DETAILED ACTIONS
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 .
Information Disclosure Statement
The information disclosure statements (IDS) submitted on 01/07/2025 and 08/09/2024. The submission is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statements are being considered by the examiner.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION. —The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
Claims 11, 17 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor, or for pre-AIA the applicant regards as the invention.
Regarding claim 11, claim 11 recites the limitation “I) repeating steps f-k until a stop criterion is reached” and it is indefinite. It is not clear from the claim language what is the “stop criterion”. Claim does not define the criterions and how the criterions are evaluated to stop the iteration. For the sake of prosecution, the examiner interpreted the limitation under its broadest reasonable interpretation as “stop criterion” as when the lower or upper limits” are met.
Regarding claim 17, claim 17 recites the limitation “I) repeating steps f-k until a stop criterion is reached” and it is indefinite. It is not clear from the claim language what is the “stop criterion”. Claim does not define the criterions and how the criterions are evaluated to stop the iteration. For the sake of prosecution, the examiner interpreted the limitation under its broadest reasonable interpretation as “stop criterion” as when the lower or upper limits” are met.
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- 22 are rejected under 35 U.S.C. 103 as being unpatentable over Bahena Tapia et al. (US 2020/0364607 A1, hereinafter Tapia, IDS ref.) and in view of WANG et al. (US 2021/0097453 A1, hereinafter Wang).
Regarding Claim 1, Tapia teaches,
A computer-implemented method for detecting an anomaly Tapia, Figure 1, Anomaly detection service 130, [0095] “anomaly detection services 130 includes logic for monitoring time-series data for anomalies”) in application monitoring data in a distributed computing environment (Tapia, [0039] “ a time series may be collected from one or more software and/or hardware resources and capture various performance attributes of the computing resources from which the sample data points were collected”) comprising:
receiving, by a computer processor (Tapia, Figure 1, Data collector 120, collector 120 receives data from agents 114a-j over one or more data communication networks”) training time series data for a given
variable, y, in the distributed computing environment (Tapia, Figure 2, [0053] Evaluation logic 206 receives the trained upper and lower limits and monitors a set of input dataset denoted D comprising data points {di, d2 , ... dj}. The input dataset D may be the same as the training dataset T or a different dataset, depending on the particular implementation”);
training, by the computer processor, a first probabilistic model using the
training time series data (Tapia, Figure 2, [0047], Anomaly detection services 130 may comprise logic for generating machine-learning (ML) models, monitoring time series signals for anomalies using the ML models”. [0050] “Anomaly detection services 130 includes training logic 200 and evaluation logic 206. Training logic 200 receives as input a training dataset denoted T comprising
data points { t1 , t2 , ... ti}”.),
where the first probabilistic model predicts a lower quantile value corresponding to a lower quantile for the given variable (Tapia, Figure 2, [0051] “Training logic 200 comprises quantile estimator 202 and quantile probability estimator 204. [0052] Quantile probability estimator 204 receives, as input, the size i (i.e., the number of data points) of training dataset T. In some embodiments, quantile probability estimator 204 computes two order statistics using the Wilks method. The order statistics may then be converted to corresponding quantile probabilities including a lower quantile probability quantile denoted LQ and an upper quantile
probability denoted UQ”);
receiving, by a computer processor, time series data for the given
variable, y, in the distributed computing environment (Tapia, Figure 5, [0097] The evaluation process includes receiving timeseries data for evaluation (operation 502). The set of timeseries data may be provided on-demand, periodically, or on
a continuous/streaming basis. For example, anomaly detection services 130 may monitor one or more streams of resource metrics associated with targets 112a-I”);
extracting, by the computer processor, a time series, (y0,... , Y n-i), for the
given variable, where the time series comprises the last n values of the time
series data (Tapia, [0057] In some embodiments, the anomaly detection system
maintains a sliding window oft-digest structures. The sliding window may comprise a shifting array of elements referred to herein as buckets. Each bucket includes a lower level t-digest structure, which may be represented as a set of
centroids. A default value of 20 buckets with a maximum of 100 centroids each was tested and observed to yield scalable and accurate results. Using this value, the maximum number of centroids across all buckets was 2000. However, the
number of centroids and t-digest structures may vary from implementation to implementation”);
`calculating, by the computer processor, a lower quantile forecast using the
time series and the first probabilistic model (Tapia, Figure 2, [0051] “quantile probability estimator 204 computes two order statistics using the Wilks method. The order statistics may then be converted to corresponding quantile probabilities including a lower quantile probability quantile denoted LQ”);
calculating, by the computer processor, a low threshold for the given
variable and a high threshold for the given variable based on the lower quantile
forecast and the higher quantile forecast (Tapia, Figure 2, [0052] “quantile probability estimator 204 queries quantile estimator 202 for the corresponding quantiles,
which may be determined as a function of the sliding window oft-digest structures. These quantiles may be used as the lower and upper limits in the trained anomaly detection model”);
receiving, by the computer processor, a new value, yn, for the given
variable (Tapia, Figure 5, 0097] The evaluation process includes receiving timeseries data for evaluation (operation 502). The set of timeseries
data may be provided on-demand, periodically, or on a continuous/streaming basis. For example, anomaly detection services 130 may monitor one or more streams of resource metrics associated with targets 112a-i.”) ;
comparing, by the computer processor, the new value of the given
variable to the low threshold and to the high threshold (Tapia, [0098]” In some embodiments, the evaluation process compares one or more data points within the time-series data to the trained upper and lower limits of the tolerance interval
to determine whether the limits have been crossed (operation 504). In the context of CPU utilization, for instance, the process may determine whether an evaluation data point is below the lower quantile or above the upper quantile”).
; and
marking, by the computer processor, the new value of the given variable
as an anomaly in response to the new value being one of less than the low
threshold or greater than the high threshold (Tapia, [0098] “If the
evaluation data point falls outside the conforming range of values that are between the two limits, then the process classifies the evaluation data point as anomalous. Conversely, if the evaluation data point is within the limits, then
the evaluation data point is not classified as anomalous”).
Tapia teaches in [0051] quantile probabilities including a lower quantile probability quantile denoted LQ and an upper quantile probability denoted UQ. However Tapia is silent on using multiple probabilistic models.
