Research And Application Of Time Series Analysis

Time series data is ubiquitous in many fields such as transportation,finance,logistics,astronomy,etc.It contains a lot of valuable information.Therefore,the mining of time series has attracted the attention of many researchers.Time series mining mainly includes the analysis of association rules.The linear representation of the sequence,comparison of similarity,pattern extraction,abnormality detection,and time series prediction are of great significance for the study of time series.Time series data usually has a large amount of data and is analyzed on the original time series data.The calculation is complicated,and the storage cost is high.The execution efficiency of the algorithm is very low.Sometimes an approximate compressed sequence is used instead of the original time series.Improve the execution efficiency of the algorithm,so it is very important to study the linear representation of the time series.This article studies the time series from two perspectives:the linear representation of time series and the prediction of time series.In order to achieve time-series compression,this paper proposes a PLR_TSIP algorithm based on the segmentation of important points in time series.This method not only considers the overall fitting error of the segmented time series,but also takes into account the segmentation time.The length of the sequence.Through analysis and comparison of multiple sets of experimental data with other segmentation algorithms,this algorithm not only improves the accuracy of the fitting but also improves the efficiency of the time series compression.This paper uses the stock data for time series forecast research.In order to better study the stock index prediction problem,this paper proposes a stock index prediction method based on PCA and LSTM model,which includes the application of principal component analysis for dimension reduction and the use of LSTM network for training and forecasting..In the experiment,the LSTM model was used to predict the Nasdaq stock index data and the S&P 500 index data.Through adjusting the parameters of the LSTM network,a relatively good experimental result was achieved.The experimental results show that this method not only improves the training Efficiency also improves the accuracy of the forecast.