US Patent Application 17804429. TECHNIQUES FOR GENERATING A MODEL FOR TIMESERIES DATA FORECASTING simplified abstract

From WikiPatents
Revision as of 15:09, 3 December 2023 by Wikipatents (talk | contribs) (Creating a new page)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

TECHNIQUES FOR GENERATING A MODEL FOR TIMESERIES DATA FORECASTING

Organization Name

Microsoft Technology Licensing, LLC==Inventor(s)==

[[Category:Chepuri Shri Krishna of Bangalore (IN)]]

[[Category:Swarnim Narayan of Bangalore (IN)]]

[[Category:Kiran Rama of Bangalore (IN)]]

[[Category:Ivan Barrientos of Seattle WA (US)]]

[[Category:Vijay Srinivas Agneeswaran of Bangalore (IN)]]

TECHNIQUES FOR GENERATING A MODEL FOR TIMESERIES DATA FORECASTING - A simplified explanation of the abstract

This abstract first appeared for US patent application 17804429 titled 'TECHNIQUES FOR GENERATING A MODEL FOR TIMESERIES DATA FORECASTING

Simplified Explanation

- The patent application describes a method for generating a model to forecast time series data. - The method involves using one or more layers, where each layer includes generating short range and long range outputs for each input in the time series data set. - The short range output is generated using a causal convolution process, which considers inputs from the time series data set that are associated with timestamps within a certain time threshold before the input timestamp. - The long range output is generated using a transformer process, which is based on the short range outputs from the causal convolution process. - The transformer process considers inputs from the time series data set that are associated with timestamps before the input timestamp. - The method aims to improve the accuracy of time series data forecasting by incorporating both short range and long range information from the time series data set.


Original Abstract Submitted

Described are examples for generating a model for forecasting time series data. For a timeseries data set, one or more layers can be provided, where each layer in the one or more layers includes, for each timeseries data input in at least a portion of multiple timeseries data inputs, generating, for the timeseries data input, a short range output from a causal convolution process that is based on timeseries data inputs from the timeseries data set that are associated with timestamps within a threshold time before the timestamp of the timeseries data input, and generating, for the timeseries data input, a long range output from a transformer process based on the short range outputs from the causal convolution process for each timeseries data input from at least the portion of the multiple timeseries data inputs that are associated with timestamps before the timestamp of the timeseries data input.