Tapia is silent on training, by the computer processor, a second probabilistic model using the training time series data,
where the second probabilistic model predicts a higher quantile value corresponding to a higher quantile for the given variable, and the higher quantile is larger than the lower quantile;
calculating, by the computer processor, a higher quantile forecast using
the time series and the second probabilistic model;
However, Wang teaches on training, by the computer processor, a second probabilistic model using the training time series data (Wang, [0006] “obtaining several results of point forecasting by utilizing multiple set of training data or multiple models first, and then performing quantile regression averaging on
those results of point forecasting to obtain a result of quantile forecasting”)
where the second probabilistic model predicts a higher quantile value corresponding to a higher quantile for the given variable, and the higher quantile is larger than the lower quantile; calculating, by the computer processor, a higher quantile forecast using the time series and the second probabilistic model (Wang,[0021] “ to obtain K=3M q-quantile forecasting models, wherein q is one of values in a preset set ranging between 0 and 1; establishing an optimization model with an objective function for minimizing the quantile loss for the second data set D2 by using the K=3M q-quantile forecasting models, determining a weight for each quantile regression model, to obtain a load ensemble forecasting model corresponding to the q-quantile; and repeating the step of obtaining quantile forecasting models and the step of calculating a load ensemble forecasting
model by traversing the preset set of values of q, to obtain a power load ensemble forecasting model with respect to different quantiles q for predicting the power load in the power system” NOTE: determining weight for each quantile( high or low values ) regression model provided better probabilistic value [0056] an optimal quantile ensemble forecasting model may be established based on the trained quantile probabilistic forecasting model, determining the weights for different quantile probabilistic forecasting methods the method according to the present disclosure may present certain weights for various individual forecasting
methods quickly”,
It would have been obvious to a person having ordinary skill in the art before the effective filing date to modify Tapia’s method for predicting anomalies to incorporate a multiple forecasting models as taught by Wang and obtain an accurate improved probabilistic load forecast (Wang, [0056]). It would have been obvious to a person of ordinary skill to include the well-known multiple models along with the quantile regression, and determining the weights for different quantile to implement an ensemble of multiple forecasting results in order to yield the predicted results of generating accurate forecast, yet with higher accuracy (KSR).
Regarding Claim 2, combination of Tapia and Wang teaches the computer-implemented method of claim 1,
Tapia further teaches further comprising:
appending, by the computer processor, the new value, Yn, of the given
variable, y, at the end of the time series; extracting, by the computer processor, a next time series, (Yv ... , Yn), for the given variable, y, where the next time series comprises the last n values of the time series; (Tapia, [0057] In some embodiments, the anomaly detection system
maintains a sliding window oft-digest structures. The sliding window may comprise a shifting array of elements referred to herein as buckets. Each bucket includes a lower level t-digest structure, which may be represented as a set of
centroids. A default value of 20 buckets with a maximum of 100 centroids each was tested and observed to yield scalable and accurate results. Using this value, the maximum number of centroids across all buckets was 2000. However, the
number of centroids and t-digest structures may vary from implementation to implementation. This approach constrains the memory requirements for the t-digest structures to a constant size, providing scalability in the memory
dimension. The space complexity of the t-digest may be expressed as O(k), where k is the number of centroids and each centroid uses 0(1) space. [0058] Once a bucket is full, a new bucket may be generated to accommodate new data points”);
calculating, by the computer processor, a lower quantile forecast using the
next time series and the first probabilistic model; Tapia, Figure 2, [0051] “quantile probability estimator 204 computes two order statistics using the Wilks method. The order statistics may then be converted to corresponding quantile probabilities including a lower quantile probability quantile denoted LQ”);
calculating, by the computer processor, a low threshold for the given
variable and a high threshold for the given variable based on the lower quantile
forecast and the higher quantile forecast Tapia, Figure 2, [0052] “quantile probability estimator 204 queries quantile estimator 202 for the corresponding quantiles,
which may be determined as a function of the sliding window oft-digest structures. These quantiles may be used as the lower and upper limits in the trained anomaly detection model”);
receiving, by the computer processor, a next new value, y n+ 1, for the
given variable (Tapia, Figure 5, 0097] The evaluation process includes receiving timeseries data for evaluation (operation 502). The set of timeseries
data may be provided on-demand, periodically, or on a continuous/streaming basis. For example, anomaly detection services 130 may monitor one or more streams of resource metrics associated with targets 112a-i.”) ;
comparing, by the computer processor, the next new value of the given
variable to the low threshold and to the high threshold (Tapia, [0098]” In some embodiments, the evaluation process compares one or more data points within the time-series data to the trained upper and lower limits of the tolerance interval
to determine whether the limits have been crossed (operation 504). In the context of CPU utilization, for instance, the process may determine whether an evaluation data point is below the lower quantile or above the upper quantile”).
; and
marking, by the computer processor, the next new value of the given
variable as an anomaly in response to the next new value being one of less than
the low threshold or greater than the high threshold. (Tapia, [0098] “If the
evaluation data point falls outside the conforming range of values that are between the two limits, then the process classifies the evaluation data point as anomalous. Conversely, if the evaluation data point is within the limits, then
the evaluation data point is not classified as anomalous”).
Tapia teaches in [0051] quantile probabilities including a lower quantile probability quantile denoted LQ and an upper quantile probability denoted UQ. However, Tapia is silent on using multiple probabilistic models.
Tapia is silent on calculating, by the computer processor, a higher quantile forecast using the next time series and the second probabilistic model.
However, Wang teaches calculating, by the computer processor, a higher quantile forecast using the next time series and the second probabilistic model (Wang, [0006] “obtaining several results of point forecasting by utilizing multiple set of training data or multiple models first, and then performing quantile regression averaging on those results of point forecasting to obtain a result of quantile forecasting”) .
(Wang,[0021] “ to obtain K=3M q-quantile forecasting models, wherein q is one of values in a preset set ranging between 0 and 1; establishing an optimization model with an objective function for minimizing the quantile loss for the second data set D2 by using the K=3M q-quantile forecasting models, determining a weight for each quantile regression model, to obtain a load ensemble forecasting model corresponding to the q-quantile”). NOTE: determining weight for each quantile( high or low values ) regression model provided better probabilistic value [0056] an optimal quantile ensemble forecasting model may be established based on the trained quantile probabilistic forecasting model, determining the weights for different quantile probabilistic forecasting methods the method according to the present disclosure may present certain weights for various individual forecasting methods quickly”,
It would have been obvious to a person having ordinary skill in the art before the effective filing date to modify Tapia’s method for predicting anomalies to incorporate a multiple forecasting models as taught by Wang and obtain an accurate improved probabilistic load forecast (Wang, [0056]). It would have been obvious to a person of ordinary skill to include the well-known multiple models along with the quantile regression, and determining the weights for different quantile to implement an ensemble of multiple forecasting results in order to yield the predicted results of generating accurate forecast, yet with higher accuracy (KSR).
