network traffic. A deep belief network (DBN) with Gaussian models was proposed in [13] to forecast the traffic demand in wireless mesh networks.
In [II, Page 2]: “ autoencoders were utilized to extract the spatial features in [15], then LSTM was adopted after to perform the traffic prediction. In [16–18], convolution neural network (CNN) and LSTM were combined to construct a spatial temporal neural network for traffic prediction, where the CNN concentrates on the spatial features. In [19], the attention mechanism was introduced into the Conv-LSTM network
[BRI: training a separate CNN–LSTM and linear/nonlinear fitting models, then combine their forecasts does represent an ensembled forecasting model.]
- collecting resource usage data associated with the wired or wireless tower
in [3, Page 4]: “ the wire-less traffic and the deployment of MBSs and SBSs in an area highly depend on the spatial/region information, the propagation environment and the population distribution. Hence, to analyze the characteristics of an area, several related datasets are introduced in this work: POIs, social activities, and the number of BSs deployed [20]. All these dataset are transformed to represent the 10000 square grids like the cellular traffic dataset”
In [IV, Page 5]:
PNG
media_image2.png
207
1082
media_image2.png
Greyscale
In [Abstract, Page 1]: “ we propose a novel data-driven intelligent BS sleeping mechanism based on a traffic prediction model that measures the BSs’ capacity in different region,
In [Abstract, Page 1]: “ the proposed BS sleeping strategy using an approximated non-linear model of capacity function achieves a near-optimal energy-saving performance with relatively low complexity”
- forecasting a resource need associated with the wired or wireless tower using the ensemble model, wherein the forecasted resource need is for allocating a resource to the wired or wireless tower;
In [2, Page 2]: “ Various approaches have been proposed to reduce the energy consumption of a mobile cellular network, which can generally be divided into the two categories: 1) improving the energy efficiency of the network components, and 2) turning off some of the network resources selectively. The BS sleeping strategy is a typical approach in the second category. Deactivating some deployed BSs during off-peak hours, not only the energy consumption but also the inter-cell interference in the network can be reduced.
In [V, Page 7]: “ In this section, MBSs’ and SBSs’ capacity functions,
C
m
(
·) and
C
s
(·), are modeled, to reveal the maximum traffic volume (i.e., throughput) that a number of MBSs or SBSs can provide in an area with a characteristic vector r. Different from the theoretical capacity function, such as Shannon Theorem, in this section, the neural network is utilized to construct the capacity model.
PNG
media_image3.png
581
672
media_image3.png
Greyscale
In [V, Page 7]: “ With these results, we can clear see the impact of the environment characteristics and the number of BSs on wireless maximum traffic, which verifies our modeling of the capacity functions
C
m
(
n
m
,r) and
C
s
(
n
s
,r). That is, the capacity is highly dependent on r. Since the number of BSs deployed in a square grid determines the provided wireless resource,
In [V, Page 8]: “ we choose it as one parameter of the capacity function [35]. Meanwhile, the geographic environment greatly impacts the propagation model of wireless signals [36]. Hence, the area characteristic vector is chosen as another parameter of the capacity function.
- and adding the resource usage data to the historic time series data to form a dynamic self- adaptation loop with respect to the wired or wireless tower.
In [1, Page 1]: the BS sleeping strategy aiming at reducing the base operating power of BSs, becomes an attractive method to save energy consumption. Various strategies to enable dynamic on-off for BSs have been proposed in these years,
In [1, Page 1]: “ the BS sleeping mode was designed according to a deterministic traffic changes over time,
In [1, Page 1]: “ the BS sleeping mode was designed according to a deterministic traffic changes over time. A similar idea was proposed in [5], where some BSs are shut down during the night time. All of these works utilize the periodic traffic fluctuation to save energy by switching some BSs to sleep mode.
In [VII, Page 11]: “ In this section, extensive experiments are conducted to evaluate the performance of our cellular traffic prediction model and the proposed dynamic BS sleeping strategy. For MBS and SBS, the parameters are adopted according to [40, 44]: 1) the fixed operating energy consumption: 10W and 5W; 2) the number of antennas: 3 and 1, and the circuit components energy consumption of antennas: 1W and 0.8W; 3) power amplifier efficiency: 0.128 and 0.12, and input power of BS: 43dBm and 33dBm; 4) the efficiency of load consumption: 1.15 · 10−9J/bit and 1.05 · 10−9J/bit
A dynamic base station (BS) sleeping strategy can indeed be designed to adapt parameters based on fixed energy consumption, number of antennas, and other resource metrics, and to integrate these into a self-adapting loop using historic time series data. This approach is relevant for both wired and wireless towers, as it can be applied to any BS type with measurable resource usage.
Lin does not explicitly disclose:
- generating an ensemble forecast model adjusted over time on-the-fly by the continuously collected historic time series data, by computing a cost associated with each of the selected base forecast models,
- determining a set of parameters to be used for generating the ensemble forecast models based on the costs associated respectively with the selected base forecast models,
- and creating the ensemble forecast model based on the selected base forecast models in accordance with the set of parameters;
However, Nara discloses:
- generating an ensemble forecast model adjusted over time on-the-fly by the continuously collected historic time series data, by computing a cost associated with each of the selected base forecast models,
In [0013] “ As described herein, an aspect of the present invention includes generating and recommending forecasting models for ensembles. Given a collection of forecasting models (such as weather models, hydrology models, climate models, etc.), historically-related observations, current model forecasts, socio-economic cost models, and/or user constraints on forecast lead time and computational resources, at least one embodiment of the invention includes determining an optimal ensemble of models and corresponding model configurations to be run in a given situation”,
In [0006]: “ In one aspect of the present invention, techniques for generating an ensemble of forecasting models are provided",
In [0016]: “ the resource and cost estimation engine 116 provide input to the model and parameter selection engine 110, which also receives input in the form of budget information, computational resource information, and cost information 104"
and in [0031]: "In further reference to FIG. 1, and as detailed herein, the model and parameter selection engine 110 determines, based on… an ensemble of models and parameterizations that are suited for the given event, as well as the availability of budget and/or computational resources" [i.e., the model and parameter selection engine generates an ensemble forecast model based on computed costs associated with the forecasting models it receives]),
- determining a set of parameters to be used for generating the ensemble forecast models based on the costs associated respectively with the selected base forecast models,
In [0024]: “ The resource and cost estimation engine 116, as depicted in FIG. 1, can be used to decide on the feasibility of running a configuration based on user-specified constraints. Examples of user-specified constraints can include the available number of processors, total memory available, model output required in under two hours, etc. Based on historical data pertaining, for example, to computational run-times, memory requirements, and hardware resource information, at least one embodiment of the invention includes creating models for hardware configuration and/or computation time/memory cost for a particular model configuration… This can include categorical features corresponding to various parameters such as physics options, cumulus parameterization, etc., or valued features such as number of grid points, time-step used, clock frequency of the processors, etc”,
and [0026]: “ The model and parameter selection engine 110 determines the final set of model configurations for a particular event. This selection depends on multiple factors such as socio-economic cost associated with the event and the lead times required for an accurate forecast, as well as the user-specified constraints on cost and computational resource availability. As noted, the resource and cost estimation engine 116 computes the computational time and/or memory and resource requirements for each model configuration" [i.e., a set of model configuration parameters associated with the constraints on operating costs and resources are determined by the resource and cost estimation engine, which feeds the parameters into the model and parameter selection engine, for generating the ensemble forecast models”.
- and creating the ensemble forecast model based on the selected base forecast models in accordance with the set of parameters;
In [0031]: “ In further reference to FIG. 1, and as detailed herein, the model and parameter selection engine 110 determines, based on… the estimated resource requirement and cost (from the resource and cost estimation engine 116), an ensemble of models and parameterizations that are suited for the given event, as well as the availability of budget and/or computational resources" [i.e., the parameters from the resource and cost estimation engine are fed into the model and the parameter selection engine to generate an ensemble forecast model”,
and in [0035]: “ Step 208 includes determining an ensemble of one or more of the multiple forecasting models to apply to the given environmental event based on (i) said estimated accuracy value for each of the multiple forecasting models, and (ii) said cost and said one or more resource requirements for each of the multiple forecasting models" [i.e., forecasting models are used in the selection/generation of the ensemble forecast model”.
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Lin and Nara.
Lin teaches a forecasting model that collects time series data from a wireless tower (base stations) and combines linear and non-linear forecast models.
Nara teaches ensembled forecasting model.
One of ordinary skill would have motivation to combine Lin and Nara that collects a forecasting model that includes a weather model within the context of historically related observatories and determines an optimal ensemble model (Nara [0013]).
In regard to claim 3: (Currently Amended):
Lin discloses:
- if the historic time series data exhibits seasonality, selecting one or more non-linear forecast base models from the available forecast models as the plurality of candidate base forecast models;
In [I, Page 1]: “ city CBD is full of tall buildings while the rural area is usually scattered with low-density houses. Therefore, an accurate capacity model of different types of BSs considering various region characteristics should be constructed”,
In [VII, Page 13]: “ compared with the energy consumption with all active MBSs and SBSs as 100%, the energy saved by the BS sleeping strategy using non-linear model is more obvious in city CBD. In more detail, the energy saved in city CBD approaches 63% in one week, and near 54% of the energy is saved in suburban regions. This is because the large number of BSs deployed and the traffic disparity between the peak and off-peak time”
- and if the historic time series data does not exhibit seasonality, selecting one or more non- linear forecast base models and one or more linear forecast base models from the available forecast models as the plurality of candidate base forecast models
In [VI, Page 9]: “ In this section, the optimal BS sleeping strategy is inves tigated which aims to minimize the energy consumption in a given area.
In [VI, Page 9]: “ In this work, we aim to develop a data-driven BS sleeping strategy based on public data with limited general cellular networks [39], e.g., overall traffic and BSs’s number, therefore, we define a cellular energy saving problem to seek the optimal BSs’ number in an area. According to [40], the energy consumption of a BS mainly includes the following parts: 1) the fixed operating energy consumption of a BS, 2) the energy consumption of the circuit components in BS’ antennas which is proportional to the number of antennas, 3) the energy consumption of the power amplifier, and 4) the energy consumption on traffic load which is linear with the traffic provided.
In [VI, Page 9]: “ the network throughput is usually assumed to increase linearly with respect to the number of deployed BSs, especially in sparse scenarios where the aggregated interference is relatively small [41,42].
C
m
(
n
m
,r) and
C
s
(
n
s
,r).
In [VI, Page 9]: “ In Fig. 13, it can be seen that, for a given square grid with characteristics r, 1)
C
s
(
n
s
,r)almost increases linearly with
n
s
in the investigated density range, and 2)
C
m
(
n
m
,r) increase”,
In [VI, Page 10]: “linearly when the density of MBSs is less than 50/km2, i.e., 3 MBSs in a square grid. Inspired by these results, linear capacity model can be used to approximately fit the results of
C
s
(
n
s
,r) and
C
m
(
n
m
,r)
In [VI, Page 10]: “ Apparently, the rapidly increasing aggregated inter-cell interference in the network densification will suppress the growth of the network throughput, such kind of results have been verified by many theoretical analysis on dense and ultra dense networks [43]. Even with various kinds of interference coordination and cancelation techniques, the linear increase of network capacity cannot be achieved in the dense networks
In [VI, Page 10]:
To better approach the real ca pacity achieved by the datasets, a non-linear function to fit the outcome of
C
m
(
n
m
,such as a1n2 m+b1nm+c1. For SBSs, due to the density range in the dataset, this performance degradation does not appear in the outcome. Since the linear model works well and leads to a smallest fitting error, the linear model is still preferred for
C
s
(
n
s
,r).
