Long-time series data processing method based on hierarchical deep network

Abstract

An embodiment of the present invention provides a long-time series data processing method based on a hierarchical deep network, including: S1, inputting the long-time series data into a trained hierarchical deep network model; S2, using the convolutional layer The multi-pairs of convolution kernels and pooling kernels perform single-dimensional convolution operations on the long-term time-series data to obtain the first compressed time-series data set, and connect all compressed time-series data in the first compressed time-series data set to form The first multi-dimensional time-series data; S3, using the convolution layer to acquire multiple multi-dimensional time-series data, and combining the multi-dimensional time-series data whose length is less than a first preset length among the multiple multi-dimensional time-series data to form a first data set ; S4, using the recursive layer and the fully connected layer to process the first data set to obtain a classification result vector of the long-term series data. The automatic processing of long data by the hierarchical deep network model is realized, and the results are accurate.

Description

technical field [0001] Embodiments of the present invention relate to the technical field of computer data analysis, and more specifically, relate to a long-term sequence data processing method based on a hierarchical deep network. Background technique [0002] In most industrial productions, in order to facilitate the management personnel to observe the working status of the production equipment in real time, there will be sensors attached to the equipment to generate real-time working condition data. Experienced experts in the field can judge the corresponding production by observing its changes over a period of time Whether the working conditions of the equipment have changed. This type of problem is really a classification problem. For example, in electric power production enterprises, experts can judge the abnormality of the motor by observing the voltage change of the motor, and make further processing in time. However, as the design of production equipment tends to ...

AI Technical Summary

Problems solved by technology

Although deep learning technology has made great achievements in various fields, there is still no good solution for the classification of time series data. The reason is that there is still no good method to extract the characteristics of time series data.

Images