A time sequence prediction system fusing a time attention mechanism

Abstract

The invention provides a time sequence prediction system fusing a time attention mechanism. The time sequence prediction system comprises a preprocessing module and a prediction module. Wherein the preprocessing module processes original records and data to form a time sequence, and the prediction module uses a coding- decoding structure encodes the time sequence and predicts required informationand data contained in the time sequence; And the prediction module comprises an angle regularization module, and the angle regularization module adjusts the influence weight obtained by the coding information obtained by the data of each time point in the time sequence.

Description

technical field [0001] The present invention relates to a time series forecasting system, in particular to a time series forecasting system incorporating a temporal attention mechanism. Background technique [0002] Time series refers to a set of data arranged in time order, and any ordered real-valued sequence can be treated as a time series. Time series data widely exist in all aspects of social life, such as financial income, meteorological research, medical diagnosis, road traffic, power system, commodity demand, industrial production, network security, etc. With the continuous development of information technology, the amount of time series data is increasing day by day, and the data dimension is getting higher and higher. [0003] Time series forecasting refers to analyzing a given time series and predicting future changes. The analysis and forecasting of time series data is of great significance in many applications, and getting forecasts earlier and getting more ac...

AI Technical Summary

Problems solved by technology

[0009] The invention uses angle regularization to improve the problem of information loss due to the long tail effect brought by the temporal attention mechanism in the time series prediction problem

Images