[BRI: in less dense area (rural) seasonality based optimization may not be required as may be represented by the historic time series data, linear and non-linear models can be combined. For dense regions (urban with optimization based on seasonality) due to small fitting errors, a linear model may be sufficient)
In regard to claim 7: (Currently Amended):
Lin does not explicitly disclose:
- wherein the ensemble forecast model corresponds to a weighted sum of the plurality of base forecast models
- wherein the set of parameters correspond to weights to be applied to the respective base forecast models
However, Nara discloses:
- wherein the ensemble forecast model corresponds to a weighted sum of the plurality of base forecast models
In [0028]: “Parameterization, as used herein, includes weights and/or reliabilities associated with different models, which can be used to combine forecasts from multiple models" [i.e., the ensemble forecast is formed by combining multiple base models according to specific weights, effectively forming a weighted sum]),
- wherein the set of parameters correspond to weights to be applied to the respective base forecast models
In [0035]: “Step 208 includes determining an ensemble of one or more of the multiple forecasting models… The determining step can include determining an ensemble of one or more of the multiple forecasting models and a parameterization for each of the one or more of the multiple forecasting models to apply to the given environmental event" [i.e., the ensemble is formed by selecting specific base forecasting models and separate weights are applied to each selected model's forecast output to combine them into a single ensemble prediction”.
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Lin and Nara.
Lin teaches a forecasting model that collects time series data from a wireless tower (base stations) and combines linear and non-linear forecast models.
Nara teaches ensembled forecasting model.
One of ordinary skill would have motivation to combine Lin and Nara that collects a forecasting model that includes a weather model within the context of historically related observatories and determines an optimal ensemble model (Nara [0013]).
In regard to claim 8: (Currently Amended):
Lin discloses:
- determining whether historic time series continuously collected from a wired or wireless tower includes seasonality;
In [1, Page 1]: “ In [4], the BS sleeping mode was designed according to a deterministic traffic changes over time. A similar idea was proposed in [5], where some BSs are shut down during the night time. All of these works utilize the periodic traffic fluctuation to save energy by switching some BSs to sleep mode. In Fig. 1, we plot the wireless traffic of Milan city at 10am and 10pm. From the figure, we can observe a significant traffic disparity between different regions for a given time instance. Meanwhile, comparing the traffic at 10am and that at 10pm, we can find that some regions tend to have traffic peaks in the day time, while some regions behave in a opposite way. Hence, in [6–9], the authors propose to selectively shut down BS resources in the network during off-peak hours based on the traffic monitor”
PNG
media_image1.png
298
526
media_image1.png
Greyscale
in [1, Page 1]: “ the accurate capacity model of BSs is another key factor of sleeping strategy. It is known that macro-cell BSs (MBSs) aim at providing an umbrella coverage while small-cell BSs (SBSs) target at throughput enhancement in traffic hotspots.
in [1, Page 1]: “ Network operators are deploying a large number of base stations (BSs) to meet the traffic requirements in peak hours. Hence, network densification has become an irresistible trend since 3G network. With the development of dense and ultra dense small-cell networks, the excessive power consumption has become a major issue of network operation. For example, BSs account for two-thirds of the total energy consumption in a wireless network [1,2]. Even if there is no or little traffic load, a BS can still consume more than 90% of their peak energy. The ultra-dense deployment of small-cell BSs (SBSs) leads to a low energy utilization in most time [3].
In [1, Page 2]: “ We propose a novel cellular traffic prediction method, where a multi-graph convolutional network (MGCN) is used to extract the spatial features, and a multi-channel long short-term memory (LSTM) network is adopted to efficiently capture the major frequency components in the traffic data. In addition, an attention scheme is designed to distinguish the significance of traffic sequences at different time”
- based on the determination, selecting, from available forecast models, either both of linear and non-linear forecast models or non-linear forecast models as a plurality of candidate base forecast models
In [1, Page 2]: “Based on the predicted traffic demand and the learned MBS and SBS capacity models, the optimal numbers of active MBSs and SBSs are obtained to minimize the power consumption in a given area. Due to the intractable expression of the capability functions obtained by the neural network, we have used both linear and non-linear fitting models to derive the optimal strategies without the high complexity of exhaustive searching”,
In [1, Page 2]: “ a novel cellular traffic prediction method, where a multi-graph convolutional network (MGCN) is used to extract the spatial features, and a multi-channel long short-term memory (LSTM) network is adopted to efficiently capture the major frequency components in the traffic data. In addition, an attention scheme is designed to distinguish the significance of traffic sequences at different time”,
In [II, Page 2]: “Many works on traffic forecasting are based on neural networks”,
In [II, Page 2]: “the differential evolution algorithm was combined with the back propagation algorithm to optimize the fuzzy neural network forecasting network traffic. A deep belief network (DBN) with Gaussian models was proposed in [13] to forecast the traffic demand in wireless mesh networks.
In [II, Page 2]: “ autoencoders were utilized to extract the spatial features in [15], then LSTM was adopted after to perform the traffic prediction. In [16–18], convolution neural network (CNN) and LSTM were combined to construct a spatial temporal neural network for traffic prediction, where the CNN concentrates on the spatial features. In [19], the attention mechanism was introduced into the Conv-LSTM network
[BRI: training a separate CNN–LSTM and linear/nonlinear fitting models, then combine their forecasts does represent an ensembled forecasting model.]
- collecting resource usage data associated with the wired or wireless tower
in [3, Page 4]: “ the wire-less traffic and the deployment of MBSs and SBSs in an area highly depend on the spatial/region information, the propagation environment and the population distribution. Hence, to analyze the characteristics of an area, several related datasets are introduced in this work: POIs, social activities, and the number of BSs deployed [20]. All these dataset are transformed to represent the 10000 square grids like the cellular traffic dataset”
In [IV, Page 5]:
PNG
media_image2.png
207
1082
media_image2.png
Greyscale
In [Abstract, Page 1]: “ we propose a novel data-driven intelligent BS sleeping mechanism based on a traffic prediction model that measures the BSs’ capacity in different region,
In [Abstract, Page 1]: “ the proposed BS sleeping strategy using an approximated non-linear model of capacity function achieves a near-optimal energy-saving performance with relatively low complexity”
- forecasting a resource need associated with the wired or wireless tower using the ensemble model, wherein the forecasted resource need is for allocating a resource to the wired or wireless tower;
In [2, Page 2]: “ Various approaches have been proposed to reduce the energy consumption of a mobile cellular network, which can generally be divided into the two categories: 1) improving the energy efficiency of the network components, and 2) turning off some of the network resources selectively. The BS sleeping strategy is a typical approach in the second category. Deactivating some deployed BSs during off-peak hours, not only the energy consumption but also the inter-cell interference in the network can be reduced.
In [V, Page 7]: “ In this section, MBSs’ and SBSs’ capacity functions,
C
m
(
·) and
C
s
(·), are modeled, to reveal the maximum traffic volume (i.e., throughput) that a number of MBSs or SBSs can provide in an area with a characteristic vector r. Different from the theoretical capacity function, such as Shannon Theorem, in this section, the neural network is utilized to construct the capacity model.
PNG
media_image3.png
581
672
media_image3.png
Greyscale
In [V, Page 7]: “ With these results, we can clear see the impact of the environment characteristics and the number of BSs on wireless maximum traffic, which verifies our modeling of the capacity functions
C
m
(
n
m
,r) and
C
s
(
n
s
,r). That is, the capacity is highly dependent on r. Since the number of BSs deployed in a square grid determines the provided wireless resource,
In [V, Page 8]: “ we choose it as one parameter of the capacity function [35]. Meanwhile, the geographic environment greatly impacts the propagation model of wireless signals [36]. Hence, the area characteristic vector is chosen as another parameter of the capacity function.
- and adding the resource usage data to the historic time series data to form a dynamic self- adaptation loop with respect to the wired or wireless tower.
In [1, Page 1]: the BS sleeping strategy aiming at reducing the base operating power of BSs, becomes an attractive method to save energy consumption. Various strategies to enable dynamic on-off for BSs have been proposed in these years,
In [1, Page 1]: “ the BS sleeping mode was designed according to a deterministic traffic changes over time,
In [1, Page 1]: “ the BS sleeping mode was designed according to a deterministic traffic changes over time. A similar idea was proposed in [5], where some BSs are shut down during the night time. All of these works utilize the periodic traffic fluctuation to save energy by switching some BSs to sleep mode.
In [VII, Page 11]: “ In this section, extensive experiments are conducted to evaluate the performance of our cellular traffic prediction model and the proposed dynamic BS sleeping strategy. For MBS and SBS, the parameters are adopted according to [40, 44]: 1) the fixed operating energy consumption: 10W and 5W; 2) the number of antennas: 3 and 1, and the circuit components energy consumption of antennas: 1W and 0.8W; 3) power amplifier efficiency: 0.128 and 0.12, and input power of BS: 43dBm and 33dBm; 4) the efficiency of load consumption: 1.15 · 10−9J/bit and 1.05 · 10−9J/bit
A dynamic base station (BS) sleeping strategy can indeed be designed to adapt parameters based on fixed energy consumption, number of antennas, and other resource metrics, and to integrate these into a self-adapting loop using historic time series data. This approach is relevant for both wired and wireless towers, as it can be applied to any BS type with measurable resource usage.
Lin does not explicitly disclose:
- A machine readable and non-transitory medium having information recorded thereon, wherein the information, when read by the machine, causes the machine to perform the following steps
- generating an ensemble forecast model adjusted over time on-the-fly by the continuously collected historic time series data, by computing a cost associated with each of the selected base forecast models,
- determining a set of parameters to be used for generating the ensemble forecast models based on the costs associated respectively with the selected base forecast models,
- and creating the ensemble forecast model based on the selected base forecast models in accordance with the set of parameters;
However, Nara discloses:
- A machine readable and non-transitory medium having information recorded thereon, wherein the information, when read by the machine, causes the machine to perform the following steps
In [0008]: “Another aspect of the invention or elements thereof can be implemented in the form of an article of manufacture tangibly embodying computer readable instructions which, when implemented, cause a computer to carry out a plurality of method steps, as described herein”
In [0008]: “ the means can include hardware module(s) or a combination of hardware and software modules, wherein the software modules are stored in a tangible computer-readable storage medium (or multiple such media)”.
