Multivariable time series prediction method for multi-scale adaptive graph learning

Abstract

The invention discloses a multivariable time sequence prediction method for multi-scale adaptive graph learning. The multivariable time sequence prediction method comprises the following steps: inputting a training sample set into a multi-scale pyramid network to obtain a multi-scale initial subsequence set of each training sample; obtaining an adjacent matrix of each scale based on an adaptive graph learning module; inputting each time sequence and the adjacent matrix of the corresponding scale into a graph neural network at the same time to obtain a representation sequence of each time sequence, inputting the representation sequence into a time convolutional network to obtain a final sub-sequence of each scale, and inputting a multi-scale final sub-sequence set into a multi-scale fusion module, inputting the multi-scale fusion data of each training sample into a multi-layer convolutional neural network for mapping to obtain a multivariable time sequence predicted value of each training sample; determining model parameters based on the loss function so as to determine a multivariable time series prediction model of multi-scale adaptive graph learning; the method can accurately predict the predicted value of the selected scene.

Description

Technical field [0001] The present invention relates to the field of multivariate time sequence prediction, and more particularly to a multivariate time series prediction method for multi-scale adaptive graph learning. Background technique [0002] Multivariate time sequences are generally existed in a variety of realistic scenarios, such as urban traffic flow and home electricity consumption in urban neighborhoods. Multivariate Time Series Prediction is a method based on a set of historical observation time sequence predicting future trends. In recent years, it has been widely studied. It has broad application space, for example, can plan a better driving route according to the forecast of each intersection traffic flow; design investment strategies through multi-stock price forecasts of the recent stock market. [0003] Compared to single variable time sequence prediction, multivariate time series prediction requires simultaneous considering time correlation within a single var...

AI Technical Summary

Problems solved by technology

[0005] First, existing studies only consider temporal correlations on a single scale, and cannot effectively capture multi-scale temporal patterns (such as daily, weekly, monthly, and other specific periodic patterns, etc.)

Secondly, the existing research uses some prior knowledge or artificial experience to define the adjacency matrix with fixed weight to represent the correlation between variables, and it is difficult to effectively describe the implicit correlation between variables.

Images