Regarding Claim 3, combination of Tapia and Wang teaches the computer-implemented method of claim 1,
Tapia further teaches wherein calculating the low and high thresholds (Tapia, Figure 4, step 412) comprises:
determining a minimum value for the given variable using the lower
quantile forecast and the higher quantile forecast and determining a maximum
value for the given variable using the lower quantile forecast and the higher
quantile forecast, thereby defining a range for the given variable (Tapia, [0089] The process next computes the lower and upper quantile probabilities as a function of the lower and upper indexes, respectively (operation 406). The indexes may be
converted to quantile probabilities by dividing the value by the number of data points. In the preceding example with 120 data points, for instance, the lower quantile probability may be computed as 3/120=0.025 (2.5 th percentile), and the
upper quantile probability may be computed as 118/120=0. 9833 (98 ½rd percentile”);
generating random numbers evenly distributed between zero and one for each random number, mapping the given random number to a corresponding value in the range, thereby generating a set a value for the given variable. (Tapia, [0092] Once computed, the upper-layer t-digest structure may be used to determine the quantiles of interest. For example, the quantiles approximating the 0.025 and 0.9833 percentiles may be queried in the preceding example. It is noted that the t-digest structures provide relatively accurate approximations at the tail ends of the distribution, which is optimal for anomaly detection systems”); and determining the low threshold and the high threshold from the set of values for the given variable (Tapia, Figure 4, [0093] The process next builds the lower and upper limits of the anomaly detection model (operation 412). The lower and upper limits may be set to the quantiles returned at operation 410. These values may be updated in real-time as the distribution of data changes based on the input data stream”).
Regarding Claim 4, combination of Tapia and Wang teaches the computer-implemented method of claim 3,
Tapia further teaches wherein the mapping the given random number to a corresponding value in the range is by linear interpolation (Tapia, [0092] “The quantiles of interest may then be computed by interpolating between the two centroids. For instance, linear interpolation may be used to compute a sort of weighted averages between centroids”).
Regarding Claim 5, combination of Tapia and Wang teaches the computer-implemented method of claim 3,
Tapia further teaches where the low threshold is set to a quantile value lower than the lower quantile value and the high threshold is set to a quantile value higher than the higher quantile value (Tapia, [0098] for instance, the process may determine whether an evaluation data point is below the lower quantile or above the upper quantile. If the evaluation data point falls outside the conforming range of
values that are between the two limits”).
Regarding Claim 6, combination of Tapia and Wang teaches the computer-implemented method of claim 1,
Tapia further teaches wherein at least one of the first probabilistic model and the second probabilistic model (Tapia, Figure 2, [0047], Anomaly detection services 130 may comprise logic for generating machine-learning (ML) models, monitoring time series signals for anomalies using the ML models”. Probability estimator 204, figure 2. “)
Tapia is silent on the model is further defined as a linear regression model.
However, Wang teaches further defined as a linear regression model. (Wang, [0008] The methods for generating various forecasting models may applying a specific forecasting model to different training sets, so as to obtain diverse parameters of the models; or for the same set of training data, to train multiple forecasting models such as the linear regression model”)
It would have been obvious to a person having ordinary skill in the art before the effective filing date to modify Tapia’s method for predicting anomalies to incorporate multiple forecasting linear regression models as taught by Wang and obtain an accurate improved probabilistic load forecast (Wang, [0056]). It would have been obvious to a person of ordinary skill to include the well-known multiple models along with the quantile regression, and determining the weights for different quantile to implement an ensemble of multiple forecasting results in order to yield the predicted results of generating accurate forecast, yet with higher accuracy (KSR).
Regarding Claim 7, combination of Tapia and Wang teaches the computer-implemented method of claim 1,
Tapia further teaches training at least one of the first probabilistic model for anomalies using the ML models”).
Tapia is silent on using training the second probabilistic model using a Regularized Smoothed Iterative Least Square method.
However, Wang teaches using training the second probabilistic model (Wang, [0006] “obtaining several results of point forecasting by utilizing multiple set of training data or multiple models first, and then performing quantile regression averaging on those results of point forecasting to obtain a result of quantile forecasting”) , using a Regularized Smoothed Iterative Least Square method (Wang, [0011] Quantile regression mainly derive a regression model for estimating the conditional quantile of the output variables. Compared with the traditional least square regression”).
It would have been obvious to a person having ordinary skill in the art before the effective filing date to modify Tapia’s method for predicting anomalies to incorporate a multiple forecasting models as taught by Wang and obtain an accurate improved probabilistic load forecast (Wang, [0056]). It would have been obvious to a person of ordinary skill to include the well-known multiple models along with the quantile regression, and determining the weights for different quantile to implement an ensemble of multiple forecasting results in order to yield the predicted results of generating accurate forecast, yet with higher accuracy (KSR).
Regarding Claim 8, combination of Tapia and Wang teaches the computer-implemented method of claim 7,
Tapia is silent on where training at least one of the first probabilistic model and the second probabilistic model minimizes the pinball loss.
However, Wang teaches where training at least one of the first probabilistic model and the second probabilistic model minimizes the pinball loss (Wang, [0020] “Another object of the present disclosure is to provide a quantile probabilistic load ensemble forecasting with an objective function for minimizing the pinball loss, to further improve the accuracy of the quantile probabilistic short-term load forecasting”)
It would have been obvious to a person having ordinary skill in the art before the effective filing date to modify Tapia’s method for predicting anomalies to incorporate a multiple forecasting models training with time series data to minimize the pinball loss as taught by Wang and obtain an accurate improved probabilistic load forecast (Wang, [0014], [0020],[0056]). It would have been obvious to a person of ordinary skill to include the well-known multiple models along with the quantile regression, and determining the weights for different quantile to implement an ensemble of multiple forecasting results in order to yield the predicted results of generating accurate forecast, yet with higher accuracy (KSR).
Regarding Claim 9, combination of Tapia and Wang teaches the computer-implemented method of claim 1,
Tapia is silent on further comprises training at least one of the first probabilistic model and the second probabilistic model using a Coordinate Descent method.
However, Wang teaches comprises training at least one of the first probabilistic model and the second probabilistic model using a Coordinate Descent method (Wang, teaches the coordinate descent algorithm for quantile regression see the following paragraph [0011] Quantile regression mainly derive a regression model for estimating the conditional quantile of the output variables. Compared with the traditional least square regression, the quantile regression may provide more detailed information for the uncertainty of the output variables. The quantile regression may be described as a typical optimization model, as expressed by
PNG
media_image1.png
48
219
media_image1.png
Greyscale
the following equation:
[0012] In the above equation, i represents an index of a training sample for the model, N represents a total number of the training samples for the model, x, represents an input of the i,h training sample, y, represents an output of the i,h training sample, q represents quantile to be regressed and has a value between O and 1, ~( q) represents a parameter to be estimated for the quantile regression model of the quantile q, and pq represents a loss function for the quantile regression of the quantile q”).
