DETAILED ACTION
This action is responsive to Applicant’s reply filed 13 April 2026. This action is made non-final.
Status of the Claims
Claims 1, 11 and 16 are amended.
Claim status is currently pending and under examination for Claims 1-20 of which independent claims are 1, 11 and 16.
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 .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on April 13, 2026 has been entered.
Response to Amendment
Applicant’s amendments to the Claims have overcome each and every 101 rejections previously set forth in the Final Office Action mailed January 13th 2026.
Applicant’s arguments regarding the art rejections are moot in view of the new grounds of rejection necessitated by Applicant’s amendment.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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.
The following are the references relied upon in the rejections below:
Sánchez-Fernández, Alvar, et al. "Fault detection based on time series modeling and multivariate statistical process control." Chemometrics and Intelligent Laboratory Systems 182 (2018): 57-69.
Mariia, Matskevichus, et al. "Model selection approach for time series forecasting." 2019 IEEE 13th International Conference on Application of Information and Communication Technologies (AICT). IEEE, 2019.
Claims 1-4, 9, 11-13, 16-18 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Sánchez-Fernández in view of Mariia.
Regarding Claim 1, Sánchez-Fernández teaches:
A system for generating a prediction of time-series data from data sets, the system comprising ((P. 58, Sec. 2, ¶1) “it is necessary to identify m different models, for a system with m sensors measuring m system variables, from the measurement data recorded by the sensors. In particular, for each system variable
x
i
,
i
=
1
,
…
,
m
, we are looking for a forecasting time-series model fi based on a subset or a transformation of the other variables which are best able to predict this variable”):
memory storing computer program instructions; and one or more processors configured to execute the computer program instructions to ((P. 63, Sec. 4.1.1, ¶2) “model each variable; … modelling, this is done to avoid over-fitting models, especially for the neural networks”
A computer is implied by modeling neural networks, which further implies a memory storing instructions that are executable by a processor.):
retrieve, for each product of a set of products, the time-series data including a plurality of time-value pairs (The Examiner interprets “product” according to its broadest reasonable interpretation (BRI) in view of the applicant’s specification (at [0002, 0040]) as encompassing system variables that measure sensor data over time.
(P. 58, Sec. 2, ¶1) “it is necessary to identify m different models, for a system with m sensors measuring m system variables, from the measurement data recorded by the sensors. In particular, for each system variable
x
i
,
i
=
1
,
…
,
m
, we are looking for a forecasting time-series model … the prediction model fi for each variable
x
i
,
i
=
1
,
…
,
m
includes the necessary time-lags on a subset of other variables, which are the best to explain this variable. So, the data set containing the variable to be modelled is spanned to include lags along the m original non-delayed variables as inputs to the model.”
System variables (‘products’) that have time-lags are included in a data set used as input for a forecasting time-series model fi, therefore variables have time-series data with a plurality of time-value pairs.);
select a first product from the set of products ((P. 58, Sec. 2, ¶1) “for each system variable
x
i
,
i
=
1
,
…
,
m
, we are looking for a forecasting time-series model”
(P. 59, Sec. 2.1, ¶1) “To implement the dynamic feature selection, the relationship between two variables
x
t
i
and
x
l
j
at two different time instants is calculated as the absolute value of the correlation coefficient”
A model is chosen for each system variable
x
i
, therefore the i-th system variable
x
t
i
is the first product from the set.);
compute correlation values between the first product and a plurality of other products from the set of products, for one or more degrees of lag, each correlation value being between the first product and one of the plurality of other products, to obtain a set of correlation values representing respective correlations between the first product and the plurality of other products assessed at prior times ((P. 58, Sec. 2.1, ¶1) “The interaction among different measured variables might be more appropriately represented on the basis of different time-delays. In order to model every variable, it is necessary to know which are the best relations for using. So a dynamic feature selection concerning auto- and cross-correlation with different time-delays is carried out. First, the matrix X (n samples
×
m variables), with the original variables in normal operation conditions, is augmented, … in the following manner Xa …”
