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 .
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claims 1-7, 9, 11-17, 19, 21-27 and 29 are rejected under 35 U.S.C. 102(a) (1) as being anticipated by Juan L. Bejarano-Luque [A Deep-Learning Model for Estimating the Impact of Social Events on Traffic Demand on a Cell Basis].
Regarding claim 1, Juan teaches:
1. A computer-implemented method for generating forecasts of unscheduled future events (i.e. the proposed deep-learning traffic forecast model- fig. 1), comprising:
at a processor (i.e. Intel quad-core processor, clock frequency of 1.8 GHz and 16 GB of RAM- page 71683), obtaining training data for a predictive model (i.e. A. Stage 1: Data Collection And Pre-Processing… the data collection period and geographical area must be large to have enough events to train a deep-learning model. It is shown later that a nationwide measurement dataset of a couple of months is enough to obtain reliable results. Network layout is provided by the network operator. Finally, information on past social events is collected in this work by combining different source- page 71676), wherein the training data comprises at least one past event (i.e. information about past events is also publicly available- page 71677) and at least one high-confidence future event (i.e. Event discovery platforms, on-line calendars and ticketing applications. These on-line platforms offer information of upcoming events in each city- page 71676);
at the processor, training the predictive model using the training data (i.e. Social event information is usually publicly available for future events, so that it can be easily collected by mobile operators for traffic prediction purposes. However, once the event has taken place, such information is normally deleted, hampering the creation of a large dataset to train a deep-learning model. Such a difficulty may explain why social event information has not yet been considered in state-of- the-art traffic forecasting models. To the authors' knowledge, no method has been proposed for traffic forecasting in mobile networks based on deep learning considering social events as an input. Hence, the main contributions of this work are: a) the preliminary analysis quantifying the impact of different types of events on cellular traffic, and b) a method for cellular traffic forecasting at a cell level on an hourly basis that takes into account social events- page 71675, ¶4…With these short time series, a deep-learning model is trained to predict the real traffic in the cell during the considered time window- page 71678, ¶3);
at the processor, applying the trained model to generate forecasts of unscheduled future events (i.e. The resulting model is combined with a traffic forecast module based on a multi-task deep-learning architecture to predict the hourly traffic series with scheduled mass events.- Abstract… Once the model is trained, the low computational load of the underlying operations allows easy integration in radio planning tools- page 71684, ¶2… The proposed method can predict the local traffic pattern generated by social events 24 hours in advance, which is needed to plan mobile network infrastructures (e.g., cell on wheels)- page 71684, ¶3); and
at an output device, outputting the forecasts of unscheduled future events (i.e. The output of the model is the predicted time series of cell traffic on an hourly basis- page 71676, ¶5… Thus, the output of the model is the time series for the next 24 hours in normal conditions predicted per serving cell- page 71678, ¶1).
Regarding claim 2, Juan teaches all the limitations of claim 1 and Juan further teaches:
wherein:each past event comprises an event having an event occurrence time prior to the present time (i.e. In this work, a convolutional AE is tested to predict a time series (daily cell traffic) from past values of the same series and a second time series (event occurrence series)- page 71676, ¶3); and each high-confidence future event comprises a previously scheduled event having an event occurrence time subsequent to the present time (i.e. Social event information is usually publicly available for future events, so that it can be easily collected by mobile operators for traffic prediction purposes- page 71675, ¶4).
Regarding claim 3, Juan teaches all the limitations of claim 2 and Juan further teaches:
wherein training the predictive model comprises:
at the processor, generating a feature matrix from the training data; andat the processor, training the predictive model using the generated feature matrix (i.e. The MLP is trained with a back-propagation method [56] that optimizes a selected loss function (e.g., mean absolute error) using a gradient descent algorithm- page 71675, ¶5…Fig. 3 shows the three different architectures tested for modeling the impact of events. The first architecture, shown in Fig. 3(a), is a MLP. The input is a bi-dimensional matrix, consisting of the traffic forecast- page 71678, ¶4).