In [0048]: “ A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media
- generating an ensemble forecast model adjusted over time on-the-fly by the continuously collected historic time series data, by computing a cost associated with each of the selected base forecast models,
In [0013] “ As described herein, an aspect of the present invention includes generating and recommending forecasting models for ensembles. Given a collection of forecasting models (such as weather models, hydrology models, climate models, etc.), historically-related observations, current model forecasts, socio-economic cost models, and/or user constraints on forecast lead time and computational resources, at least one embodiment of the invention includes determining an optimal ensemble of models and corresponding model configurations to be run in a given situation”,
In [0006]: “ In one aspect of the present invention, techniques for generating an ensemble of forecasting models are provided",
In [0016]: “ the resource and cost estimation engine 116 provide input to the model and parameter selection engine 110, which also receives input in the form of budget information, computational resource information, and cost information 104"
and in [0031]: "In further reference to FIG. 1, and as detailed herein, the model and parameter selection engine 110 determines, based on… an ensemble of models and parameterizations that are suited for the given event, as well as the availability of budget and/or computational resources" [i.e., the model and parameter selection engine generates an ensemble forecast model based on computed costs associated with the forecasting models it receives]),
- determining a set of parameters to be used for generating the ensemble forecast models based on the costs associated respectively with the selected base forecast models,
In [0024]: “ The resource and cost estimation engine 116, as depicted in FIG. 1, can be used to decide on the feasibility of running a configuration based on user-specified constraints. Examples of user-specified constraints can include the available number of processors, total memory available, model output required in under two hours, etc. Based on historical data pertaining, for example, to computational run-times, memory requirements, and hardware resource information, at least one embodiment of the invention includes creating models for hardware configuration and/or computation time/memory cost for a particular model configuration… This can include categorical features corresponding to various parameters such as physics options, cumulus parameterization, etc., or valued features such as number of grid points, time-step used, clock frequency of the processors, etc”,
and [0026]: “ The model and parameter selection engine 110 determines the final set of model configurations for a particular event. This selection depends on multiple factors such as socio-economic cost associated with the event and the lead times required for an accurate forecast, as well as the user-specified constraints on cost and computational resource availability. As noted, the resource and cost estimation engine 116 computes the computational time and/or memory and resource requirements for each model configuration" [i.e., a set of model configuration parameters associated with the constraints on operating costs and resources are determined by the resource and cost estimation engine, which feeds the parameters into the model and parameter selection engine, for generating the ensemble forecast models”.
- and creating the ensemble forecast model based on the selected base forecast models in accordance with the set of parameters;
In [0031]: “ In further reference to FIG. 1, and as detailed herein, the model and parameter selection engine 110 determines, based on… the estimated resource requirement and cost (from the resource and cost estimation engine 116), an ensemble of models and parameterizations that are suited for the given event, as well as the availability of budget and/or computational resources" [i.e., the parameters from the resource and cost estimation engine are fed into the model and the parameter selection engine to generate an ensemble forecast model”,
and in [0035]: “ Step 208 includes determining an ensemble of one or more of the multiple forecasting models to apply to the given environmental event based on (i) said estimated accuracy value for each of the multiple forecasting models, and (ii) said cost and said one or more resource requirements for each of the multiple forecasting models" [i.e., forecasting models are used in the selection/generation of the ensemble forecast model”.
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Lin and Nara.
Lin teaches a forecasting model that collects time series data from a wireless tower (base stations) and combines linear and non-linear forecast models.
Nara teaches ensembled forecasting model.
One of ordinary skill would have motivation to combine Lin and Nara that collects a forecasting model that includes a weather model within the context of historically related observatories and determines an optimal ensemble model (Nara [0013]).
In regard to claim 10: (Currently Amended):
Lin discloses:
- if the historic time series data exhibits seasonality, selecting one or more non-linear forecast base models from the available forecast models as the plurality of candidate base forecast models;
In [I, Page 1]: “ city CBD is full of tall buildings while the rural area is usually scattered with low-density houses. Therefore, an accurate capacity model of different types of BSs considering various region characteristics should be constructed”,
In [VII, Page 13]: “ compared with the energy consumption with all active MBSs and SBSs as 100%, the energy saved by the BS sleeping strategy using non-linear model is more obvious in city CBD. In more detail, the energy saved in city CBD approaches 63% in one week, and near 54% of the energy is saved in suburban regions. This is because the large number of BSs deployed and the traffic disparity between the peak and off-peak time”
- and if the historic time series data does not exhibit seasonality, selecting one or more non- linear forecast base models and one or more linear forecast base models from the available forecast models as the plurality of candidate base forecast models
In [VI, Page 9]: “ In this section, the optimal BS sleeping strategy is inves tigated which aims to minimize the energy consumption in a given area.
In [VI, Page 9]: “ In this work, we aim to develop a data-driven BS sleeping strategy based on public data with limited general cellular networks [39], e.g., overall traffic and BSs’s number, therefore, we define a cellular energy saving problem to seek the optimal BSs’ number in an area. According to [40], the energy consumption of a BS mainly includes the following parts: 1) the fixed operating energy consumption of a BS, 2) the energy consumption of the circuit components in BS’ antennas which is proportional to the number of antennas, 3) the energy consumption of the power amplifier, and 4) the energy consumption on traffic load which is linear with the traffic provided.
In [VI, Page 9]: “ the network throughput is usually assumed to increase linearly with respect to the number of deployed BSs, especially in sparse scenarios where the aggregated interference is relatively small [41,42].
C
m
(
n
m
,r) and
C
s
(
n
s
,r).
In [VI, Page 9]: “ In Fig. 13, it can be seen that, for a given square grid with characteristics r, 1)
C
s
(
n
s
,r)almost increases linearly with
n
s
in the investigated density range, and 2)
C
m
(
n
m
,r) increase”,
In [VI, Page 10]: “linearly when the density of MBSs is less than 50/km2, i.e., 3 MBSs in a square grid. Inspired by these results, linear capacity model can be used to approximately fit the results of
C
s
(
n
s
,r) and
C
m
(
n
m
,r)
In [VI, Page 10]: “ Apparently, the rapidly increasing aggregated inter-cell interference in the network densification will suppress the growth of the network throughput, such kind of results have been verified by many theoretical analysis on dense and ultra dense networks [43]. Even with various kinds of interference coordination and cancelation techniques, the linear increase of network capacity cannot be achieved in the dense networks
In [VI, Page 10]:
To better approach the real ca pacity achieved by the datasets, a non-linear function to fit the outcome of
C
m
(
n
m
,such as a1n2 m+b1nm+c1. For SBSs, due to the density range in the dataset, this performance degradation does not appear in the outcome. Since the linear model works well and leads to a smallest fitting error, the linear model is still preferred for
C
s
(
n
s
,r).
[BRI: in less dense area (rural) seasonality based optimization may not be required as may be represented by the historic time series data, linear and non-linear models can be combined. For dense regions (urban with optimization based on seasonality) due to small fitting errors, a linear model may be sufficient)
In regard to claim 14: (Currently Amended):
Lin does not explicitly disclose:
- wherein the ensemble forecast model corresponds to a weighted sum of the plurality of base forecast models
- wherein the set of parameters correspond to weights to be applied to the respective base forecast models
However, Nara discloses:
- wherein the ensemble forecast model corresponds to a weighted sum of the plurality of base forecast models
In [0028]: “Parameterization, as used herein, includes weights and/or reliabilities associated with different models, which can be used to combine forecasts from multiple models" [i.e., the ensemble forecast is formed by combining multiple base models according to specific weights, effectively forming a weighted sum]),
- wherein the set of parameters correspond to weights to be applied to the respective base forecast models
In [0035]: “Step 208 includes determining an ensemble of one or more of the multiple forecasting models… The determining step can include determining an ensemble of one or more of the multiple forecasting models and a parameterization for each of the one or more of the multiple forecasting models to apply to the given environmental event" [i.e., the ensemble is formed by selecting specific base forecasting models and separate weights are applied to each selected model's forecast output to combine them into a single ensemble prediction”.
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Lin and Nara.
Lin teaches a forecasting model that collects time series data from a wireless tower (base stations) and combines linear and non-linear forecast models.
Nara teaches ensembled forecasting model.
One of ordinary skill would have motivation to combine Lin and Nara that collects a forecasting model that includes a weather model within the context of historically related observatories and determines an optimal ensemble model (Nara [0013]).
In regard to claim 15: (Currently Amended)
Lin discloses:
- determining whether historic time series continuously collected from a wired or wireless tower includes seasonality;
In [1, Page 1]: “ In [4], the BS sleeping mode was designed according to a deterministic traffic changes over time. A similar idea was proposed in [5], where some BSs are shut down during the night time. All of these works utilize the periodic traffic fluctuation to save energy by switching some BSs to sleep mode. In Fig. 1, we plot the wireless traffic of Milan city at 10am and 10pm. From the figure, we can observe a significant traffic disparity between different regions for a given time instance. Meanwhile, comparing the traffic at 10am and that at 10pm, we can find that some regions tend to have traffic peaks in the day time, while some regions behave in a opposite way. Hence, in [6–9], the authors propose to selectively shut down BS resources in the network during off-peak hours based on the traffic monitor”
PNG
media_image1.png
298
526
media_image1.png
Greyscale
in [1, Page 1]: “ the accurate capacity model of BSs is another key factor of sleeping strategy. It is known that macro-cell BSs (MBSs) aim at providing an umbrella coverage while small-cell BSs (SBSs) target at throughput enhancement in traffic hotspots.
in [1, Page 1]: “ Network operators are deploying a large number of base stations (BSs) to meet the traffic requirements in peak hours. Hence, network densification has become an irresistible trend since 3G network. With the development of dense and ultra dense small-cell networks, the excessive power consumption has become a major issue of network operation. For example, BSs account for two-thirds of the total energy consumption in a wireless network [1,2]. Even if there is no or little traffic load, a BS can still consume more than 90% of their peak energy. The ultra-dense deployment of small-cell BSs (SBSs) leads to a low energy utilization in most time [3].
In [1, Page 2]: “ We propose a novel cellular traffic prediction method, where a multi-graph convolutional network (MGCN) is used to extract the spatial features, and a multi-channel long short-term memory (LSTM) network is adopted to efficiently capture the major frequency components in the traffic data. In addition, an attention scheme is designed to distinguish the significance of traffic sequences at different time”
- based on the determination, selecting, from available forecast models, either both of linear and non-linear forecast models or non-linear forecast models as a plurality of candidate base forecast models
In [1, Page 2]: “Based on the predicted traffic demand and the learned MBS and SBS capacity models, the optimal numbers of active MBSs and SBSs are obtained to minimize the power consumption in a given area. Due to the intractable expression of the capability functions obtained by the neural network, we have used both linear and non-linear fitting models to derive the optimal strategies without the high complexity of exhaustive searching”,
In [1, Page 2]: “ a novel cellular traffic prediction method, where a multi-graph convolutional network (MGCN) is used to extract the spatial features, and a multi-channel long short-term memory (LSTM) network is adopted to efficiently capture the major frequency components in the traffic data. In addition, an attention scheme is designed to distinguish the significance of traffic sequences at different time”,
In [II, Page 2]: “Many works on traffic forecasting are based on neural networks”,
In [II, Page 2]: “the differential evolution algorithm was combined with the back propagation algorithm to optimize the fuzzy neural network forecasting network traffic. A deep belief network (DBN) with Gaussian models was proposed in [13] to forecast the traffic demand in wireless mesh networks.