(P. 59, Sec. 2.1, ¶1) “To implement the dynamic feature selection, the relationship between two variables
x
t
i
and
x
l
j
at two different time instants is calculated as the absolute value of the correlation coefficient:
PNG
media_image1.png
52
689
media_image1.png
Greyscale
where
i
,
j
=
1
,
…
,
m
and
t
,
l
=
1
,
…
,
L
. This coefficient is a direct measure of the correlation between variables and, after the calculation of vector
R
i
for the i-th variable, the dynamic feature selection is carried out by selecting the variables in matrix Xa with high correlation values”
A correlation coefficient is calculated between the i-th system variable
x
t
i
(‘first product’) and
x
l
j
(a second variable at a different time instant/delay), therefore a correlation coefficient is a correlation value between a first product and one other product assessed at a prior time for one or more degrees of lag. A correlation coefficient is computed between variable
x
t
i
and each variable
x
l
j
, where j represents a different j-th system variable (therefore computing correlation values between the first product and a plurality of other products, for one or more degrees of lag to obtain a set of correlation values representing respective correlations between the first product and the plurality of other products assessed at prior times).);
select a subset of products from the set of products based at least in part on the correlation values ((P. 59, Sec. 2.1, ¶1) “the dynamic feature selection is carried out by selecting the variables in matrix Xa with high correlation values”);
select, based on … the time-series data comprising a number of time points per series and associated with the first product and the subset of products, a type of machine learning model from among a plurality of model types comprising at least one of a regression-based model, a decision-tree-based model, or a deep-learning model ((P. 63, Sec. 4.1.1, ¶2-3) “The first step is the dynamic feature selection, i.e., to discover which delayed variables must be used to model each variable; … The faultless training data (500 observations) were used to model each variable with its corresponding dynamic features as inputs. … the developed time-series models were: ARIMA models, with the structure ARIMA(p,q,i) … Neural network models. The NN used is always a Perceptron Multilayer network (MLP) with three layers”
(P. 64, Sec. 4.1.1, ¶1) “The model with the lowest rMSE value using test data was selected”
(P. 61, Sec. 3, ¶3) “the goal is to find the best time-series descriptive model for every system variable with the dynamic features selected in step 1 from training data … Some measures of the error committed in forecasting each test time series with each respective model are used to select the preference of these models. These error measures are: rMSE (root-mean-square error)”);
provide the time-series data associated with each product from the subset of products and the first product to the selected machine learning model trained configured to predict a future value of the first product based on values of the subset of products at the prior times ((P. 61, Sec. 3, ¶5) “the control limits of every chart used in this methodology need to be calculated. They are based on the residuals, calculated as the difference between the prediction for each variable using the respective time-series model chosen in step 2 and the actual measurement of this variable”
(P. 64, Sec. 4.1.1, ¶2) “Now a forecast is obtained for every variable using the respective model and then compared with its actual measured value, resulting in a residual”
(P. 68, Sec. 5, ¶1) “proposed methodology explicitly accounts for the dynamic relations in the process data through dynamic feature selection, … After that the residuals, which are the difference between the process measurements and the output of the model, are monitored”
Residuals are calculated by finding the difference between process measurements (process data used to determine the correlated, delayed features of the i-th system variable) and a forecast obtained from a chosen model, therefore providing time-series data associated with each product from the subset of products and the first product (correlated features of the i-th variable) to the selected machine learning model configured to predict a future value of the first product (i-th system variable) based on values of the subset of products at prior times.);
and obtain, from the selected machine learning model, prediction data representing a set of predicted values for the first product at one or more future times ((P. 64, Sec. 4.1.1, ¶2) “Now a forecast is obtained for every variable using the respective model”),
wherein the selected machine learning model is configured to generate the prediction data for the first product ((P. 58, Sec. 2, ¶1) “for each system variable
x
i
,
i
=
1
,
…
,
m
, we are looking for a forecasting time-series model fi based on a subset or a transformation of the other variables which are best able to predict this variable”).
