Research On And Application Of Multivariate Time Series Classification And Clustering Algorithms Based On Deep Learning

Multivariate time series is a kind of popular data from such as biomedicine,weather forecasting,network intrusion detection and moving object simulation.The classification and clustering of multivariate time series are both important topics of time series data mining and have important research value.Because multivariate time series data is usually highdimensional data.It is data arranged in chronological order,and the sequence length of multivariate time series data is not necessarily equal.Multivariate time series data's unique complexity makes classification and clustering of multivariate time series data very difficult.In recent years,due to the rise of deep learning,more and more traditional machine learning problems can be solved by deep learning methods and have achieved better performance.This article explores the study of multivariate time series classification and clustering algorithms based on deep learning.At present,some papers have done research on related aspects,but the research of multivariate time series classification and clustering based on deep learning is not much.Some problems still need to be solved.The research and application of multivariate time series classification and clustering based on deep learning in this paper have research value and practical significance.In order to improve the existing algorithms and to solve the problems of multivariate time series classification and clustering,this paper has conducted in-depth research on multivariate time series classification and clustering algorithms based on deep learning.The main research work in this paper is as follows:1)In this paper,neural network models for multivariate time series classification and clustering are designed,which can successfully classify and cluster multivariate time series data;2)Uses L2 Norm regularization for network model parameters in the classification neural network.Norm regularization reduces the problem of overfitting in neural network model;3)The use of deep separable convolutions in neural network model for multivariate time series classification can reduce the number of model parameters and speed up the speed of model training,Making the model more lightweight;4)The use of unsupervised autoencoder networks for multivariate time series clustering,improving the performance of multivariate time series clustering;5)The LSTM network in the autoencoder network structure is improved,and the Attention LSTM is used instead of the LSTM,so that the model has a more powerful memory and improves the performance of the model.6)The experimental results show that the network model of the multivariate time series classification in this paper is more accurate than the current algorithms on most data sets,and the training parameters in the network model are greatly reduced.The network model of multivariate time series clustering in this paper has also greatly improved the accuracy of clustering,which is higher than the accuracy of current algorithms.The future direction of the author is to further improve the network model to improve the performance of time and accuracy,and strive to apply the proposed network model on the key areas of multivariate time series data analysis.