It would have been obvious to a person having ordinary skill in the art before the effective filing date to modify Tapia’s method for predicting anomalies to incorporate a multiple forecasting models training with time series data to minimize the pinball loss as taught by Wang and obtain an accurate improved probabilistic load forecast (Wang, [0014], [0020],[0056]). It would have been obvious to a person of ordinary skill to include the well-known multiple models along with the quantile regression, and determining the weights for different quantile to implement an ensemble of multiple forecasting results in order to yield the predicted results of generating accurate forecast, yet with higher accuracy (KSR).
Regarding Claim 10, combination of Tapia and Wang teaches the computer-implemented method of claim 1,
Tapia further teaches wherein the lower quantile and the higher quantile are symmetrical in relation to the median. (Tapia, Figure 4, step 410, [0091] The process further determines the quantiles from the t-digest distribution approximation ( operation 410).[0092] A pair of centroids enclosing the requested quantile probability may be determined (…) where the value p represents a fraction of data up to the centroid in question (also referred to as the quantile for the approximate mean of the centroid c,) and is approximated for centroid c, by summing the weights for all of the centroids in the t-digest structure before c,. The quantiles of
interest may then be computed by interpolating between the two centroids. For instance, linear interpolation may be used to compute a sort of weighted averages between centroids “)
Regarding Claim 11, Tapia teaches,
A computer-implemented method for detecting an anomaly (Tapia, Figure 1, Anomaly detection service 130, [0095] “anomaly detection services 130 includes logic for monitoring time-series data for anomalies”) in application monitoring data in a distributed computing environment (Tapia, [0039] “ a time series may be collected from one or more software and/or hardware resources and capture various performance attributes of the computing resources from which the sample data points were collected”) comprising:
a) receiving, by a computer processor (Tapia, Figure 1, Data collector 120, collector 120 receives data from agents 114a-j over one or more data communication networks”) training time series data for a given variable, y, in the distributed computing environment (Tapia, Figure 2, [0053] Evaluation logic 206 receives the trained upper and lower limits and monitors a set of input dataset denoted D comprising data points {di, d2 , ... dj}. The input dataset D may be the same as the training dataset T or a different dataset, depending on the particular implementation”);
b) training, by the computer processor, a first probabilistic model using the
training time series data (Tapia, Figure 2, [0047], Anomaly detection services 130 may comprise logic for generating machine-learning (ML) models, monitoring time series signals for anomalies using the ML models”. [0050] “Anomaly detection services 130 includes training logic 200 and evaluation logic 206. Training logic 200 receives as input a training dataset denoted T comprising
data points { t1 , t2 , ... ti}”.),where the first probabilistic model predicts a lower quantile value corresponding to a lower quantile for the given variable (Tapia, Figure 2, [0051] “Training logic 200 comprises quantile estimator 202 and quantile probability estimator 204. [0052] Quantile probability estimator 204 receives, as input, the size i (i.e., the number of data points) of training dataset T. In some embodiments, quantile probability estimator 204 computes two order statistics using the Wilks method. The order statistics may then be converted to corresponding quantile probabilities including a lower quantile probability quantile denoted LQ and an upper quantile probability denoted UQ”);
d) receiving, by a computer processor, time series data for the given
variable, y, in the distributed computing environment (Tapia, Figure 5, [0097] The evaluation process includes receiving timeseries data for evaluation (operation 502). The set of timeseries data may be provided on-demand, periodically, or on
a continuous/streaming basis. For example, anomaly detection services 130 may monitor one or more streams of resource metrics associated with targets 112a-I”);
e) computing multiple paths (Tapia, Figures 2-5) of forecasts by:
i) extracting, by the computer processor, a time series, (y0 , ... , Yn-i),
for the given variable, where the time series comprises the last n data points of
the time series data (Tapia, [0057] In some embodiments, the anomaly detection system maintains a sliding window oft-digest structures. The sliding window may comprise a shifting array of elements referred to herein as buckets. Each bucket includes a lower level t-digest structure, which may be represented as a set of
centroids. A default value of 20 buckets with a maximum of 100 centroids each was tested and observed to yield scalable and accurate results. Using this value, the maximum number of centroids across all buckets was 2000. However, the
number of centroids and t-digest structures may vary from implementation to implementation”);
ii) calculating, by the computer processor, a lower quantile forecast
using the time series and the first probabilistic model (Tapia, Figure 2, [0051] “quantile probability estimator 204 computes two order statistics using the Wilks method. The order statistics may then be converted to corresponding quantile probabilities including a lower quantile probability quantile denoted LQ”);
iv) determining a minimum value for the given variable using the
lower quantile forecast and the higher quantile forecast and determining a
maximum value for the given variable using the lower quantile forecast and the
higher quantile forecast (Tapia, Figure 2, [0052] “quantile probability estimator 204 queries quantile estimator 202 for the corresponding quantiles,
which may be determined as a function of the sliding window oft-digest structures. These quantiles may be used as the lower and upper limits in the trained anomaly detection model”);, thereby defining a range for the given variable;
v) generating a random number evenly distributed between zero
and one;(Tapia, [0033] “A q-quantile or just quantile, as used herein, refers
to a cut point for dividing a data distribution into intervals. (…) . A tolerance interval may be defined with a lower limit quantile probability and an upper limit quantile probability ( e.g., 0.5 and 0.95). The associated quantiles (e.g., the metric
values representing the 5th and 95th percentiles) for the upper and lower quantile probabilities, however, may vary depending on the distribution of values in the training dataset” also see [0092] ).