In [II, Page 2]: “ autoencoders were utilized to extract the spatial features in [15], then LSTM was adopted after to perform the traffic prediction. In [16–18], convolution neural network (CNN) and LSTM were combined to construct a spatial temporal neural network for traffic prediction, where the CNN concentrates on the spatial features. In [19], the attention mechanism was introduced into the Conv-LSTM network
[BRI: training a separate CNN–LSTM and linear/nonlinear fitting models, then combine their forecasts does represent an ensembled forecasting model.]
- collecting resource usage data associated with the wired or wireless tower
in [3, Page 4]: “ the wire-less traffic and the deployment of MBSs and SBSs in an area highly depend on the spatial/region information, the propagation environment and the population distribution. Hence, to analyze the characteristics of an area, several related datasets are introduced in this work: POIs, social activities, and the number of BSs deployed [20]. All these dataset are transformed to represent the 10000 square grids like the cellular traffic dataset”
In [IV, Page 5]:
PNG
media_image2.png
207
1082
media_image2.png
Greyscale
In [Abstract, Page 1]: “ we propose a novel data-driven intelligent BS sleeping mechanism based on a traffic prediction model that measures the BSs’ capacity in different region,
In [Abstract, Page 1]: “ the proposed BS sleeping strategy using an approximated non-linear model of capacity function achieves a near-optimal energy-saving performance with relatively low complexity”
- forecasting a resource need associated with the wired or wireless tower using the ensemble model, wherein the forecasted resource need is for allocating a resource to the wired or wireless tower;
In [2, Page 2]: “ Various approaches have been proposed to reduce the energy consumption of a mobile cellular network, which can generally be divided into the two categories: 1) improving the energy efficiency of the network components, and 2) turning off some of the network resources selectively. The BS sleeping strategy is a typical approach in the second category. Deactivating some deployed BSs during off-peak hours, not only the energy consumption but also the inter-cell interference in the network can be reduced.
In [V, Page 7]: “ In this section, MBSs’ and SBSs’ capacity functions,
C
m
(
·) and
C
s
(·), are modeled, to reveal the maximum traffic volume (i.e., throughput) that a number of MBSs or SBSs can provide in an area with a characteristic vector r. Different from the theoretical capacity function, such as Shannon Theorem, in this section, the neural network is utilized to construct the capacity model.
PNG
media_image3.png
581
672
media_image3.png
Greyscale
In [V, Page 7]: “ With these results, we can clear see the impact of the environment characteristics and the number of BSs on wireless maximum traffic, which verifies our modeling of the capacity functions
C
m
(
n
m
,r) and
C
s
(
n
s
,r). That is, the capacity is highly dependent on r. Since the number of BSs deployed in a square grid determines the provided wireless resource,
In [V, Page 8]: “ we choose it as one parameter of the capacity function [35]. Meanwhile, the geographic environment greatly impacts the propagation model of wireless signals [36]. Hence, the area characteristic vector is chosen as another parameter of the capacity function.
- and adding the resource usage data to the historic time series data to form a dynamic self- adaptation loop with respect to the wired or wireless tower.
In [1, Page 1]: the BS sleeping strategy aiming at reducing the base operating power of BSs, becomes an attractive method to save energy consumption. Various strategies to enable dynamic on-off for BSs have been proposed in these years,
In [1, Page 1]: “ the BS sleeping mode was designed according to a deterministic traffic changes over time,
In [1, Page 1]: “ the BS sleeping mode was designed according to a deterministic traffic changes over time. A similar idea was proposed in [5], where some BSs are shut down during the night time. All of these works utilize the periodic traffic fluctuation to save energy by switching some BSs to sleep mode.
In [VII, Page 11]: “ In this section, extensive experiments are conducted to evaluate the performance of our cellular traffic prediction model and the proposed dynamic BS sleeping strategy. For MBS and SBS, the parameters are adopted according to [40, 44]: 1) the fixed operating energy consumption: 10W and 5W; 2) the number of antennas: 3 and 1, and the circuit components energy consumption of antennas: 1W and 0.8W; 3) power amplifier efficiency: 0.128 and 0.12, and input power of BS: 43dBm and 33dBm; 4) the efficiency of load consumption: 1.15 · 10−9J/bit and 1.05 · 10−9J/bit
A dynamic base station (BS) sleeping strategy can indeed be designed to adapt parameters based on fixed energy consumption, number of antennas, and other resource metrics, and to integrate these into a self-adapting loop using historic time series data. This approach is relevant for both wired and wireless towers, as it can be applied to any BS type with measurable resource usage.
Lin does not explicitly disclose:
- A system, comprising: a data preprocessor implemented by a processor and configured for
- a performance based model selector implemented by a processor and configured for
- an ensemble model based forecaster implemented by a processor and configured for
- an integrated model ensemble unit implemented by a processor and configured for generating an ensemble forecast model adjusted over time on-the-fly by the continuously collected historic time series data, by computing a cost associated with each of the selected base forecast models,
- determining a set of parameters to be used for generating the ensemble forecast models based on the costs associated respectively with the selected base forecast models,
- and creating the ensemble forecast model based on the selected base forecast models in accordance with the set of parameters;
However, Nara discloses:
- A system, comprising: a data preprocessor implemented by a processor and configured for
In [0006]: “ In one aspect of the present invention, techniques for generating an ensemble of forecasting models are provided. An exemplary computer-implemented method can include steps of identifying a given environmental event from multiple items of input data”,
In [0008]: “ Furthermore, another aspect of the invention or elements thereof can be implemented in the form of an apparatus including a memory and at least one processor that is coupled to the memory and configured to perform noted method steps”,
- a performance based model selector implemented by a processor and configured for
In [0024]: “ Furthermore, another aspect of the invention or elements thereof can be implemented in the form of an apparatus including a memory and at least one processor that is coupled to the memory and configured to perform noted method steps”,
In [0006]: “Identifying a given environmental event from multiple items of input data; estimating an accuracy value for each of multiple forecasting models applied to an environmental event related to the given environmental event based on historical data”,
- an ensemble model based forecaster implemented by a processor and configured for
In [0006]: “ multiple forecasting models applied to an environmental event related to the given environmental event based on historical data; computing a cost and one or more resource requirements for each of the multiple forecasting models; and determining an ensemble of one or more of the multiple forecasting models to apply to the given environmental event [ Note: See [0008] for supporting by the processor]
- an integrated model ensemble unit implemented by a processor and configured for generating an ensemble forecast model adjusted over time on-the-fly by the continuously collected historic time series data, by computing a cost associated with each of the selected base forecast models,
In [0006]: “ multiple forecasting models applied to an environmental event related to the given environmental event based on historical data; computing a cost and one or more resource requirements for each of the multiple forecasting models; and determining an ensemble of one or more of the multiple forecasting models to apply to the given environmental event”,
In [0008]: “ Furthermore, another aspect of the invention or elements thereof can be implemented in the form of an apparatus including a memory and at least one processor that is coupled to the memory and configured to perform noted method steps”,
[ BRI: an apparatus can support as an integrated ensemble unit with the context of processor of [0008]]
In [0013] “ As described herein, an aspect of the present invention includes generating and recommending forecasting models for ensembles. Given a collection of forecasting models (such as weather models, hydrology models, climate models, etc.), historically-related observations, current model forecasts, socio-economic cost models, and/or user constraints on forecast lead time and computational resources, at least one embodiment of the invention includes determining an optimal ensemble of models and corresponding model configurations to be run in a given situation”,
In [0006]: “ In one aspect of the present invention, techniques for generating an ensemble of forecasting models are provided",
In [0016]: “ the resource and cost estimation engine 116 provide input to the model and parameter selection engine 110, which also receives input in the form of budget information, computational resource information, and cost information 104"
and in [0031]: "In further reference to FIG. 1, and as detailed herein, the model and parameter selection engine 110 determines, based on… an ensemble of models and parameterizations that are suited for the given event, as well as the availability of budget and/or computational resources" [i.e., the model and parameter selection engine generates an ensemble forecast model based on computed costs associated with the forecasting models it receives]),
- determining a set of parameters to be used for generating the ensemble forecast models based on the costs associated respectively with the selected base forecast models,
In [0024]: “ The resource and cost estimation engine 116, as depicted in FIG. 1, can be used to decide on the feasibility of running a configuration based on user-specified constraints. Examples of user-specified constraints can include the available number of processors, total memory available, model output required in under two hours, etc. Based on historical data pertaining, for example, to computational run-times, memory requirements, and hardware resource information, at least one embodiment of the invention includes creating models for hardware configuration and/or computation time/memory cost for a particular model configuration… This can include categorical features corresponding to various parameters such as physics options, cumulus parameterization, etc., or valued features such as number of grid points, time-step used, clock frequency of the processors, etc”,
and [0026]: “ The model and parameter selection engine 110 determines the final set of model configurations for a particular event. This selection depends on multiple factors such as socio-economic cost associated with the event and the lead times required for an accurate forecast, as well as the user-specified constraints on cost and computational resource availability. As noted, the resource and cost estimation engine 116 computes the computational time and/or memory and resource requirements for each model configuration" [i.e., a set of model configuration parameters associated with the constraints on operating costs and resources are determined by the resource and cost estimation engine, which feeds the parameters into the model and parameter selection engine, for generating the ensemble forecast models”.
- and creating the ensemble forecast model based on the selected base forecast models in accordance with the set of parameters;
In [0031]: “ In further reference to FIG. 1, and as detailed herein, the model and parameter selection engine 110 determines, based on… the estimated resource requirement and cost (from the resource and cost estimation engine 116), an ensemble of models and parameterizations that are suited for the given event, as well as the availability of budget and/or computational resources" [i.e., the parameters from the resource and cost estimation engine are fed into the model and the parameter selection engine to generate an ensemble forecast model”,
and in [0035]: “ Step 208 includes determining an ensemble of one or more of the multiple forecasting models to apply to the given environmental event based on (i) said estimated accuracy value for each of the multiple forecasting models, and (ii) said cost and said one or more resource requirements for each of the multiple forecasting models" [i.e., forecasting models are used in the selection/generation of the ensemble forecast model”.
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Lin and Nara.
Lin teaches a forecasting model that collects time series data from a wireless tower (base stations) and combines linear and non-linear forecast models.
Nara teaches ensembled forecasting model.
One of ordinary skill would have motivation to combine Lin and Nara that collects a forecasting model that includes a weather model within the context of historically related observatories and determines an optimal ensemble model (Nara [0013]).
In regard to claim 16: (Currently Amended):
Lin discloses:
- if the historic time series data exhibits seasonality, selecting one or more non-linear forecast base models from the available forecast models as the plurality of candidate base forecast models;
In [I, Page 1]: “ city CBD is full of tall buildings while the rural area is usually scattered with low-density houses. Therefore, an accurate capacity model of different types of BSs considering various region characteristics should be constructed”,
In [VII, Page 13]: “ compared with the energy consumption with all active MBSs and SBSs as 100%, the energy saved by the BS sleeping strategy using non-linear model is more obvious in city CBD. In more detail, the energy saved in city CBD approaches 63% in one week, and near 54% of the energy is saved in suburban regions. This is because the large number of BSs deployed and the traffic disparity between the peak and off-peak time”
- and if the historic time series data does not exhibit seasonality, selecting one or more non- linear forecast base models and one or more linear forecast base models from the available forecast models as the plurality of candidate base forecast models
In [VI, Page 9]: “ In this section, the optimal BS sleeping strategy is inves tigated which aims to minimize the energy consumption in a given area.