However, Sánchez-Fernández does not teach selecting, based on data dimensions of time-series data, a type of machine learning model from among a plurality of model types, which is taught by Mariia:
select, based on data dimensions of the time-series data comprising a number of time points per series …, a type of machine learning model from among a plurality of model types comprising at least one of a regression-based model, a decision-tree-based model, or a deep-learning model ((P. 1, Abstract) “the model selection aims to estimate the performance of different model candidates in order to choose the most appropriate one. In this study we suggest exploiting specific features of time series for the optimal forecasting model selection such as length, seasonality, trend strength and others. To demonstrate reliability of feature-based approach, forecasting error distribution of LSTM Recurrent Neural Network, Linear Regression model, Holt-Winters model and ARIMA model trained on 250 time series with various characteristics were compared. Results of statistical experiments have demonstrated a significant dependence of a forecasting model on the characteristics of a series. Proposed model selection approach allows formulating a priori recommendations for choosing the optimal forecasting model for the specific time series”
Mariia discloses Table 1 (reproduced below) on P. 2 depicting a table of model selection approaches based on time series length (‘number of time points per series’) and other characteristics. Table 1 shows that for a time series with a length of 200 or more, a recurrent neural network (‘a deep learning model’) should be selected. Furthermore, Table 1 depicts that for a time series with a length of less than 200, a linear regression model, Holt-Winters model, or an ARIMA model should be selected.
PNG
media_image2.png
621
722
media_image2.png
Greyscale
);
Mariia teaches selecting forecasting models based on time series length is a known method in the art. Before the effective filing date of the claimed invention, it would have been obvious to combine the method of Sánchez-Fernández with the model selection method disclosed by Mariia to choose an optimal forecasting model based on time series length. By choosing an optimal forecasting model based on time series length, forecasting errors can be minimized, thereby improving a model’s ability to generate reliable predictions.
Regarding Claims 2, 12 and 17, the combined method of Sánchez-Fernández/Mariia teaches:
the system of claim 1, wherein the machine learning model is configured to generate the prediction data for only the first product ((P. 58, Sec. 2, ¶1) “for each system variable
x
i
,
i
=
1
,
…
,
m
, we are looking for a forecasting time-series model fi based on a subset or a transformation of the other variables which are best able to predict this variable”).
Regarding Claim 3, claim 3 is an obvious extension of claim 1. The Examiner finds that it would have been obvious before the effective filing date of the claimed invention to generate prediction data for only some products with a reasonable expectation of success since there are finite quantities of predictions that can be made. Given that the combination of Sánchez-Fernández/Mariia teaches the system of claim 1, there are only three possible quantities of predictions that can be made: (1) make predictions for all products; (2) make predictions for some products only; or (3) make no predictions. The advantage of making predictions for all products is that it would result in some predictions that are useful. The advantage of making no predictions is that no computing resources are used. Making predictions for only some products offers system engineers a way to balance these two concerns, and therefore is obvious. See MPEP § 2143(I)(E) "Obvious to try" rationale. For the forgoing reasons, claim 3 is obvious in view of the combination of Sánchez-Fernández/Mariia.
Regarding Claims 4, 13 and 18, the combined method of Sánchez-Fernández/Mariia teaches:
the system of claim 1, further comprising: determining a ranking of correlation values from the set of correlation values ((P. 59, Sec. 2.1, ¶1) “the dynamic feature selection is carried out by selecting the variables in matrix Xa with high correlation values”
(P. 63, Sec. 4.1.1, ¶2) “The 10 most correlated delayed variables are used for each variable for the next step: modelling, this is done to avoid over-fitting models, especially for the neural networks.”),
wherein the ranking of correlation values indicates which products from the set of products have data trends that are most strongly correlated with a first data trend of the first product ((P. 59, Sec. 2.1, ¶1) “To implement the dynamic feature selection, the relationship between two variables
x
t
i
and
x
l
j
at two different time instants is calculated as the absolute value of the correlation coefficient: … This coefficient is a direct measure of the correlation between variables and, after the calculation of vector
R
i
for the i-th variable, the dynamic feature selection is carried out by selecting the variables in matrix Xa with high correlation values”
A correlation coefficient is calculated between the i-th system variable
x
t
i
(first product) and
x
l
j
(a second variable at a different time delay), therefore a correlation coefficient is a correlation value between a first product and one other product (and therefore a correlation coefficient indicates how correlated a data trend of a first product is with a data trend of a second product). Selecting the variables with the highest correlation values represents the variables (products) that are strongly correlated with the i-th system variable
x
t
i
(first product).),
wherein the subset of products have the top N correlation values from the ranking ((P. 63, Sec. 4.1.1, ¶2) “The 10 most correlated delayed variables are used for each variable for the next step: modelling”).