vi) generating a forecast value for the given variable by mapping
the random number to a corresponding value in the range (Tapia, Figure 5, step 504, [0098] In some embodiments, the evaluation process compares one or more data points within the time-series data to the trained upper and lower limits of the tolerance interval to determine whether the limits have been crossed (operation
504”);
vii) appending, by the computer processor, the forecast value of the
given variable, y, at the end of the time series (Tapia, [0079] In some embodiments, a tolerance interval is characterized by the parameters p and β. The parameter p, which is referred to as the "coverage percentage", represents the proportion of the total population of values that one wants to cover with the interval. The parameter β is the confidence level, which in this case represents the probability that the interval actually covers p percent of the population. The term "population" is used in a generic sense and may refer to the total set of possible values for a random variable x. For example, it may be a metric from a system where anomaly detection is desired. [0080] the coverage percentage p of interval { xr, xn-r+l} behaves like a random variable with PDF g(p)=Beta(p; n-2r+l, 2r) where n is the number of data points in the training dataset, and candidate index
r
ϵ
1
,
n
The result is that such distribution only depends on the sample size n and the selected index r.”) ;
viii) repeating steps i-vii until a stop criterion is reached (Tapia, Figure 5, [0102] The process further includes determining whether to continue monitoring the time-series data (operation 510). Monitoring may be stopped at any point on demand, based on predefined time limits, or based on any other criteria. The process may stream or periodically receive time-series data generated by targets ll2a-i for evaluation. The process may be repeated for remaining data points in the received timeseries dataset and/or as new time-series data is received to continue evaluating resource behavior within the computing environment”)
f) identifying, by the computer processor, quantiles in the multiple paths of
forecasts (Tapia, Figures 2-5, [0093] Figure 4, The process next builds the lower and upper limits of the anomaly detection model (operation 412). The lower
and upper limits may be set to the quantiles returned at operation 410. These values may be updated in real-time as the distribution of data changes based on the input data streams.); and
g) reporting the quantiles to the method customer. (Tapia, Figure 1, [ 0108] In some embodiments, anomaly detection services 130 includes an interface, such as a GUI, CLI, and/or API, for presenting and responding to detected anomalies. For
example, a GUI interface may present an interactive visualization to a user upon detecting an anomaly”)
Tapia teaches determining high and low quantile values and predicting anomaly based on the limits. In [0051] quantile probabilities including a lower quantile probability quantile denoted LQ and an upper quantile probability denoted UQ.[ 0092] Once computed, the upper-layer t-digest structure may be used to determine the quantiles of interest. For example, the quantiles approximating the 0.025 and 0.9833 percentiles may be queried in the preceding example. Tapia is silent on using multiple probabilistic models.
Tapia is silent on c) training, by the computer processor, a second probabilistic model using the training time series data, where the second probabilistic model predicts a higher quantile value corresponding to a higher quantile for the given variable,
and the higher quantile is larger than the lower quantile; iii) calculating, by the computer processor, a higher quantile forecast using the time series and the second probabilistic model;
However, Wang teaches on training, by the computer processor, a second probabilistic model using the training time series data (Wang, [0006] “obtaining several results of point forecasting by utilizing multiple set of training data or multiple models first, and then performing quantile regression averaging on
those results of point forecasting to obtain a result of quantile forecasting”)
where the second probabilistic model predicts a higher quantile value corresponding to a higher quantile for the given variable, and the higher quantile is larger than the lower quantile; calculating, by the computer processor, a higher quantile forecast using the time series and the second probabilistic model (Wang,[0021] “ to obtain K=3M q-quantile forecasting models, wherein q is one of values in a preset set ranging between 0 and 1; establishing an optimization model with an objective function for minimizing the quantile loss for the second data set D2 by using the K=3M q-quantile forecasting models, determining a weight for each quantile regression model, to obtain a load ensemble forecasting model corresponding to the q-quantile; and repeating the step of obtaining quantile forecasting models and the step of calculating a load ensemble forecasting model by traversing the preset set of values of q, to obtain a power load ensemble forecasting model with respect to different quantiles q for predicting the power load in the power system” NOTE: determining weight for each quantile( high or low values ) regression model provided better probabilistic value [0056] an optimal quantile ensemble forecasting model may be established based on the trained quantile probabilistic forecasting model, determining the weights for different quantile probabilistic forecasting methods the method according to the present disclosure may present certain weights for various individual forecasting
methods quickly”).
It would have been obvious to a person having ordinary skill in the art before the effective filing date to modify Tapia’s method for predicting anomalies to incorporate a multiple forecasting models as taught by Wang and obtain an accurate improved probabilistic load forecast (Wang, [0056]). It would have been obvious to a person of ordinary skill to include the well-known multiple models along with the quantile regression, and determining the weights for different quantile to implement an ensemble of multiple forecasting results in order to yield the predicted results of generating accurate forecast, yet with higher accuracy (KSR).
Regarding Claim 12, combination of Tapia and Wang teaches the method of claim 11,
Tapia further teaches wherein at least one of the first probabilistic model and the second probabilistic model (Tapia, Figure 2, [0047], Anomaly detection services 130 may comprise logic for generating machine-learning (ML) models, monitoring time series signals for anomalies using the ML models” . Probability estimator 204, figure 2. “)
Tapia is silent on the model is further defined as a linear regression model.
However, Wang teaches further defined as a linear regression model. (Wang, [0008] The methods for generating various forecasting models may applying a specific forecasting model to different training sets, so as to obtain diverse parameters of the models; or for the same set of training data, to train multiple forecasting models such as the linear regression model”)
It would have been obvious to a person having ordinary skill in the art before the effective filing date to modify Tapia’s method for predicting anomalies to incorporate multiple forecasting linear regression models as taught by Wang and obtain an accurate improved probabilistic load forecast (Wang, [0056]). It would have been obvious to a person of ordinary skill to include the well-known multiple models along with the quantile regression, and determining the weights for different quantile to implement an ensemble of multiple forecasting results in order to yield the predicted results of generating accurate forecast, yet with higher accuracy (KSR).
.
Regarding Claim 13, combination of Tapia and Wang teaches the method of claim 11,
Tapia further teaches training at least one of the first probabilistic model for anomalies using the ML models”).
Tapia is silent on using training the second probabilistic model using a Regularized Smoothed Iterative Least Square method.
However, Wang teaches using training the second probabilistic model (Wang, [0006] “obtaining several results of point forecasting by utilizing multiple set of training data or multiple models first, and then performing quantile regression averaging on those results of point forecasting to obtain a result of quantile forecasting”) , using a Regularized Smoothed Iterative Least Square method (Wang, [0011] Quantile regression mainly derive a regression model for estimating the conditional quantile of the output variables. Compared with the traditional least square regression”).
It would have been obvious to a person having ordinary skill in the art before the effective filing date to modify Tapia’s method for predicting anomalies to incorporate a multiple forecasting models as taught by Wang and obtain an accurate improved probabilistic load forecast (Wang, [0056]). It would have been obvious to a person of ordinary skill to include the well-known multiple models along with the quantile regression, and determining the weights for different quantile to implement an ensemble of multiple forecasting results in order to yield the predicted results of generating accurate forecast, yet with higher accuracy (KSR).
Regarding Claim 14, combination of Tapia and Wang teaches the method of claim 13,
Tapia is silent on where training at least one of the first probabilistic model and the second probabilistic model minimizes the pinball loss.