4. The method of claim 3, wherein the feature matrix comprises a two- dimensional feature matrix, and wherein:
a first dimension of the feature matrix represents event occurrence time; and a second dimension of the feature matrix represents time at which each event is scheduled (i.e. the traffic forecast,
T
^
fW(t, c), and the event occurrence vector, Ew(t, c, e), in the event time window- page 71678, ¶4).
Regarding claim 5, Juan teaches all the limitations of claim 2 and Juan further teaches:
wherein training the predictive model comprises training the predictive model using a combination of at least one past event (i.e. A. Stage 1: Data Collection And Pre-Processing… the data collection period and geographical area must be large to have enough events to train a deep-learning model. It is shown later that a nationwide measurement dataset of a couple of months is enough to obtain reliable results. Network layout is provided by the network operator. Finally, information on past social events is collected in this work by combining different source- page 71676) and at least one high-confidence future event (i.e. Social event information is usually publicly available for future events, so that it can be easily collected by mobile operators for traffic prediction purposes. However, once the event has taken place, such information is normally deleted, hampering the creation of a large dataset to train a deep-learning model. Such a difficulty may explain why social event information has not yet been considered in state-of- the-art traffic forecasting models. To the authors' knowledge, no method has been proposed for traffic forecasting in mobile networks based on deep learning considering social events as an input. Hence, the main contributions of this work are: a) the preliminary analysis quantifying the impact of different types of events on cellular traffic, and b) a method for cellular traffic forecasting at a cell level on an hourly basis that takes into account social events- page 71675, ¶4…With these short time series, a deep-learning model is trained to predict the real traffic in the cell during the considered time window- page 71678, ¶3).
Regarding claim 6, Juan teaches all the limitations of claim 2 and Juan further teaches:
wherein the predictive model comprises a machine learning model (i.e. Multilayer Perceptron (MLP)- page . 71675, ¶3).
Regarding claim 7, Juan teaches all the limitations of claim 2 and Juan further teaches:
wherein applying the trained model to generate forecasts of unscheduled future events comprises: receiving inference data; and applying the trained model to the inference data (i.e. For simplicity, it is assumed that the impact of an event is restricted to a single cell (the one serving the central location of the event) and a limited time window (11 hours centered at the central hour of the event)… the model only needs to update the daily traffic forecast in the selected time window around the event in the serving cell- page 71678, ¶2-5).
Regarding claim 9, Juan teaches all the limitations of claim 1 and Juan further teaches:
wherein the predictive model comprises a time series forecasting model (i.e. The proposed method aims to increase accuracy in cell traffic forecasting, correcting a classical deep-learning time series model with spatio-temporal information of scheduled events- ¶AGE 71676, ¶4).
Regarding claims 11-17, and 19, computer-readable medium storing instructions claims 11-17, 19 correspond to the same method as claimed in claims 1-7 and 9, respectively, and therefore are also rejected for the same rationale as listed above.
Regarding claims 21-27, and 29, apparatus claims 21-27, and 9 are drawn to the apparatus using/performing the same method as claimed in claims 1-7 and 9, respectively. Therefore, apparatus claims 21-27, and 29 correspond to method claims 1-7 and 9, respectively, and are rejected for the same rationale as used above
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 of this title, 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.
The factual inquiries set forth in Graham v. John Deere Co., 383 U.S. 1, 148 USPQ 459 (1966), that are applied for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 8, 18 and 28 are rejected under 35 U.S.C. 103 as being unpatentable over Juan L. Bejarano-Luque [A Deep-Learning Model for Estimating the Impact of Social Events on Traffic Demand on a Cell Basis] in view of Turgut Aykin [US 8577706 B1].
Regarding claim 8, Juan teaches all the limitations of claim 1.
However, Juan does not teach explicitly:
wherein: each past event comprises an employee leave of absence having an event occurrence time prior to the present time; and each high-confidence future event comprises a previously scheduled employee leave of absence having an event occurrence time subsequent to the present time.