In [VI, Page 9]: “ In this work, we aim to develop a data-driven BS sleeping strategy based on public data with limited general cellular networks [39], e.g., overall traffic and BSs’s number, therefore, we define a cellular energy saving problem to seek the optimal BSs’ number in an area. According to [40], the energy consumption of a BS mainly includes the following parts: 1) the fixed operating energy consumption of a BS, 2) the energy consumption of the circuit components in BS’ antennas which is proportional to the number of antennas, 3) the energy consumption of the power amplifier, and 4) the energy consumption on traffic load which is linear with the traffic provided.
In [VI, Page 9]: “ the network throughput is usually assumed to increase linearly with respect to the number of deployed BSs, especially in sparse scenarios where the aggregated interference is relatively small [41,42].
C
m
(
n
m
,r) and
C
s
(
n
s
,r).
In [VI, Page 9]: “ In Fig. 13, it can be seen that, for a given square grid with characteristics r, 1)
C
s
(
n
s
,r)almost increases linearly with
n
s
in the investigated density range, and 2)
C
m
(
n
m
,r) increase”,
In [VI, Page 10]: “linearly when the density of MBSs is less than 50/km2, i.e., 3 MBSs in a square grid. Inspired by these results, linear capacity model can be used to approximately fit the results of
C
s
(
n
s
,r) and
C
m
(
n
m
,r)
In [VI, Page 10]: “ Apparently, the rapidly increasing aggregated inter-cell interference in the network densification will suppress the growth of the network throughput, such kind of results have been verified by many theoretical analysis on dense and ultra dense networks [43]. Even with various kinds of interference coordination and cancelation techniques, the linear increase of network capacity cannot be achieved in the dense networks
In [VI, Page 10]:
To better approach the real ca pacity achieved by the datasets, a non-linear function to fit the outcome of
C
m
(
n
m
,such as a1n2 m+b1nm+c1. For SBSs, due to the density range in the dataset, this performance degradation does not appear in the outcome. Since the linear model works well and leads to a smallest fitting error, the linear model is still preferred for
C
s
(
n
s
,r).
[BRI: in less dense area (rural) seasonality based optimization may not be required as may be represented by the historic time series data, linear and non-linear models can be combined. For dense regions (urban with optimization based on seasonality) due to small fitting errors, a linear model may be sufficient)
In regard to claim 20: (Currently Amended):
Lin does not explicitly disclose:
- wherein the ensemble forecast model corresponds to a weighted sum of the plurality of base forecast models
- wherein the set of parameters correspond to weights to be applied to the respective base forecast models
However, Nara discloses:
- wherein the ensemble forecast model corresponds to a weighted sum of the plurality of base forecast models
In [0028]: “Parameterization, as used herein, includes weights and/or reliabilities associated with different models, which can be used to combine forecasts from multiple models" [i.e., the ensemble forecast is formed by combining multiple base models according to specific weights, effectively forming a weighted sum]),
- wherein the set of parameters correspond to weights to be applied to the respective base forecast models
In [0035]: “Step 208 includes determining an ensemble of one or more of the multiple forecasting models… The determining step can include determining an ensemble of one or more of the multiple forecasting models and a parameterization for each of the one or more of the multiple forecasting models to apply to the given environmental event" [i.e., the ensemble is formed by selecting specific base forecasting models and separate weights are applied to each selected model's forecast output to combine them into a single ensemble prediction”.
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Lin and Nara.
Lin teaches a forecasting model that collects time series data from a wireless tower (base stations) and combines linear and non-linear forecast models.
Nara teaches ensembled forecasting model.
One of ordinary skill would have motivation to combine Lin and Nara that collects a forecasting model that includes a weather model within the context of historically related observatories and determines an optimal ensemble model (Nara [0013]).
Claims 4-5, 11-12 and 17-18 are rejected under 35 U.S.C. 103 as being unpatentable over
Jiansheng Lin, et al (hereinafter Lin) A data-driven base station sleeping strategy based on traffic prediction." IEEE Transactions on Network Science and Engineering 11.6 (2021): 5627-5643.
Balakrishnan Narayanaswamy (hereinafter Nara ) US 20150253463 A1;
and further in view of Xi Cheng (hereinafter Cheng) US 20210357402 A1;
In regard to claim 4: (Currently Amended):
Lin and Nara do not explicitly disclose:
- performing linear regression on smoothed historic time series data to generate linear regression result;
- generating detrended historic time series data based on the smoothed historic time series data and the linear regression result;
- performing auto-correlation on the smoothed time series data and the detrended historic time series data to generate auto-correlation results;
- and determining whether the historic time series data exhibits seasonality based on the auto-correlation results.
However, Cheng discloses:
- performing linear regression on smoothed historic time series data to generate linear regression result
in [0034]: “The preprocess stage 510 receives an input time series 502 and performs data frequency handling 512, NULL imputation 514 (i.e., determining and/or rejecting any nulls in the input time series 502), holiday effect modeling 516, and anomaly detection 518” [i.e., the preprocessing operations provide a cleaned and smoothed input time series] and paragraph [0036]: “After preprocessing, the training stage 520 begins with a seasonal and trend decomposition using local regression (STL) module 522 which generates a de-seasoned component 523 a and a seasonal component 523 b” [i.e., the STL module uses local regression to smooth the input time series, producing a de-seasoned (smoothed historic time series data) component as it represents the non-seasonal variations in the data], "The STL module 522 estimates nonlinear relationships and decomposes a time series into multiple components 412 a-c (FIG. 4)" [i.e., the de-seasoned component from the STL module serves as a regression estimate of the smoothed series”,
- generating detrended historic time series data based on the smoothed historic time series data and the linear regression result
in [0036]: “After preprocessing, the training stage 520 begins with a seasonal and trend decomposition using local regression (STL) module 522 which generates a de-seasoned component 523” [i.e., the STL module produces a smoothed/de-seasoned component 523a which inherently generates detrended historic time series data”,
- performing auto-correlation on the smoothed time series data and the detrended historic time series data to generate auto-correlation results
In [0024]: “The model trainer 210 may generate and/or train each model 212 with different parameters, for example, the model trainer 210 may generate and train a plurality of autoregressive integrated moving average (ARIMA) models with different orders of the autoregressive models (i.e., the number of time lags and commonly represented as the parameter p)” [i.e., the autoregressive order parameter p depends on the auto-correlation of input data, and is used in encoding correlation patterns in the time series data] and paragraph [0036]: “The de-seasoned component 523 a is processed via a Kwtatkowsi-Phillips-Schmidt-Shin (KPSS) test module 524 and generates a plural of ARIMA models in parallel at 526” [i.e., the de-seasoned (smoothed) component is processed via KPSS and used for generating ARIMA models in parallel; these steps inherently involve the auto-correlation calculations on the smoothed and de-trended data of the input time series (historic time series data) generating auto-correlation results”,
- and determining whether the historic time series data exhibits seasonality based on the auto-correlation results
in [0024]: “The model trainer 210 may generate and/or train each model 212 with different parameters, for example, the model trainer 210 may generate and train a plurality of autoregressive integrated moving average (ARIMA) models with different orders of the autoregressive models (i.e., the number of time lags and commonly represented as the parameter p)” [i.e., ARIMA's autoregressive parameter p and lag structure captures temporal correlations of de-seasoned series], paragraph [0025]: “As discussed in more detail below, the model trainer 210 may generate and train forecasting models 212 capable of modeling many different aspects of time series. For example, the forecast models 212 may account for seasonal effects” and paragraph [0036]: “The de-seasoned component 523 a is processed via a Kwtatkowsi-Phillips-Schmidt-Shin (KPSS) test module 524 and generates a plural of ARIMA models in parallel at 526” [i.e., the ARIMA models are trained on the de-seasoned component and yields autocorrelation coefficients that quantify repeating/seasonal patterns in the input time series data, inherently determining whether the data exhibits seasonality]).
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Lin, Nara and Cheng.
Lin teaches a forecasting model that collects time series data from a wireless tower (base stations) and combines linear and non-linear forecast models.
Nara teaches ensembled forecasting model.
Cheng teaches auto-correlation on the smoothed historic time series data.
One of ordinary skill would have motivation to combine Lin, Nara and Cheng’s method in order to “accurately forecast future trends incorporating the time component (Cheng [0003]).
In regard to claim 5: (Original):
Lin and Nara do not explicitly disclose:
- wherein the auto-correlation results include: a first auto-correlation metric obtained via auto-correlation on the smoothed historic time series data; and a second auto-correlation metric obtained via auto-correlation on the detrended historic time series data
However, Cheng discloses:
- wherein the auto-correlation results include: a first auto-correlation metric obtained via auto-correlation on the smoothed historic time series data; and a second auto-correlation metric obtained via auto-correlation on the detrended historic time series data
In [0024]: “The model trainer 210 may generate and/or train each model 212 with different parameters, for example, the model trainer 210 may generate and train a plurality of autoregressive integrated moving average (ARIMA) models with different orders of the autoregressive models (i.e., the number of time lags and commonly represented as the parameter p)” [i.e., the lag-order (p) and resulting coefficients from ARIMA model generation encode the specific temporal correlation patterns that are identified in the de-seasoned/smoothed time series input data], paragraph [0034]: “Referring now to FIG. 5… The preprocess stage 510 receives an input time series 502 and performs data frequency handling 512, NULL imputation 514 (i.e., determining and/or rejecting any nulls in the input time series 502), holiday effect modeling 516, and anomaly detection 518” [i.e., the input time series undergoes a form of smoothing via NULL imputation and anomaly detection] and paragraph [0036]: “The de-seasoned component 523 a is processed via a Kwtatkowsi-Phillips-Schmidt-Shin (KPSS) test module 524 and generates a plural of ARIMA models in parallel at 526” [i.e., the generation of ARIMA models with a specific lag order produces corresponding auto-regressive coefficients which represent the strength of the temporal correlation identified at first and second time lags (first and second auto-correlation metrics) within the de-seasoned component (encompassing detrended & smoothed input time series data”.
In regard to claim 11: (Currently Amended):
Lin and Nara do not explicitly disclose:
- performing linear regression on smoothed historic time series data to generate linear regression result;
- generating detrended historic time series data based on the smoothed historic time series data and the linear regression result;
- performing auto-correlation on the smoothed time series data and the detrended historic time series data to generate auto-correlation results;
- and determining whether the historic time series data exhibits seasonality based on the auto-correlation results.
