Construction method of multivariable long-sequence time sequence prediction model based on Transform framework

Abstract

The invention discloses a construction method of a multivariable long-sequence time sequence prediction model based on a Transform framework, the prediction model, and provides a multivariable long-sequence time sequence prediction method based on the Transform framework. The technical problems that according to an existing multivariable long-sequence time sequence prediction method, a long-time input sequence cannot be efficiently processed, calculation complexity and space consumption are high, potential spatial correlation between variables is not fully utilized, and a stable periodic mode of the long-sequence time sequence is neglected are solved.

Description

technical field [0001] The present invention relates to the field of computer data processing technology and performance improvement technology, in particular to time series prediction technology, specifically a method for constructing a multivariate long-sequence time series prediction model based on Transformer framework, a prediction model and a prediction method. Background technique [0002] Multivariate time series forecasting has been widely used in scenarios such as transportation planning, energy consumption, financial management, weather forecasting, and disease spread analysis, constantly reshaping modern society. For example, forecast traffic flow to plan the best driving route, and forecast the stock market to design the best investment strategy. In these practical applications, an urgent need is to extend the prediction time to the distant future, which is of great significance for long-term planning and preventive warning. Performing accurate multivariate lon...

AI Technical Summary

Problems solved by technology

[0010] In view of this, the purpose of the first aspect of the present invention is to provide a method for building a multivariate long-sequence time series prediction model based on the Transformer framework; the purpose of the second aspect of the present invention is to provide a multivariate method based on the Transformer framework Long-sequence time-series prediction model; the purpose of the third aspect of the present invention is to provide a multi-variable long-sequence time-series prediction method based on the Transformer framework, to solve the existing multi-variable long-sequence time-series prediction method that cannot efficiently process long Technical issues such as temporal input sequences, high computational complexity and space consumption, underutilization of potential spatial correlations between variables, and ignorance of stable periodic patterns in long-sequence time series

Images