However, Wang teaches where training at least one of the first probabilistic model and the second probabilistic model minimizes the pinball loss (Wang, [0020] “Another object of the present disclosure is to provide a quantile probabilistic load ensemble forecasting with an objective function for minimizing the pinball loss, to further improve the accuracy of the quantile probabilistic short-term load forecasting”)
It would have been obvious to a person having ordinary skill in the art before the effective filing date to modify Tapia’s method for predicting anomalies to incorporate a multiple forecasting models training with time series data to minimize the pinball loss as taught by Wang and obtain an accurate improved probabilistic load forecast (Wang, [0014], [0020],[0056]). It would have been obvious to a person of ordinary skill to include the well-known multiple models along with the quantile regression, and determining the weights for different quantile to implement an ensemble of multiple forecasting results in order to yield the predicted results of generating accurate forecast, yet with higher accuracy (KSR).
Regarding Claim 15, combination of Tapia and Wang teaches the method of claim 11,
Tapia is silent on further comprises training at least one of the first probabilistic model and the second probabilistic model using a Coordinate Descent method.
However, Wang teaches comprises training at least one of the first probabilistic model and the second probabilistic model using a Coordinate Descent method (Wang, teaches the coordinate descent algorithm for quantile regression see the following paragraph [0011] Quantile regression mainly derive a regression model for estimating the conditional quantile of the output variables. Compared with the traditional least square regression, the quantile regression may provide more detailed information for the uncertainty of the output variables. The quantile regression may be described as a typical optimization model, as expressed by
PNG
media_image1.png
48
219
media_image1.png
Greyscale
the following equation:
[0012] In the above equation, i represents an index of a training sample for the model, N represents a total number of the training samples for the model, x, represents an input of the i,h training sample, y, represents an output of the i,h training sample, q represents quantile to be regressed and has a value between O and 1, ~( q) represents a parameter to be estimated for the quantile regression model of the quantile q, and pq represents a loss function for the quantile regression of the quantile q”).
It would have been obvious to a person having ordinary skill in the art before the effective filing date to modify Tapia’s method for predicting anomalies to incorporate a multiple forecasting models training with time series data to minimize the pinball loss as taught by Wang and obtain an accurate improved probabilistic load forecast (Wang, [0014], [0020],[0056]). It would have been obvious to a person of ordinary skill to include the well-known multiple models along with the quantile regression, and determining the weights for different quantile to implement an ensemble of multiple forecasting results in order to yield the predicted results of generating accurate forecast, yet with higher accuracy (KSR).
Regarding Claim 16, combination of Tapia and Wang teaches the method of claim 11,
Tapia further teaches wherein the lower quantile and the higher quantile are symmetrical in relation to the median. (Tapia, Figure 4, step 410, [0091] The process further determines the quantiles from the t-digest distribution approximation ( operation 410).[0092] A pair of centroids enclosing the requested quantile probability may be determined (…) where the value p represents a fraction of data up to the centroid in question (also referred to as the quantile for the approximate mean of the centroid c,) and is approximated for centroid c, by summing the weights for all of the centroids in the t-digest structure before c,. The quantiles of
interest may then be computed by interpolating between the two centroids. For instance, linear interpolation may be used to compute a sort of weighted averages between centroids “)
Regarding Claim 17,
Tapia teaches,
A computer-implemented method for detecting an anomaly (Tapia, Figure 1, Anomaly detection service 130, [0095] “anomaly detection services 130 includes logic for monitoring time-series data for anomalies”) in application monitoring data in a distributed computing environment (Tapia, [0039] “ a time series may be collected from one or more software and/or hardware resources and capture various performance attributes of the computing resources from which the sample data points were collected”) comprising:
a) receiving, by a computer processor (Tapia, Figure 1, Data collector 120, collector 120 receives data from agents 114a-j over one or more data communication networks”) training time series data for a given variable, y, in the distributed computing environment (Tapia, Figure 2, [0053] Evaluation logic 206 receives the trained upper and lower limits and monitors a set of input dataset denoted D comprising data points {di, d2 , ... dj}. The input dataset D may be the same as the training dataset T or a different dataset, depending on the particular implementation”);
b) training, by the computer processor, a first probabilistic model using the
training time series data (Tapia, Figure 2, [0047], Anomaly detection services 130 may comprise logic for generating machine-learning (ML) models, monitoring time series signals for anomalies using the ML models”. [0050] “Anomaly detection services 130 includes training logic 200 and evaluation logic 206. Training logic 200 receives as input a training dataset denoted T comprising
data points { t1 , t2 , ... ti}”.),where the first probabilistic model predicts a lower quantile value corresponding to a lower quantile for the given variable (Tapia, Figure 2, [0051] “Training logic 200 comprises quantile estimator 202 and quantile probability estimator 204. [0052] Quantile probability estimator 204 receives, as input, the size i (i.e., the number of data points) of training dataset T. In some embodiments, quantile probability estimator 204 computes two order statistics using the Wilks method. The order statistics may then be converted to corresponding quantile probabilities including a lower quantile probability quantile denoted LQ and an upper quantile probability denoted UQ”);
d) receiving, by a computer processor, time series data for the given
variable, y, in the distributed computing environment (Tapia, Figure 5, [0097] The evaluation process includes receiving timeseries data for evaluation (operation 502). The set of timeseries data may be provided on-demand, periodically, or on
a continuous/streaming basis. For example, anomaly detection services 130 may monitor one or more streams of resource metrics associated with targets 112a-I”);
e) computing multiple paths (Tapia, Figures 2-5) of forecasts by:
i) extracting, by the computer processor, a time series, (y0 , ... , Yn-i),
for the given variable, where the time series comprises the last n data points of
the time series data (Tapia, [0057] In some embodiments, the anomaly detection system maintains a sliding window oft-digest structures. The sliding window may comprise a shifting array of elements referred to herein as buckets. Each bucket includes a lower level t-digest structure, which may be represented as a set of
centroids. A default value of 20 buckets with a maximum of 100 centroids each was tested and observed to yield scalable and accurate results. Using this value, the maximum number of centroids across all buckets was 2000. However, the
number of centroids and t-digest structures may vary from implementation to implementation”);
ii) calculating, by the computer processor, a lower quantile forecast
using the time series and the first probabilistic model (Tapia, Figure 2, [0051] “quantile probability estimator 204 computes two order statistics using the Wilks method. The order statistics may then be converted to corresponding quantile probabilities including a lower quantile probability quantile denoted LQ”);
iv) determining a minimum value for the given variable using the
lower quantile forecast and the higher quantile forecast and determining a
maximum value for the given variable using the lower quantile forecast and the
higher quantile forecast (Tapia, Figure 2, [0052] “quantile probability estimator 204 queries quantile estimator 202 for the corresponding quantiles,
which may be determined as a function of the sliding window oft-digest structures. These quantiles may be used as the lower and upper limits in the trained anomaly detection model”);, thereby defining a range for the given variable;
v) generating a random number evenly distributed between zero
and one;(Tapia, [0033] “A q-quantile or just quantile, as used herein, refers
to a cut point for dividing a data distribution into intervals. (…) . A tolerance interval may be defined with a lower limit quantile probability and an upper limit quantile probability ( e.g., 0.5 and 0.95). The associated quantiles (e.g., the metric
values representing the 5th and 95th percentiles) for the upper and lower quantile probabilities, however, may vary depending on the distribution of values in the training dataset” also see [0092] ).