However, Cheng discloses:
- performing linear regression on smoothed historic time series data to generate linear regression result
in [0034]: “The preprocess stage 510 receives an input time series 502 and performs data frequency handling 512, NULL imputation 514 (i.e., determining and/or rejecting any nulls in the input time series 502), holiday effect modeling 516, and anomaly detection 518” [i.e., the preprocessing operations provide a cleaned and smoothed input time series] and paragraph [0036]: “After preprocessing, the training stage 520 begins with a seasonal and trend decomposition using local regression (STL) module 522 which generates a de-seasoned component 523 a and a seasonal component 523 b” [i.e., the STL module uses local regression to smooth the input time series, producing a de-seasoned (smoothed historic time series data) component as it represents the non-seasonal variations in the data], "The STL module 522 estimates nonlinear relationships and decomposes a time series into multiple components 412 a-c (FIG. 4)" [i.e., the de-seasoned component from the STL module serves as a regression estimate of the smoothed series”,
- generating detrended historic time series data based on the smoothed historic time series data and the linear regression result
in [0036]: “After preprocessing, the training stage 520 begins with a seasonal and trend decomposition using local regression (STL) module 522 which generates a de-seasoned component 523” [i.e., the STL module produces a smoothed/de-seasoned component 523a which inherently generates detrended historic time series data”,
- performing auto-correlation on the smoothed time series data and the detrended historic time series data to generate auto-correlation results
In [0024]: “The model trainer 210 may generate and/or train each model 212 with different parameters, for example, the model trainer 210 may generate and train a plurality of autoregressive integrated moving average (ARIMA) models with different orders of the autoregressive models (i.e., the number of time lags and commonly represented as the parameter p)” [i.e., the autoregressive order parameter p depends on the auto-correlation of input data, and is used in encoding correlation patterns in the time series data] and paragraph [0036]: “The de-seasoned component 523 a is processed via a Kwtatkowsi-Phillips-Schmidt-Shin (KPSS) test module 524 and generates a plural of ARIMA models in parallel at 526” [i.e., the de-seasoned (smoothed) component is processed via KPSS and used for generating ARIMA models in parallel; these steps inherently involve the auto-correlation calculations on the smoothed and de-trended data of the input time series (historic time series data) generating auto-correlation results”,
- and determining whether the historic time series data exhibits seasonality based on the auto-correlation results
in [0024]: “The model trainer 210 may generate and/or train each model 212 with different parameters, for example, the model trainer 210 may generate and train a plurality of autoregressive integrated moving average (ARIMA) models with different orders of the autoregressive models (i.e., the number of time lags and commonly represented as the parameter p)” [i.e., ARIMA's autoregressive parameter p and lag structure captures temporal correlations of de-seasoned series], paragraph [0025]: “As discussed in more detail below, the model trainer 210 may generate and train forecasting models 212 capable of modeling many different aspects of time series. For example, the forecast models 212 may account for seasonal effects” and paragraph [0036]: “The de-seasoned component 523 a is processed via a Kwtatkowsi-Phillips-Schmidt-Shin (KPSS) test module 524 and generates a plural of ARIMA models in parallel at 526” [i.e., the ARIMA models are trained on the de-seasoned component and yields autocorrelation coefficients that quantify repeating/seasonal patterns in the input time series data, inherently determining whether the data exhibits seasonality]).
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Lin, Nara and Cheng.
Lin teaches a forecasting model that collects time series data from a wireless tower (base stations) and combines linear and non-linear forecast models.
Nara teaches ensembled forecasting model.
Cheng teaches auto-correlation on the smoothed historic time series data.
One of ordinary skill would have motivation to combine Lin, Nara and Cheng’s method in order to “accurately forecast future trends incorporating the time component (Cheng [0003]).
In regard to claim 12: (Original):
Lin and Nara do not explicitly disclose:
- wherein the auto-correlation results include: a first auto-correlation metric obtained via auto-correlation on the smoothed historic time series data; and a second auto-correlation metric obtained via auto-correlation on the detrended historic time series data
However, Cheng discloses:
- wherein the auto-correlation results include: a first auto-correlation metric obtained via auto-correlation on the smoothed historic time series data; and a second auto-correlation metric obtained via auto-correlation on the detrended historic time series data
In [0024]: “The model trainer 210 may generate and/or train each model 212 with different parameters, for example, the model trainer 210 may generate and train a plurality of autoregressive integrated moving average (ARIMA) models with different orders of the autoregressive models (i.e., the number of time lags and commonly represented as the parameter p)” [i.e., the lag-order (p) and resulting coefficients from ARIMA model generation encode the specific temporal correlation patterns that are identified in the de-seasoned/smoothed time series input data], paragraph [0034]: “Referring now to FIG. 5… The preprocess stage 510 receives an input time series 502 and performs data frequency handling 512, NULL imputation 514 (i.e., determining and/or rejecting any nulls in the input time series 502), holiday effect modeling 516, and anomaly detection 518” [i.e., the input time series undergoes a form of smoothing via NULL imputation and anomaly detection] and paragraph [0036]: “The de-seasoned component 523 a is processed via a Kwtatkowsi-Phillips-Schmidt-Shin (KPSS) test module 524 and generates a plural of ARIMA models in parallel at 526” [i.e., the generation of ARIMA models with a specific lag order produces corresponding auto-regressive coefficients which represent the strength of the temporal correlation identified at first and second time lags (first and second auto-correlation metrics) within the de-seasoned component (encompassing detrended & smoothed input time series data”.
In regard to claim 17: (Currently Amended):
Lin and Nara do not explicitly disclose:
- performing linear regression on smoothed historic time series data to generate linear regression result;
- generating detrended historic time series data based on the smoothed historic time series data and the linear regression result;
- performing auto-correlation on the smoothed time series data and the detrended historic time series data to generate auto-correlation results;
- and determining whether the historic time series data exhibits seasonality based on the auto-correlation results.
However, Cheng discloses:
- performing linear regression on smoothed historic time series data to generate linear regression result
in [0034]: “The preprocess stage 510 receives an input time series 502 and performs data frequency handling 512, NULL imputation 514 (i.e., determining and/or rejecting any nulls in the input time series 502), holiday effect modeling 516, and anomaly detection 518” [i.e., the preprocessing operations provide a cleaned and smoothed input time series] and paragraph [0036]: “After preprocessing, the training stage 520 begins with a seasonal and trend decomposition using local regression (STL) module 522 which generates a de-seasoned component 523 a and a seasonal component 523 b” [i.e., the STL module uses local regression to smooth the input time series, producing a de-seasoned (smoothed historic time series data) component as it represents the non-seasonal variations in the data], "The STL module 522 estimates nonlinear relationships and decomposes a time series into multiple components 412 a-c (FIG. 4)" [i.e., the de-seasoned component from the STL module serves as a regression estimate of the smoothed series”,
- generating detrended historic time series data based on the smoothed historic time series data and the linear regression result
in [0036]: “After preprocessing, the training stage 520 begins with a seasonal and trend decomposition using local regression (STL) module 522 which generates a de-seasoned component 523” [i.e., the STL module produces a smoothed/de-seasoned component 523a which inherently generates detrended historic time series data”,
- performing auto-correlation on the smoothed time series data and the detrended historic time series data to generate auto-correlation results
In [0024]: “The model trainer 210 may generate and/or train each model 212 with different parameters, for example, the model trainer 210 may generate and train a plurality of autoregressive integrated moving average (ARIMA) models with different orders of the autoregressive models (i.e., the number of time lags and commonly represented as the parameter p)” [i.e., the autoregressive order parameter p depends on the auto-correlation of input data, and is used in encoding correlation patterns in the time series data] and paragraph [0036]: “The de-seasoned component 523 a is processed via a Kwtatkowsi-Phillips-Schmidt-Shin (KPSS) test module 524 and generates a plural of ARIMA models in parallel at 526” [i.e., the de-seasoned (smoothed) component is processed via KPSS and used for generating ARIMA models in parallel; these steps inherently involve the auto-correlation calculations on the smoothed and de-trended data of the input time series (historic time series data) generating auto-correlation results”,
- and determining whether the historic time series data exhibits seasonality based on the auto-correlation results
in [0024]: “The model trainer 210 may generate and/or train each model 212 with different parameters, for example, the model trainer 210 may generate and train a plurality of autoregressive integrated moving average (ARIMA) models with different orders of the autoregressive models (i.e., the number of time lags and commonly represented as the parameter p)” [i.e., ARIMA's autoregressive parameter p and lag structure captures temporal correlations of de-seasoned series], paragraph [0025]: “As discussed in more detail below, the model trainer 210 may generate and train forecasting models 212 capable of modeling many different aspects of time series. For example, the forecast models 212 may account for seasonal effects” and paragraph [0036]: “The de-seasoned component 523 a is processed via a Kwtatkowsi-Phillips-Schmidt-Shin (KPSS) test module 524 and generates a plural of ARIMA models in parallel at 526” [i.e., the ARIMA models are trained on the de-seasoned component and yields autocorrelation coefficients that quantify repeating/seasonal patterns in the input time series data, inherently determining whether the data exhibits seasonality]).
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Lin, Nara and Cheng.
Lin teaches a forecasting model that collects time series data from a wireless tower (base stations) and combines linear and non-linear forecast models.
Nara teaches ensembled forecasting model.
Cheng teaches auto-correlation on the smoothed historic time series data.
One of ordinary skill would have motivation to combine Lin, Nara and Cheng’s method in order to “accurately forecast future trends incorporating the time component (Cheng [0003]).
In regard to claim 18: (Original):
Lin and Nara do not explicitly disclose:
- wherein the auto-correlation results include: a first auto-correlation metric obtained via auto-correlation on the smoothed historic time series data; and a second auto-correlation metric obtained via auto-correlation on the detrended historic time series data
However, Cheng discloses:
- wherein the auto-correlation results include: a first auto-correlation metric obtained via auto-correlation on the smoothed historic time series data; and a second auto-correlation metric obtained via auto-correlation on the detrended historic time series data
In [0024]: “The model trainer 210 may generate and/or train each model 212 with different parameters, for example, the model trainer 210 may generate and train a plurality of autoregressive integrated moving average (ARIMA) models with different orders of the autoregressive models (i.e., the number of time lags and commonly represented as the parameter p)” [i.e., the lag-order (p) and resulting coefficients from ARIMA model generation encode the specific temporal correlation patterns that are identified in the de-seasoned/smoothed time series input data], paragraph [0034]: “Referring now to FIG. 5… The preprocess stage 510 receives an input time series 502 and performs data frequency handling 512, NULL imputation 514 (i.e., determining and/or rejecting any nulls in the input time series 502), holiday effect modeling 516, and anomaly detection 518” [i.e., the input time series undergoes a form of smoothing via NULL imputation and anomaly detection] and paragraph [0036]: “The de-seasoned component 523 a is processed via a Kwtatkowsi-Phillips-Schmidt-Shin (KPSS) test module 524 and generates a plural of ARIMA models in parallel at 526” [i.e., the generation of ARIMA models with a specific lag order produces corresponding auto-regressive coefficients which represent the strength of the temporal correlation identified at first and second time lags (first and second auto-correlation metrics) within the de-seasoned component (encompassing detrended & smoothed input time series data”.
Claims 6, 13, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over
Jiansheng Lin, et al (hereinafter Lin) A data-driven base station sleeping strategy based on traffic prediction." IEEE Transactions on Network Science and Engineering 11.6 (2021): 5627-5643,
Balakrishnan Narayanaswamy (hereinafter Nara ) US 20150253463 A1,
and further in view of Rawan ALKURD (hereinafter ALKURD) US 2021/0266781 A1.