In the same field of endeavor, Turgut teaches:
wherein: each past event comprises an employee leave of absence having an event occurrence time prior to the present time; and each high-confidence future event comprises a previously scheduled employee leave of absence having an event occurrence time subsequent to the present time (i.e. Agents are scheduled to weekly schedule patterns referred to as tours. To describe the MILP model, it is assumed that a tour specifies a weekly work pattern that is seven days long with the following breaks in the work schedule: (i) Daily breaks in the shift schedule including two relief breaks, and one lunch break during a work day, and (ii) two non-work days (i.e. days-off). To simplify the notation used to describe the MILP model, it is further assumed that each relief break is one planning period, and lunch break two planning periods long. Extensions of the MILP model of the method to tours with more or less breaks, and with different break durations together with various other scheduling rules are disclosed in U.S. patent application Ser. No. 12/584,210, and U.S. Pat. No. 7,725,339- col 12, line 45-58).
It would have been obvious to one with ordinary skill in the art before the effective filing date of the claimed invention, to modify the teachings of Juan with the teachings of Turgut to determine the starting staffing levels to meet the performance targets if the skill group considered alone serves the target contact groups (Turgut- col 4, line 61-63).
Regarding claim 18, computer-readable medium storing instructions claim 18 corresponds to the same method as claimed in claim 8, and therefore is also rejected for the same reasons of obviousness as listed above.
Regarding claim 28, apparatus claim 16 is drawn to the apparatus using/performing the same method as claimed in claim 8. Therefore, apparatus claim 28 corresponds to method claim 8, and is rejected for the same reasons of obviousness as used above.
Claims 10, 20 and 30 are rejected under 35 U.S.C. 103 as being unpatentable over Juan L. Bejarano-Luque et al. [A Deep-Learning Model for Estimating the Impact of Social Events on Traffic Demand on a Cell Basis] in view of Wenyu Zhang et al.[Multi-label Prediction in Time Series Data using Deep Neural Networks].
Regarding claim 10, Juan teaches all the limitations of claim 1.
a Long Short- Term Memory (LSTM) layer and a fully connected layer (i.e. The basic LSTM network consists of a single hidden layer, with several LSTM neurons to model multiple hidden states, and an output layer, consisting of a fully connected layer of perceptrons to derive the forecasted value(s)- page 71675, ¶6-8)
However, Juan does not teach explicitly:
wherein the time series forecasting model comprises a multi-label deep learning regression model.
In the same field of endeavor, Wenyu teaches:
wherein the time series forecasting model comprises a multi-label deep learning regression model (i.e. This paper addresses a multi-label predictive fault classification problem for multidimensional time-series data… In this paper, LSTM models are used in a predictive network to take advantage of their ability to model long-term dependencies- page 3, col 2… the ensemble comprises gradient boosting machine, random forest and penalized logistic regression- page 7, col 1).
It would have been obvious to one with ordinary skill in the art before the effective filing date of the claimed invention, to modify the teachings of Juan with the teachings of Wenyu to address the problem of most of the state-of-the-art techniques which can’t reliably predict faults (events) over a desired future horizon (Wenyu- Abstract).
Regarding claim 20, computer-readable medium storing instructions claim 20 corresponds to the same method as claimed in claim 10, and therefore is also rejected for the same reasons of obviousness as listed above.
Regarding claim 30, apparatus claim 30 is drawn to the apparatus using/performing the same method as claimed in claim 10. Therefore, apparatus claim 30 corresponds to method claim 10, and is rejected for the same reasons of obviousness as used above.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to CLIFFORD HILAIRE whose telephone number is (571)272-8397. The examiner can normally be reached 5:30-1400.
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, SATH V PERUNGAVOOR can be reached at (571)272-7455. 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.
CLIFFORD HILAIRE
Primary Examiner
Art Unit 2488
/CLIFFORD HILAIRE/Primary Examiner, Art Unit 2488