Cross domain changeable feature fusion depth modeling method and system

Abstract

The invention provides a cross-domain variable feature fusion depth modeling method and system. The method comprises the following steps: acquiring a multi-dimensional time sequence in a past preset time period; determining features of different dimensions of the multi-dimensional time sequence; fusing the features, and generating high-dimensional features according to a fusion result; reducing the dimensionality of the high-dimensional features, and determining a multi-dimensional time sequence in a future preset time period according to a dimensionality reduction result; the method has the beneficial effects that the features of the multi-dimensional time sequence in the past preset time period are determined, the obtained features are fused, the high-dimensional features are generated according to the fusion result, and then the dimension reduction processing is performed on the high-dimensional features, so that the accuracy of the multi-dimensional time sequence is improved. According to the dimension reduction processing result, the multi-dimensional time sequence in the future preset time period is determined, so that the future multi-dimensional time sequence can be predicted according to the past multi-dimensional time sequence, and the influence on the future monitoring data can be known according to the predicted future multi-dimensional time sequence.

Description

technical field [0001] The invention relates to the technical field of multi-dimensional time series prediction, in particular to a deep modeling method and system for cross-domain multi-variable feature fusion. Background technique [0002] Based on the existing monitoring system, it can be found that the vast majority of monitoring data are multi-dimensional time series. The monitoring of multi-dimensional time series data plays a role that cannot be ignored in the process of company fault discovery. However, from the massive multi-dimensional time series data indicators It can be found that there are many kinds of indicators, and the relationship is complex; in terms of the characteristics of indicators themselves, there are periodicity, regular spikes, overall rise and fall, low peak periods, etc. In terms of influencing factors, there are holidays, temporary activities, weather, The epidemic situation and other factors; the fixed threshold monitoring strategy of the ori...

AI Technical Summary

Problems solved by technology

[0004] In view of the above-mentioned shortcomings of the existing technology, the purpose of the present invention is to provide a cross-domain variable feature fusion depth modeling method and system, which is used to solve the problem of the inability to predict the future value of multi-dimensional time series in the prior art The problem

Images