In regard to claim 6: (Currently Amended):
Lin and Nara do not explicitly disclose:
- designating multiple candidate base forecast models from the available forecast models based on whether the historic time series data exhibits seasonality;
- and identifying the plurality of candidate base forecast models from the multiple candidate base forecast models based on forecast performance of each of the multiple candidate base forecast models, by with respect to each of the multiple candidate base forecast models,
- generating a forecast result based on the historic time series data using the candidate base forecast model, computing a measure indicative of the performance of the candidate base forecast model based on the forecast result;
- computing a measure indicative of the performance of the candidate base forecast model based on the forecast result
- and selecting the plurality of candidate base forecast models from the multiple candidate base forecast models based on the measures associated respectively with the multiple candidate base forecast models
However, ALKURD discloses:
- designating multiple candidate base forecast models from the available forecast models based on whether the historic time series data exhibits seasonality;
In [0314]: “ Train and validate the model: There are several ML algorithms that can be used to build a predictive model. From a practical point of view, the best predictor candidate for our proposed framework is a deep neural network (DNN) algorithm.
In [0281]:
we perform user satisfaction prediction using the following traditional set of ML algorithms: Decision Tree (DT), K-nearest neighbor (Knn), and Random Forest (RF).
In [0281]:
hyperparameter tuning is conducted using grid-search. FIG. 22 depicts the 10-folds cross-validation prediction accuracies for each ML algorithm. Since there are six satisfaction levels, the random choice accuracy level is 0.166. FIG. 22 shows that the best performance is achieved using RF (ensemble) algorithm with an average accuracy of 0.884 compared to 0.85 for DT and Knn. It is worth mentioning that the predictors' performance should be directly related to the number of relevant context variables available to the ML algorithms.
In [0484]:
Various ML models can be utilized to build surrogates, including linear models, support vector machines [69], and Gaussian processes [70]. The present invention utilizes a big data-driven framework to build ML-based surrogate models in order to predict user satisfaction in wireless networks.
In [0039] :” In another embodiment, the context comprises one or more of the parameters: time, day, location, speed, activity, service request arrival”,
In [0067]: “ FIGS. 19a and 19b present the probability of request arrival (i.e., P{N(t, t+δ)=1)) vs. t in hours for a weekday and a weekend day, respectively”,
In [0065]: “ FIG. 17 presents a graphic representation setting out the percentage of time the user spent at each location over each time period on a weekday”,
PNG
media_image4.png
366
592
media_image4.png
Greyscale
- and identifying the plurality of candidate base forecast models from the multiple candidate base forecast models based on forecast performance of each of the multiple candidate base forecast models, by with respect to each of the multiple candidate base forecast models,
In [0281]:
hyperparameter tuning is conducted using grid-search. FIG. 22 depicts the 10-folds cross-validation prediction accuracies for each ML algorithm. Since there are six satisfaction levels, the random choice accuracy level is 0.166. FIG. 22 shows that the best performance is achieved using RF (ensemble) algorithm with an average accuracy of 0.884 compared to 0.85 for DT and Knn. It is worth mentioning that the predictors' performance should be directly related to the number of relevant context variables available to the ML algorithms.
- generating a forecast result based on the historic time series data using the candidate base forecast model, computing a measure indicative of the performance of the candidate base forecast model based on the forecast result;
In [0281]:”hyperparameter tuning is conducted using grid-search. FIG. 22 depicts the 10-folds cross-validation prediction accuracies for each ML algorithm. Since there are six satisfaction levels, the random choice accuracy level is 0.166. FIG. 22 shows that the best performance is achieved using RF (ensemble) algorithm with an average accuracy of 0.884 compared to 0.85 for DT and Knn. It is worth mentioning that the predictors' performance should be directly related to the number of relevant context variables available to the ML algorithms”,
In [0280]: “ ML and data analytics are important to extract patterns and knowledge from historical data and use it to predict user satisfaction behavior in the future”,
In [0475]: “ surrogate models, which are comparatively faster and rely on historical data and user patterns to predict user satisfaction values in real-time, may be very useful in implementing personalized wireless networks.
In [0065]: “ FIG. 17 presents a graphic representation setting out the percentage of time the user spent at each location over each time period on a weekday”,
PNG
media_image4.png
366
592
media_image4.png
Greyscale
- computing a measure indicative of the performance of the candidate base forecast model based on the forecast result
In [0281]:”hyperparameter tuning is conducted using grid-search. FIG. 22 depicts the 10-folds cross-validation prediction accuracies for each ML algorithm. Since there are six satisfaction levels, the random choice accuracy level is 0.166. FIG. 22 shows that the best performance is achieved using RF (ensemble) algorithm with an average accuracy of 0.884 compared to 0.85 for DT and Knn. It is worth mentioning that the predictors' performance should be directly related to the number of relevant context variables available to the ML algorithms”,
[BRI: accuracy is the measure of performance]
- and selecting the plurality of candidate base forecast models from the multiple candidate base forecast models based on the measures associated respectively with the multiple candidate base forecast models
In [0281]:”hyperparameter tuning is conducted using grid-search. FIG. 22 depicts the 10-folds cross-validation prediction accuracies for each ML algorithm. Since there are six satisfaction levels, the random choice accuracy level is 0.166. FIG. 22 shows that the best performance is achieved using RF (ensemble) algorithm with an average accuracy of 0.884 compared to 0.85 for DT and Knn. It is worth mentioning that the predictors' performance should be directly related to the number of relevant context variables available to the ML algorithms”,
In [0447] : “ EAs are a class of metaheuristics population-based optimization algorithms, where multiple candidate solutions are maintained in parallel. EAs are designed based on the idea of the survival of the “fittest” solution in order to evolve a population that is a good approximation of a desired global optimum [62]. The fitness of an evolved solution is a measure of its quality at solving the problem.
In [0134]: “ Current networks are designed mostly to be a “Universal Fit,” where service providers deliver services with a quality appeal to all types of users. However, user expectations of service quality are not “One Size Fits All.”
[BRI: choosing more than one model from a set of possible forecasting methods to serve as the foundation for a combined or ensemble forecast means no when no single model consistently outperforms all others across different conditions, or when the goal is to improve robustness and stability by leveraging the strengths of multiple approaches. In the context of forecasting, the “survival of the fittest” approach can indeed be used to select the best-performing candidate base forecast models from a set of alternatives. This is conceptually similar to how Genetic Algorithms (GAs) or Evolutionary Algorithms (EAs) work: they maintain a population of candidate solutions (here, forecast models), evaluate each using a fitness measure (e.g., forecast accuracy, error metrics), and iteratively improve the population by selecting, combining, and mutating models
Spec [0016] “ In the field of advanced data analytics and econometrics, different time series forecasting models and methods have been developed and applied to fit different tasks or time series data with different characteristics. There is no “one size fits all” time series forecast model that may be deployed in any situation.
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Lin, Nara and ALKURD.
Lin teaches a forecasting model that collects time series data from a wireless tower (base stations) and combines linear and non-linear forecast models.
Nara teaches ensembled forecasting model.
ALKURD teaches candidate forecast models.
One of ordinary skill would have motivation to combine Lin, Nara and ALKURD that can re-learning to improve the predictive model performance and to update the model with user behavioral changes that could occur over time (ALKURD [0170]).
In regard to claim 13: (Currently Amended):
Lin and Nara do not explicitly disclose:
- designating multiple candidate base forecast models from the available forecast models based on whether the historic time series data exhibits seasonality;
- and identifying the plurality of candidate base forecast models from the multiple candidate base forecast models based on forecast performance of each of the multiple candidate base forecast models, by with respect to each of the multiple candidate base forecast models,
- generating a forecast result based on the historic time series data using the candidate base forecast model, computing a measure indicative of the performance of the candidate base forecast model based on the forecast result;
- computing a measure indicative of the performance of the candidate base forecast model based on the forecast result
- and selecting the plurality of candidate base forecast models from the multiple candidate base forecast models based on the measures associated respectively with the multiple candidate base forecast models
However, ALKURD discloses:
- designating multiple candidate base forecast models from the available forecast models based on whether the historic time series data exhibits seasonality;
In [0314]: “ Train and validate the model: There are several ML algorithms that can be used to build a predictive model. From a practical point of view, the best predictor candidate for our proposed framework is a deep neural network (DNN) algorithm.
In [0281]:
we perform user satisfaction prediction using the following traditional set of ML algorithms: Decision Tree (DT), K-nearest neighbor (Knn), and Random Forest (RF).
In [0281]:
hyperparameter tuning is conducted using grid-search. FIG. 22 depicts the 10-folds cross-validation prediction accuracies for each ML algorithm. Since there are six satisfaction levels, the random choice accuracy level is 0.166. FIG. 22 shows that the best performance is achieved using RF (ensemble) algorithm with an average accuracy of 0.884 compared to 0.85 for DT and Knn. It is worth mentioning that the predictors' performance should be directly related to the number of relevant context variables available to the ML algorithms.
In [0484]:
Various ML models can be utilized to build surrogates, including linear models, support vector machines [69], and Gaussian processes [70]. The present invention utilizes a big data-driven framework to build ML-based surrogate models in order to predict user satisfaction in wireless networks.
In [0039] :” In another embodiment, the context comprises one or more of the parameters: time, day, location, speed, activity, service request arrival”,
In [0067]: “ FIGS. 19a and 19b present the probability of request arrival (i.e., P{N(t, t+δ)=1)) vs. t in hours for a weekday and a weekend day, respectively”,
In [0065]: “ FIG. 17 presents a graphic representation setting out the percentage of time the user spent at each location over each time period on a weekday”,
PNG
media_image4.png
366
592
media_image4.png
Greyscale
- and identifying the plurality of candidate base forecast models from the multiple candidate base forecast models based on forecast performance of each of the multiple candidate base forecast models, by with respect to each of the multiple candidate base forecast models,
In [0281]:
hyperparameter tuning is conducted using grid-search. FIG. 22 depicts the 10-folds cross-validation prediction accuracies for each ML algorithm. Since there are six satisfaction levels, the random choice accuracy level is 0.166. FIG. 22 shows that the best performance is achieved using RF (ensemble) algorithm with an average accuracy of 0.884 compared to 0.85 for DT and Knn. It is worth mentioning that the predictors' performance should be directly related to the number of relevant context variables available to the ML algorithms.
- generating a forecast result based on the historic time series data using the candidate base forecast model, computing a measure indicative of the performance of the candidate base forecast model based on the forecast result;
In [0281]:”hyperparameter tuning is conducted using grid-search. FIG. 22 depicts the 10-folds cross-validation prediction accuracies for each ML algorithm. Since there are six satisfaction levels, the random choice accuracy level is 0.166. FIG. 22 shows that the best performance is achieved using RF (ensemble) algorithm with an average accuracy of 0.884 compared to 0.85 for DT and Knn. It is worth mentioning that the predictors' performance should be directly related to the number of relevant context variables available to the ML algorithms”,
In [0280]: “ ML and data analytics are important to extract patterns and knowledge from historical data and use it to predict user satisfaction behavior in the future”,
In [0475]: “ surrogate models, which are comparatively faster and rely on historical data and user patterns to predict user satisfaction values in real-time, may be very useful in implementing personalized wireless networks.