vi) generating a forecast value for the given variable by mapping
the random number to a corresponding value in the range (Tapia, Figure 5, step 504, [0098] In some embodiments, the evaluation process compares one or more data points within the time-series data to the trained upper and lower limits of the tolerance interval to determine whether the limits have been crossed (operation
504”);
vii) appending, by the computer processor, the forecast value of the
given variable, y, at the end of the time series (Tapia, [0079] In some embodiments, a tolerance interval is characterized by the parameters p and β. The parameter p, which is referred to as the "coverage percentage", represents the proportion of the total population of values that one wants to cover with the interval. The parameter β is the confidence level, which in this case represents the probability that the interval actually covers p percent of the population. The term "population" is used in a generic sense and may refer to the total set of possible values for a random variable x. For example, it may be a metric from a system where anomaly detection is desired. [0080] the coverage percentage p of interval { xr, xn-r+l} behaves like a random variable with PDF g(p)=Beta(p; n-2r+l, 2r) where n is the number of data points in the training dataset, and candidate index
r
ϵ
1
,
n
The result is that such distribution only depends on the sample size n and the selected index r.”) ;
viii) repeating steps i-vii until a stop criterion is reached (Tapia, Figure 5, [0102] The process further includes determining whether to continue monitoring the time-series data (operation 510). Monitoring may be stopped at any point on demand, based on predefined time limits, or based on any other criteria. The process may stream or periodically receive time-series data generated by targets ll2a-i for evaluation. The process may be repeated for remaining data points in the received timeseries dataset and/or as new time-series data is received to continue evaluating resource behavior within the computing environment”)
f) identifying, by the computer processor, quantiles in the multiple paths of
forecasts (Tapia, Figures 2-5, [0093] Figure 4, The process next builds the lower and upper limits of the anomaly detection model (operation 412). The lower
and upper limits may be set to the quantiles returned at operation 410. These values may be updated in real-time as the distribution of data changes based on the input data streams.); and
g) reporting the quantiles to the method customer. (Tapia, Figure 1, [ 0108] In some embodiments, anomaly detection services 130 includes an interface, such as a GUI, CLI, and/or API, for presenting and responding to detected anomalies. For
example, a GUI interface may present an interactive visualization to a user upon detecting an anomaly”)
Tapia teaches determining high and low quantile values and predicting anomaly based on the limits. In [0051] quantile probabilities including a lower quantile probability quantile denoted LQ and an upper quantile probability denoted UQ.[ 0092] Once computed, the upper-layer t-digest structure may be used to determine the quantiles of interest. For example, the quantiles approximating the 0.025 and 0.9833 percentiles may be queried in the preceding example. Tapia is silent on using multiple probabilistic models.
Tapia is silent on c) training, by the computer processor, a second probabilistic model using the training time series data, where the second probabilistic model predicts a higher quantile value corresponding to a higher quantile for the given variable,
and the higher quantile is larger than the lower quantile; iii) calculating, by the computer processor, a higher quantile forecast using the time series and the second probabilistic model;
However, Wang teaches on training, by the computer processor, a second probabilistic model using the training time series data (Wang, [0006] “obtaining several results of point forecasting by utilizing multiple set of training data or multiple models first, and then performing quantile regression averaging on
those results of point forecasting to obtain a result of quantile forecasting”)
where the second probabilistic model predicts a higher quantile value corresponding to a higher quantile for the given variable, and the higher quantile is larger than the lower quantile; calculating, by the computer processor, a higher quantile forecast using the time series and the second probabilistic model (Wang,[0021] “ to obtain K=3M q-quantile forecasting models, wherein q is one of values in a preset set ranging between 0 and 1; establishing an optimization model with an objective function for minimizing the quantile loss for the second data set D2 by using the K=3M q-quantile forecasting models, determining a weight for each quantile regression model, to obtain a load ensemble forecasting model corresponding to the q-quantile; and repeating the step of obtaining quantile forecasting models and the step of calculating a load ensemble forecasting model by traversing the preset set of values of q, to obtain a power load ensemble forecasting model with respect to different quantiles q for predicting the power load in the power system” NOTE: determining weight for each quantile( high or low values ) regression model provided better probabilistic value [0056] an optimal quantile ensemble forecasting model may be established based on the trained quantile probabilistic forecasting model, determining the weights for different quantile probabilistic forecasting methods the method according to the present disclosure may present certain weights for various individual forecasting
methods quickly”).
It would have been obvious to a person having ordinary skill in the art before the effective filing date to modify Tapia’s method for predicting anomalies to incorporate a multiple forecasting models as taught by Wang and obtain an accurate improved probabilistic load forecast (Wang, [0056]). It would have been obvious to a person of ordinary skill to include the well-known multiple models along with the quantile regression, and determining the weights for different quantile to implement an ensemble of multiple forecasting results in order to yield the predicted results of generating accurate forecast, yet with higher accuracy (KSR).
Regarding Claim 18, combination of Tapia and Wang teaches the method of claim 17,
Tapia further teaches wherein at least one of the first probabilistic model and the second probabilistic model (Tapia, Figure 2, [0047], Anomaly detection services 130 may comprise logic for generating machine-learning (ML) models, monitoring time series signals for anomalies using the ML models” . Probability estimator 204, figure 2. “)
Tapia is silent on the model is further defined as a linear regression model.
However, Wang teaches further defined as a linear regression model. (Wang, [0008] The methods for generating various forecasting models may applying a specific forecasting model to different training sets, so as to obtain diverse parameters of the models; or for the same set of training data, to train multiple forecasting models such as the linear regression model”)
It would have been obvious to a person having ordinary skill in the art before the effective filing date to modify Tapia’s method for predicting anomalies to incorporate multiple forecasting linear regression models as taught by Wang and obtain an accurate improved probabilistic load forecast (Wang, [0056]). It would have been obvious to a person of ordinary skill to include the well-known multiple models along with the quantile regression, and determining the weights for different quantile to implement an ensemble of multiple forecasting results in order to yield the predicted results of generating accurate forecast, yet with higher accuracy (KSR).