Regarding Claim 9, the combined method of Sánchez-Fernández/Mariia teaches:
the system of claim 1, wherein: the time-series data includes first time-series data associated with the first product and second time-series data associated with a second product from the set of products ((P. 58, Sec. 2, ¶1) “it is necessary to identify m different models, for a system with m sensors measuring m system variables, from the measurement data recorded by the sensors. In particular, for each system variable
x
i
,
i
=
1
,
…
,
m
, we are looking for a forecasting time-series model … the prediction model fi for each variable
x
i
,
i
=
1
,
…
,
m
includes the necessary time-lags on a subset of other variables, which are the best to explain this variable. So, the data set containing the variable to be modelled is spanned to include lags along the m original non-delayed variables as inputs to the model.”
System variables (‘products’) are included in a data set used as input for a forecasting time-series model fi, therefore all variables have time-series data.);
the first time-series data comprises a first plurality of time-value pairs, wherein each time-value pair of the first plurality of time-value pairs represents a value associated with the first product at each of a first set of times ((P. 59, Sec. 2.1, ¶1) “To implement the dynamic feature selection, the relationship between two variables
x
t
i
and
x
l
j
at two different time instants is calculated as the absolute value of the correlation coefficient”
A correlation coefficient is calculated between the i-th system variable
x
t
i
(first product) and
x
l
j
(second product) at two different time instants. Therefore, the i-th system variable
x
t
i
has time-series data comprised of time-value pairs (each representing a value) associated with the i-th variable at a time instant (‘first set of times’).);
the second time-series data comprises a second plurality of time-value pairs, wherein each time-value pair of the second plurality of time-value pairs represents a value associated with the second product at each of a second set of times (The i-th system variable
x
t
i
(first product) and
x
l
j
(second product) are compared at two different time instants. Therefore, the j-th system variable
x
l
j
has time-series data comprised of time-value pairs (each representing a value) associated with the j-th variable at a time instant (‘second set of times’).);
the first set of times being discrete and captured at a first temporal frequency ((P. 66, Sec. 4.2, Last Paragraph) “the method used here is applied over a set of 7 variables, which are easier to obtain in a real plant … the measurements were recorded every 8 h”
The time series of each variable are discrete since measurements were recorded every 8 hours. The i-th system variable
x
t
i
(first product) is compared at a different time instant than variable
x
l
j
, therefore the i-th system variable is captured at a first temporal frequency.);
the second set of times being discrete and captured at a second temporal frequency ((P. 58, Sec. 2, ¶1) “L denotes the maximal lags and will be set to a concrete value calculated with a dynamic feature selection”
The time series of each variable are discrete since measurements were recorded every 8 hours. The variable
x
l
j
is delayed (lagged) and is compared at a different time instant than the i-th system variable, therefore variable
x
l
j
is captured at a second temporal frequency.);
the first temporal frequency and the second temporal frequency differ (The temporal frequencies of the i-th system variable
x
t
i
and variable
x
l
j
differ because they are compared at different time instants and variable
x
l
j
is lagged.).
Regarding Claim 11, the rejection of claim 1 is incorporated. The difference in scope being:
a non-transitory computer readable medium having instructions recorded thereon for generating a prediction of time-series data from data sets, the instructions when executed by a computer having at least one programmable processor cause operations comprising ((P. 63, Sec. 4.1.1, ¶2) “modelling, this is done to avoid over-fitting models, especially for the neural networks”
A computer is implied by modeling neural networks, which further implies a non-transitory computer readable medium having instructions that are executable by a processor of the computer.).
Regarding Claim 16, the rejection of claim 1 is incorporated. The difference in scope being:
a method for implementation by at least one programmable processor ((P. 63, Sec. 4.1.1, ¶2) “modelling, this is done to avoid over-fitting models, especially for the neural networks”
A computer is implied by modeling neural networks, which further implies a programmable processor.).
