Localized temporal model forecasting

Patent metadata

Field	Value
Publication number	US-12067501-B2
Application number	US-202217712876-A
Country	US
Kind code	B2
Filing date	Apr 4, 2022
Priority date	Jan 14, 2016
Publication date	Aug 20, 2024
Grant date	Aug 20, 2024

Abstract

Disclosed herein are systems, computer-readable media, and methods related to modeling on multivariate time series data overlaid with event data. In particular, some examples involve selecting one or more historical time series data arrays similar to a recent time series data array and filtering the similar historical time series data arrays based on event data. Some examples can also involve training a localized temporal forecasting model using the filtered historical time series data arrays. Some examples can include building and/or training the localized temporal forecasting model at or near a time that a forecast is needed.

First claim

The invention claimed is:

1. A computing system comprising:
at least one processor;
a non-transitory computer-readable medium; and
program instructions stored on the non-transitory computer-readable medium that are executable by the at least one processor to cause the computing system to:
a) obtain a first time series data array representing a set of one or more parameters associated with an asset of a system;
b) identify one or more historical time series data arrays similar to the first time series data array, the one or more historical time series data arrays representing respective past values of the set of one or more parameters;
c) obtain first event data associated with the first time series data array and historical event data associated with each of the one or more historical time series data arrays;
d) filter the one or more historical time series data arrays based on a comparison of the first event data and the historical event data associated with each of the one or more historical time series data arrays to obtain one or more filtered historical time series data arrays;
e) train a time series forecast model using the one or more filtered historical time series data arrays as training data for a machine learning process that is executed by the computing system to produce the trained time series forecast model, and wherein the machine learning process employs recurrent neural network and/or support vector regression;
f) generate a forecast of a future value of at least one parameters of the set of one or more parameters using the trained time series forecast model;
g) detect a forecasted occurrence of an anonymous condition of the asset of the system based on the forecast of the future value of at least one parameter, and
h) control the asset of the system to affect the behavior of the asset of the system based on the anonymous condition as detected.

2. The computing system of claim 1 , wherein a)-f) are repeated based on one of a period of time expiring or receiving a request to generate a forecast.

3. The computing system of claim 1 , wherein the program instructions are further executable by the at least one processor to cause the computing system to:
detect a temporal change;
based on the detected change, repeat a) to obtain an updated first time series data array representing the set of one or more parameters at a time subsequent to a time of a previous first time series data array; and
repeat b)-f) based on the updated first time series data array.

4. The computing system of claim 1 , wherein the program instructions are further executable by the at least one processor to cause the computing system to:
detect a change in at least one parameters of the set of one or more parameters;
based on the detected change, repeat a) to obtain an updated first time series data array representing the set of one or more parameters at a time subsequent to a time of a previous first time series data array; and
repeat b)-f) based on the updated first time series data array.

5. The computing system of claim 1 , wherein the program instructions are further executable by the at least one processor to cause the computing system to transmit the forecast to another system.

6. The computing system of claim 1 , wherein the first event data indicates an event that has occurred at a time of the first time series data array and that is expected to affect behavior of the system subsequent to the time of the first time series data array.

7. The computing system of claim 1 , wherein the respective historical event data associated with each of the one or more filtered historical time series data arrays indicates an event similar to the first event.

8. A non-transitory computer-readable medium having program instructions stored thereon that are executable by at least one processor of a computing system to cause the computing system to:
a) obtain a first time series data array representing a set of one or more parameters associated with an asset of a system;
b) identify one or more historical time series data arrays similar to the first time series data array, the one or more historical time series data arrays representing respective past values of the set of one or more parameters;
c) obtain first event data associated with the first time series data array and historical event data associated with each of the one or more historical time series data arrays;
d) filter the one or more historical time series data arrays based on a comparison of the first event data and the historical event data associated with each of the one or more historical time series data arrays to obtain one or more filtered historical time series data arrays;
e) train a time series forecast model using the one or more filtered historical time series data arrays as training data for a machine learning process that is executed by the computing system to produce the trained time series forecast model, and wherein the machine learning process employs recurrent neural network and/or support vector regression;
f) generate a forecast of a future value of at least one parameters of the set of one or more parameters using the trained time series forecast model;
g) detect a forecasted occurrence of an anonymous condition the asset of the system based on the forecast of the future value of at least one parameter; and
h) control the asset of the system to affect the behavior of the asset of the system based on the anonymous condition as detected.

9. The non-transitory computer-readable medium of claim 8 , wherein a)-f) are repeated based on one of a period of time expiring or receiving a request to generate a forecast.

10. The non-transitory computer-readable medium of claim 8 , wherein the program instructions are further executable by the at least one processor to cause the computing system to:
detect a temporal change;
based on the detected change, repeat a) to obtain an updated first time series data array representing the set of one or more parameters at a time subsequent to a time of a previous first time series data array; and
repeat b)-f) based on the updated first time series data array.

11. The non-transitory computer-readable medium of claim 8 , wherein the program instructions are further executable by the at least one processor to cause the computing system to:
detect a change in at least one parameter of the set of one or more parameters;
based on the detected change, repeat a) to obtain an updated first time series data array representing the set of one or more parameters at a time subsequent to a time of a previous first time series data array; and
repeat b)-f) based on the updated first time series data array.

12. The non-transitory computer-readable medium of claim 8 , wherein the program instructions are further executable by the at least one processor to cause the computing system to transmit the forecast to another system.

13. The non-transitory computer-readable medium of claim 8 , wherein the first event data indicates an event that has occurred at a time of the first time series data array and that is expected to affect behavior of the system subsequent to the time of the first time series data array.

14. The non-transitory computer-readable medium of claim 8 , wherein the respective historical event data associated with each of the one or more filtered historical time series data arrays indicates an event similar to the first event.

15. A computer-implemented method comprising:
a) obtaining a first time series data array representing a set of one or more parameters associated with an asset of a system;
b) ident

Assignees

• Uptake Tech Inc

Inventors

• Herzog James

Classifications

• G06N5/045

  Explanation of inference; Explainable artificial intelligence [XAI]; Interpretable artificial intelligence · CPC title

• G06N20/00 Primary

  Machine learning · CPC title

• G06N5/04 Primary

  Inference or reasoning models · CPC title