| CPC G06F 18/29 (2023.01) [G06F 18/211 (2023.01); G06F 18/2163 (2023.01); G06N 20/00 (2019.01)] | 17 Claims |
|
1. A method for visualizing time series models, the method comprising:
obtaining, by a visualization engine, a time series dataset, wherein the time series dataset comprises a plurality of forecasting model features;
receiving, in response to obtaining the time series data set, instructions to generate a future forecast model for the time series dataset, wherein the future forecast model comprises:
a future prediction of the times series dataset; and
at least one forecasting model feature visualization output illustrating an effect of at least one of the forecasting model features on the future prediction,
wherein the at least one forecasting model feature visualization output to be generated is based on a selection of a forecasting model feature from among the forecasting model features and an explainability type, and information of the selection is included in the instructions, wherein the explainability type notifies the visualization engine that the at least one forecasting model feature visualization output is a visual decomposition graph;
generating, based on the instructions and by applying the time series data to a machine learning model comprising a time series forecaster, the future forecast model, wherein the time series data further comprises a primary time series covering a period of time, and the future prediction is generated for the primary time series, wherein generating the at least one forecasting model feature visualization output comprises generating a linear importance visualization graph,
wherein generating the linear importance visualization graph comprises:
generating, using the machine learning model:
a first forecast version of the primary time series over the period of time without the forecasting model feature and a second forecasting model feature among the forecasting model features;
a second forecast version of the primary time series over the period of time using only the forecasting model feature;
a third forecast version of the primary time series over the period of time using both the forecasting model feature and the second forecasting model feature;
plotting the primary time series and the first forecast version of the primary time series on the linear importance visualization graph,
introducing, using a first animated effect, the forecasting model feature onto the linear importance visualization graph,
executing a second animated effect to show a merging of the forecasting model feature with the first forecast version of the primary time series on the linear importance visualization graph,
replacing, using a third animated effect and after the forecasting model feature is overlapped with the first forecast version of the primary time series using the second animated effect, both the forecasting model feature and the first forecast version of the primary time series with the second forecast version of the primary time series on the linear importance visualization graph; and
displaying, based on the generating and to a user on a display of a computing device, the future forecast model.
|