Regarding Claim 20, the combined method of Sánchez-Fernández/Mariia teaches:
the method of claim 16, wherein: when the data dimensions of the time-series data have 2-40 time points per series, select LASSO, when the data dimensions of the time-series data have 40-5000 time points per series, select Random Forests,
and when the data dimensions of the time-series data have 5000 or more time points per series, select Deep Learning (Mariia discloses “the model selection aims to estimate the performance of different model candidates in order to choose the most appropriate one. In this study we suggest exploiting specific features of time series for the optimal forecasting model selection such as length, seasonality, trend strength and others. To demonstrate reliability of feature-based approach, forecasting error distribution of LSTM Recurrent Neural Network, Linear Regression model, Holt-Winters model and ARIMA model trained on 250 time series with various characteristics were compared. Results of statistical experiments have demonstrated a significant dependence of a forecasting model on the characteristics of a series. Proposed model selection approach allows formulating a priori recommendations for choosing the optimal forecasting model for the specific time series” (P. 1, Abstract).
Mariia discloses Table 1 (reproduced below) on P. 2 depicting a table of model selection approaches based on time series length (‘time points’) and other features. Table 1 shows that for a time series with a length of 200 or more, a recurrent neural network (a deep learning model) should be selected.
PNG
media_image2.png
621
722
media_image2.png
Greyscale
).
Mariia teaches performing machine learning model selection based on time series length (‘time points’) is a known method in the art. Before the effective filing date of the claimed invention, it would have been obvious to modify the combined method of Sánchez-Fernández/Mariia with the technique disclosed by Mariia to train an optimal machine learning model. By selecting a machine learning model to train based on time series length, an appropriate model can be chosen to capture underlying patterns and trends since model complexity affects how complex relationships and patterns are learned. Therefore, training a more complex model for longer time series data can yield a more accurate model with reliable predictions.
The claimed invention in the instant application is directed to having contingent limitations. Per MPEP § 2111.04(ii) “the broadest reasonable interpretation of a method (or process) claim having contingent limitations requires only those steps that must be performed and does not include steps that are not required to be performed because the condition(s) precedent are not met. For example, assume a method claim requires step A if a first condition happens and step B if a second condition happens. If the claimed invention may be practiced without either the first or second condition happening, then neither step A or B is required by the broadest reasonable interpretation of the claim.” The claimed invention may be practiced without selecting each of the LASSO, Random Forest, and Deep Learning models. If time-series data with 5000 or more time points per series is acquired, the invention can still be practiced without having to select a LASSO or a Random Forest model. In another example, for time series data with 2 time points per series, a LASSO model is selected and the claimed invention can still be performed without selecting a Random Forest or Deep Learning model. When one model is selected, the other steps are not required to occur to practice the invention, therefore, the method steps of claim 20 are not required to be performed under a broadest reasonable interpretation of the claim and are obvious over the prior art.
The following are the references relied upon in the rejections below:
Polizzotto (US 10810637 B2)
Claims 5, 14, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Sánchez-Fernández in view of Mariia, further in view of Polizzotto.
Regarding Claims 5, 14 and 19, the combined method of Sánchez-Fernández/Mariia teaches the system of claim 1, however, the combination does not teach a product comprising an environmental, social, and governance metric, which is taught by Polizzotto:
wherein at least one product of the set of products comprises an environmental, social, and governance (ESG) metric, and wherein the ESG metric is one of a carbon metric, an ESG fund ratings metric, or an ESG product involvement metric (Polizzotto discloses “a responsibility score may define the manner in which a client (e.g. a corporation or business entity) is viewed with respect to their social responsibility. For example, a current responsibility score associated with the client may include one or more of: an environmental score; a social score; and a governance score, wherein one example of such a responsibility score is an ESG score. As is known in the art, an ESG score is defined using various ESG scoring criteria. Further and as discussed above, social platform promotion process 10 may recommend social platforms (chosen from social platform pool 56) that may address perceived social responsibility issues associated with a client, wherein these social responsibility issues may often be identified by social platform promotion process 10 examining a responsibility score. Accordingly, social platform promotion process 10 may be configured to predict how a responsibility score may change when a client contributes to one of the social platforms recommended by social platform promotion process 10” (Col. 17, line 62 to Col. 18, line 14).