In [0065]: “ FIG. 17 presents a graphic representation setting out the percentage of time the user spent at each location over each time period on a weekday”,
PNG
media_image4.png
366
592
media_image4.png
Greyscale
- computing a measure indicative of the performance of the candidate base forecast model based on the forecast result
In [0281]:”hyperparameter tuning is conducted using grid-search. FIG. 22 depicts the 10-folds cross-validation prediction accuracies for each ML algorithm. Since there are six satisfaction levels, the random choice accuracy level is 0.166. FIG. 22 shows that the best performance is achieved using RF (ensemble) algorithm with an average accuracy of 0.884 compared to 0.85 for DT and Knn. It is worth mentioning that the predictors' performance should be directly related to the number of relevant context variables available to the ML algorithms”,
[BRI: accuracy is the measure of performance]
- and selecting the plurality of candidate base forecast models from the multiple candidate base forecast models based on the measures associated respectively with the multiple candidate base forecast models
In [0281]:”hyperparameter tuning is conducted using grid-search. FIG. 22 depicts the 10-folds cross-validation prediction accuracies for each ML algorithm. Since there are six satisfaction levels, the random choice accuracy level is 0.166. FIG. 22 shows that the best performance is achieved using RF (ensemble) algorithm with an average accuracy of 0.884 compared to 0.85 for DT and Knn. It is worth mentioning that the predictors' performance should be directly related to the number of relevant context variables available to the ML algorithms”,
In [0447] : “ EAs are a class of metaheuristics population-based optimization algorithms, where multiple candidate solutions are maintained in parallel. EAs are designed based on the idea of the survival of the “fittest” solution in order to evolve a population that is a good approximation of a desired global optimum [62]. The fitness of an evolved solution is a measure of its quality at solving the problem.
In [0134]: “ Current networks are designed mostly to be a “Universal Fit,” where service providers deliver services with a quality appeal to all types of users. However, user expectations of service quality are not “One Size Fits All.”
[BRI: choosing more than one model from a set of possible forecasting methods to serve as the foundation for a combined or ensemble forecast means no when no single model consistently outperforms all others across different conditions, or when the goal is to improve robustness and stability by leveraging the strengths of multiple approaches. In the context of forecasting, the “survival of the fittest” approach can indeed be used to select the best-performing candidate base forecast models from a set of alternatives. This is conceptually similar to how Genetic Algorithms (GAs) or Evolutionary Algorithms (EAs) work: they maintain a population of candidate solutions (here, forecast models), evaluate each using a fitness measure (e.g., forecast accuracy, error metrics), and iteratively improve the population by selecting, combining, and mutating models
Spec [0016] “ In the field of advanced data analytics and econometrics, different time series forecasting models and methods have been developed and applied to fit different tasks or time series data with different characteristics. There is no “one size fits all” time series forecast model that may be deployed in any situation.
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Lin, Nara and ALKURD.
Lin teaches a forecasting model that collects time series data from a wireless tower (base stations) and combines linear and non-linear forecast models.
Nara teaches ensembled forecasting model.
ALKURD teaches candidate forecast models.
One of ordinary skill would have motivation to combine Lin, Nara and ALKURD that can re-learning to improve the predictive model performance and to update the model with user behavioral changes that could occur over time (ALKURD [0170]).
In regard to claim 19: (Currently Amended):
Lin and Nara do not explicitly disclose:
- designating multiple candidate base forecast models from the available forecast models based on whether the historic time series data exhibits seasonality;
- and identifying the plurality of candidate base forecast models from the multiple candidate base forecast models based on forecast performance of each of the multiple candidate base forecast models, by with respect to each of the multiple candidate base forecast models,
- generating a forecast result based on the historic time series data using the candidate base forecast model, computing a measure indicative of the performance of the candidate base forecast model based on the forecast result;
- computing a measure indicative of the performance of the candidate base forecast model based on the forecast result
- and selecting the plurality of candidate base forecast models from the multiple candidate base forecast models based on the measures associated respectively with the multiple candidate base forecast models
However, ALKURD discloses:
- designating multiple candidate base forecast models from the available forecast models based on whether the historic time series data exhibits seasonality;
In [0314]: “ Train and validate the model: There are several ML algorithms that can be used to build a predictive model. From a practical point of view, the best predictor candidate for our proposed framework is a deep neural network (DNN) algorithm.
In [0281]:
we perform user satisfaction prediction using the following traditional set of ML algorithms: Decision Tree (DT), K-nearest neighbor (Knn), and Random Forest (RF).
In [0281]:
hyperparameter tuning is conducted using grid-search. FIG. 22 depicts the 10-folds cross-validation prediction accuracies for each ML algorithm. Since there are six satisfaction levels, the random choice accuracy level is 0.166. FIG. 22 shows that the best performance is achieved using RF (ensemble) algorithm with an average accuracy of 0.884 compared to 0.85 for DT and Knn. It is worth mentioning that the predictors' performance should be directly related to the number of relevant context variables available to the ML algorithms.
In [0484]:
Various ML models can be utilized to build surrogates, including linear models, support vector machines [69], and Gaussian processes [70]. The present invention utilizes a big data-driven framework to build ML-based surrogate models in order to predict user satisfaction in wireless networks.
In [0039] :” In another embodiment, the context comprises one or more of the parameters: time, day, location, speed, activity, service request arrival”,
In [0067]: “ FIGS. 19a and 19b present the probability of request arrival (i.e., P{N(t, t+δ)=1)) vs. t in hours for a weekday and a weekend day, respectively”,
In [0065]: “ FIG. 17 presents a graphic representation setting out the percentage of time the user spent at each location over each time period on a weekday”,
PNG
media_image4.png
366
592
media_image4.png
Greyscale
- and identifying the plurality of candidate base forecast models from the multiple candidate base forecast models based on forecast performance of each of the multiple candidate base forecast models, by with respect to each of the multiple candidate base forecast models,
In [0281]:
hyperparameter tuning is conducted using grid-search. FIG. 22 depicts the 10-folds cross-validation prediction accuracies for each ML algorithm. Since there are six satisfaction levels, the random choice accuracy level is 0.166. FIG. 22 shows that the best performance is achieved using RF (ensemble) algorithm with an average accuracy of 0.884 compared to 0.85 for DT and Knn. It is worth mentioning that the predictors' performance should be directly related to the number of relevant context variables available to the ML algorithms.
- generating a forecast result based on the historic time series data using the candidate base forecast model, computing a measure indicative of the performance of the candidate base forecast model based on the forecast result;
In [0281]:”hyperparameter tuning is conducted using grid-search. FIG. 22 depicts the 10-folds cross-validation prediction accuracies for each ML algorithm. Since there are six satisfaction levels, the random choice accuracy level is 0.166. FIG. 22 shows that the best performance is achieved using RF (ensemble) algorithm with an average accuracy of 0.884 compared to 0.85 for DT and Knn. It is worth mentioning that the predictors' performance should be directly related to the number of relevant context variables available to the ML algorithms”,
In [0280]: “ ML and data analytics are important to extract patterns and knowledge from historical data and use it to predict user satisfaction behavior in the future”,
In [0475]: “ surrogate models, which are comparatively faster and rely on historical data and user patterns to predict user satisfaction values in real-time, may be very useful in implementing personalized wireless networks.
In [0065]: “ FIG. 17 presents a graphic representation setting out the percentage of time the user spent at each location over each time period on a weekday”,
PNG
media_image4.png
366
592
media_image4.png
Greyscale
- computing a measure indicative of the performance of the candidate base forecast model based on the forecast result
In [0281]:”hyperparameter tuning is conducted using grid-search. FIG. 22 depicts the 10-folds cross-validation prediction accuracies for each ML algorithm. Since there are six satisfaction levels, the random choice accuracy level is 0.166. FIG. 22 shows that the best performance is achieved using RF (ensemble) algorithm with an average accuracy of 0.884 compared to 0.85 for DT and Knn. It is worth mentioning that the predictors' performance should be directly related to the number of relevant context variables available to the ML algorithms”,
[BRI: accuracy is the measure of performance]
- and selecting the plurality of candidate base forecast models from the multiple candidate base forecast models based on the measures associated respectively with the multiple candidate base forecast models
In [0281]:”hyperparameter tuning is conducted using grid-search. FIG. 22 depicts the 10-folds cross-validation prediction accuracies for each ML algorithm. Since there are six satisfaction levels, the random choice accuracy level is 0.166. FIG. 22 shows that the best performance is achieved using RF (ensemble) algorithm with an average accuracy of 0.884 compared to 0.85 for DT and Knn. It is worth mentioning that the predictors' performance should be directly related to the number of relevant context variables available to the ML algorithms”,
In [0447] : “ EAs are a class of metaheuristics population-based optimization algorithms, where multiple candidate solutions are maintained in parallel. EAs are designed based on the idea of the survival of the “fittest” solution in order to evolve a population that is a good approximation of a desired global optimum [62]. The fitness of an evolved solution is a measure of its quality at solving the problem.
In [0134]: “ Current networks are designed mostly to be a “Universal Fit,” where service providers deliver services with a quality appeal to all types of users. However, user expectations of service quality are not “One Size Fits All.”
[BRI: choosing more than one model from a set of possible forecasting methods to serve as the foundation for a combined or ensemble forecast means no when no single model consistently outperforms all others across different conditions, or when the goal is to improve robustness and stability by leveraging the strengths of multiple approaches. In the context of forecasting, the “survival of the fittest” approach can indeed be used to select the best-performing candidate base forecast models from a set of alternatives. This is conceptually similar to how Genetic Algorithms (GAs) or Evolutionary Algorithms (EAs) work: they maintain a population of candidate solutions (here, forecast models), evaluate each using a fitness measure (e.g., forecast accuracy, error metrics), and iteratively improve the population by selecting, combining, and mutating models
Spec [0016] “ In the field of advanced data analytics and econometrics, different time series forecasting models and methods have been developed and applied to fit different tasks or time series data with different characteristics. There is no “one size fits all” time series forecast model that may be deployed in any situation.
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Lin, Nara and ALKURD.
Lin teaches a forecasting model that collects time series data from a wireless tower (base stations) and combines linear and non-linear forecast models.
Nara teaches ensembled forecasting model.
ALKURD teaches candidate forecast models.
One of ordinary skill would have motivation to combine Lin, Nara and ALKURD that can re-learning to improve the predictive model performance and to update the model with user behavioral changes that could occur over time (ALKURD [0170]).
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the
examiner should be directed to TIRUMALE KRISHNASWAMY RAMESH whose telephone number is (571)272-4605. The examiner can normally be reached by phone.
Examiner interviews are available via telephone, in-person, and video conferencing
using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s
supervisor, Li B Zhen can be reached on phone (571-272-3768). The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be
obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit:
https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for
information about filing in DOCX format. For additional questions, contact the Electronic
Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO
Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/TIRUMALE K RAMESH/ Examiner, Art Unit 2121
/Li B. Zhen/Supervisory Patent Examiner, Art Unit 2121
Prosecution Insights