Regarding Claim 19, combination of Tapia and Wang teaches the method of claim 17,
Tapia further teaches training at least one of the first probabilistic model for anomalies using the ML models”).
Tapia is silent on using training the second probabilistic model using a Regularized Smoothed Iterative Least Square method.
However, Wang teaches using training the second probabilistic model (Wang, [0006] “obtaining several results of point forecasting by utilizing multiple set of training data or multiple models first, and then performing quantile regression averaging on those results of point forecasting to obtain a result of quantile forecasting”) , using a Regularized Smoothed Iterative Least Square method (Wang, [0011] Quantile regression mainly derive a regression model for estimating the conditional quantile of the output variables. Compared with the traditional least square regression”).
It would have been obvious to a person having ordinary skill in the art before the effective filing date to modify Tapia’s method for predicting anomalies to incorporate a multiple forecasting models as taught by Wang and obtain an accurate improved probabilistic load forecast (Wang, [0056]). It would have been obvious to a person of ordinary skill to include the well-known multiple models along with the quantile regression, and determining the weights for different quantile to implement an ensemble of multiple forecasting results in order to yield the predicted results of generating accurate forecast, yet with higher accuracy (KSR).
Regarding Claim 20, combination of Tapia and Wang teaches the method of claim 19,
Tapia is silent on where training at least one of the first probabilistic model and the second probabilistic model minimizes the pinball loss.
However, Wang teaches where training at least one of the first probabilistic model and the second probabilistic model minimizes the pinball loss (Wang, [0020] “Another object of the present disclosure is to provide a quantile probabilistic load ensemble forecasting with an objective function for minimizing the pinball loss, to further improve the accuracy of the quantile probabilistic short-term load forecasting”)
It would have been obvious to a person having ordinary skill in the art before the effective filing date to modify Tapia’s method for predicting anomalies to incorporate a multiple forecasting models training with time series data to minimize the pinball loss as taught by Wang and obtain an accurate improved probabilistic load forecast (Wang, [0014], [0020],[0056]). It would have been obvious to a person of ordinary skill to include the well-known multiple models along with the quantile regression, and determining the weights for different quantile to implement an ensemble of multiple forecasting results in order to yield the predicted results of generating accurate forecast, yet with higher accuracy (KSR).
Regarding Claim 21, combination of Tapia and Wang teaches the method of claim 17,
Tapia is silent on further comprises training at least one of the first probabilistic model and the second probabilistic model using a Coordinate Descent method.
However, Wang teaches comprises training at least one of the first probabilistic model and the second probabilistic model using a Coordinate Descent method (Wang, teaches the coordinate descent algorithm for quantile regression see the following paragraph [0011] Quantile regression mainly derive a regression model for estimating the conditional quantile of the output variables. Compared with the traditional least square regression, the quantile regression may provide more detailed information for the uncertainty of the output variables. The quantile regression may be described as a typical optimization model, as expressed by
PNG
media_image1.png
48
219
media_image1.png
Greyscale
the following equation:
[0012] In the above equation, i represents an index of a training sample for the model, N represents a total number of the training samples for the model, x, represents an input of the i,h training sample, y, represents an output of the i,h training sample, q represents quantile to be regressed and has a value between O and 1, ~( q) represents a parameter to be estimated for the quantile regression model of the quantile q, and pq represents a loss function for the quantile regression of the quantile q”).
It would have been obvious to a person having ordinary skill in the art before the effective filing date to modify Tapia’s method for predicting anomalies to incorporate a multiple forecasting models training with time series data to minimize the pinball loss as taught by Wang and obtain an accurate improved probabilistic load forecast (Wang, [0014], [0020],[0056]). It would have been obvious to a person of ordinary skill to include the well-known multiple models along with the quantile regression, and determining the weights for different quantile to implement an ensemble of multiple forecasting results in order to yield the predicted results of generating accurate forecast, yet with higher accuracy (KSR).
.
Regarding Claim 22, combination of Tapia and Wang teaches the method of claim 17,
Tapia further teaches wherein the lower quantile and the higher quantile are symmetrical in relation to the median. (Tapia, Figure 4, step 410, [0091] The process further determines the quantiles from the t-digest distribution approximation ( operation 410).[0092] A pair of centroids enclosing the requested quantile probability may be determined (…) where the value p represents a fraction of data up to the centroid in question (also referred to as the quantile for the approximate mean of the centroid c,) and is approximated for centroid c, by summing the weights for all of the centroids in the t-digest structure before c,. The quantiles of
interest may then be computed by interpolating between the two centroids. For instance, linear interpolation may be used to compute a sort of weighted averages between centroids”).
Conclusion
Citation of Pertinent Prior Art
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Singh et al. (US 2024/0005177 A1) recites “Monitoring may be performed for time series prediction models. Data to generate a new time series forecast may be received. A determination may be made that the data is associated with a previously generated time series forecast by a machine learning model. Performance metrics may be generated for the machine learning model according to a comparison of the data with the previously generated time series forecast. The performance metrics can then be provided for further analysis and action” (Abstract).
QIU et al. (US 2020/0311603 A1) The invention provides “A device receives historical data associated with multiple cloud computing environments, trains one or more machine learning models, with the historical data, to generate trained machine learning models that generate outputs, and trains a model with the outputs to generate a trained model. The device receives particular data, associated with a cloud computing environment, that includes data identifying usage of resources associated with the cloud computing environment, and processes the particular data, with the trained machine learning models, to generate anomaly scores indicating anomalous usage of the resources associated with the cloud computing environment. The device processes the one or more anomaly scores, with the trained model, to generate a final anomaly score indicating anomalous usage of at least one of the resources associated with the cloud computing environment, and performs one or more actions based on the final anomaly score” (Abstract).
Banubakode et al. (US 2022/0245526 A1) discloses “A server computer may receive and process a plurality of time series data to generate sparse datasets based on sparsity levels. The server computer applies a time series forecasting model to each respective subset of previous data points of the sparse datasets increasingly at the first time granularity to generate a set of prediction values and a set of residuals; applies a regression model to the set of the prediction residuals to generate a set of adjusted residuals for the sparse datasets; and generates a visualized explanation based on the
set of the prediction values and the set of adjusted residuals for one or more of the sparse datasets” (Abstract).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DILARA SULTANA whose telephone number is (571)272-3861. The examiner can normally be reached Mon-Fri, 9 AM-5:30 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, EMAN ALKAFAWI can be reached on (571) 272-4448. 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.
/DILARA SULTANA/Examiner, Art Unit 2858
06/10/2026
/EMAN A ALKAFAWI/Supervisory Patent Examiner, Art Unit 2858 6/15/2026
Prosecution Insights