Polizzotto further discloses “examples of ESG criteria used by investors include determining a company's impact on climate change or carbon emissions, water use or conservation efforts, anti-corruption policies, board diversity, human rights efforts and community development” (Col. 10, lines 4-8).).
Polizzotto teaches using an environmental, social, and governance (ESG) score (‘ESG metric’) to predict a company’s social perception is a known method in the art. Before the effective filing date of the claimed invention, it would have been obvious to modify the combined method of Sánchez-Fernández/Mariia with the ESG score disclosed by Polizzotto to predict the effect social perception has on a product’s performance. Social perceptions regarding a company can affect the performance of a product if a company is deemed to be socially or environmentally irresponsible and by calculating an ESG score, a company’s perceived social responsibility can be quantified and be used to predict a product’s future value. Therefore, including an ESG score in data analysis would yield a more accurate prediction since social perceptions influence a product’s performance.
The following are the references relied upon in the rejections below:
Maeser (US 20200327434 A1)
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over Sánchez-Fernández in view of Mariia, further in view of Maeser.
Regarding Claim 6, the combined method of Sánchez-Fernández/Mariia teaches the system of claim 1, however, the combination does not teach a regression model with a random error term, which Maeser does:
wherein a regression model that predicts the future values as implemented by the machine learning model includes a random error term (Maeser discloses “the relationship between the response (Y) and predictor variables
(
X
1
,
X
2
,
X
3
,
X
4
,
X
5
)
can be approximated by the regression models of
Y
=
f
(
X
1
)
+
E
for simple regression and
Y
=
f
(
X
1
,
X
2
,
X
3
,
X
4
,
X
5
)
+
E
for multiple regression. “E is assumed to be a random error representing the discrepancy in the approximation” and accounts for the “failure of the model to fit the data exactly” [1]” [107].).
Maeser teaches using a regression model with a random error term to make predictions is a known method in the art. Before the effective filing date of the claimed invention, it would have been obvious to modify the combined method of Sánchez-Fernández/Mariia with the regression model disclosed by Maeser to capture missing variables. By calculating a random error in a regression model, the amount that a regression model fails to fit a set of data can be measured. The random error can then be used to make observations about the model such as that there are missing influential variables and the model needs to be trained again, thereby increasing model accuracy and performance upon retraining.
The following are the references relied upon in the rejections below:
Harwalkar, Disha, et al. "Analytical Study of Correlation Between Demand and Renewable Energy Forecasting Using Data Mining/Analytics." Journal of Emerging Technologies and Innovative Research 7 (2020): 72-78.
Claim 7 is rejected under 35 U.S.C. 103 as being unpatentable over Sánchez-Fernández in view of Mariia, further in view of Harwalkar.
With respect to claim 7, the combined method of Sánchez-Fernández/Mariia teaches the system of claim 1, however, the combination does not teach calculating a correlation value using Spearman’s correlation coefficient, which is taught by Harwalkar:
wherein the correlation value is computed using Spearman's correlation coefficient (Harwalkar discloses “Spearman's rank correlation coefficient … is a nonparametric measure of rank correlation (statistical dependence between the rankings of two variables). It assesses how well the relationship between two variables can be described using a monotonic function. … Spearman's correlation assesses monotonic relationships (whether linear or not). If there are no repeated data values, a perfect Spearman correlation of +1 or −1 occurs when each of the variables is a perfect monotone function of the other. Intuitively, the Spearman correlation between two variables will be high when observations have a similar (or identical for a correlation of 1) rank (i.e. relative position label of the observations within the variable: 1st, 2nd, 3rd, etc.) between the two variables, and low when observations have a dissimilar (or fully opposed for a correlation of −1) rank between the two variables. Spearman's coefficient is appropriate for both continuous and discrete ordinal variables” (P. 74-75, Sec. 4, Last Paragraph).).
Harwalkar teaches using Spearman’s correlation coefficient to rank and correlate variables is a known method in the art. Before the effective filing date of the claimed invention, it would have been obvious to modify the combined method of Sánchez-Fernández/Mariia with the technique disclosed by Harwalkar to correlate and rank continuous and discrete ordinal variables. By correlating and ranking continuous and discrete ordinal variables, underlying relationships between variables can be discovered and variables can be ranked in order of their influence. Therefore, a better understanding of data can be achieved to make well-informed decisions.
The following are the references relied upon in the rejections below:
Yao (US 20190050711 A1)
Claim 10 is rejected under 35 U.S.C. 103 as being unpatentable over Sánchez-Fernández in view of Mariia, further in view of Yao.
Regarding Claim 10, the combined method of Sánchez-Fernández/Mariia teaches “the system of claim 9, wherein the one or more processors are further caused to” however, the combination does not teach generating intermediate values by interpolation, which is taught by Yao:
generate intermediate values for the second product at each of the first set of times of which there is no corresponding value for the second product from the second plurality of time-value pairs (Yao discloses “when different sensors acquire and record data, hardware faults or signal transmission faults may happen, the data acquisition frequencies of different sensors may also be different, thus, if the timestamps of the time series data from different sensors are different, missing values are filled into the time series data via a linear interpolation compensation method” [0078].
Yao further discloses “in step S81, when the timestamps of the time series data from different sensors are different, linear interpolation compensation is performed on the time series data with a low sampling frequency. For example, the sampling frequency of the data from the sensor 1 is 10 Hz, and the sampling frequency of the data from the sensor 2 is 100 Hz, so that the timestamps are different. The data with the sampling frequency of 10 Hz is interpolated to the high frequency of 100 Hz first, so that the data from the sensor 1 and the data from the sensor 2 are both 100 Hz and have the same timestamp” [0080-0081].),
wherein the intermediate values are determined by interpolating the second plurality of time-value pairs at each of the first set of times of which there is no corresponding value for the second product from the second plurality of time-value pairs (Yao discloses “in step S81, when the timestamps of the time series data from different sensors are different, linear interpolation compensation is performed on the time series data with a low sampling frequency. For example, the sampling frequency of the data from the sensor 1 is 10 Hz, and the sampling frequency of the data from the sensor 2 is 100 Hz, so that the timestamps are different. The data with the sampling frequency of 10 Hz is interpolated to the high frequency of 100 Hz first, so that the data from the sensor 1 and the data from the sensor 2 are both 100 Hz and have the same timestamp” [0080-0081].).
Yao teaches filling missing values of time series with low sampling frequencies by using linear interpolation is a known method in the art. Before the effective filing date of the claimed invention, it would have been obvious to modify the combined method of Sánchez-Fernández/Mariia with the interpolation method disclosed by Yao to create uniformly sampled data. By using linear interpolation to fill in missing values of a time series with a low sampling frequency, time series can be synchronized and have consistent time intervals. By using time series with uniformly sampled data in data analysis, more consistent and accurate results can be achieved since filling in missing data can remove biases and incorrect assumptions about underlying patterns.
Allowable Subject Matter
Claims 8 and 15 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
Please note that claim 20, although containing similar limitations, has a different BRI due to being a method claim containing contingent limitations. See MPEP § 2111.04(ii).
Below are the closest cited references, each of which disclose various aspects of the claimed invention:
Bledsoe et al. (US 20180300737 A1) teaches a server that selects forecasting models based on time series characteristics, such as number of samples per model.
However, none of the prior art references of record—alone or in combination—disclose or suggest the combined features recited in the dependent claims, including specifically (for claim 8):
when the data dimensions of the time-series data have 2-40 time points per series, select LASSO, when the data dimensions of the time-series data have 40-5000 time points per series, select Random Forests
Although no particular limitation seems to be novel/non-obvious in itself, the combination of limitations recited are such that the claim—when considered as a whole—is non-obvious. Dependent Claim 15 recites similar limitations, therefore the same reasoning applies.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to PEDRO J MORALES whose telephone number is (571)272-6106. The examiner can normally be reached 8:30 AM - 6:00 PM.
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, MIRANDA M HUANG can be reached at (571)270-7092. 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.
/PEDRO J MORALES/Examiner, Art Unit 2124
/VINCENT GONZALES/Primary Examiner, Art Unit 2124
Prosecution Insights