Research on Complex Time Series Prediction Algorithm Based on ENA-SARIMA-LSTM Combined Model

View at publisher Summarise with VIRAT
Authors: Zeyu Xi, Zihao Yu, Zirui Zhang
Cited by: 0
References: 10

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

In order to solve the problem that the traditional single model has insufficient prediction accuracy in dealing with complex time series with seasonal nonlinear features, this paper proposes a combined prediction algorithm that integrates elastic network regression and long short-term memory network algorithm. ENR elastic network regression is used to screen key characteristic variables, reduce data dimensions, and remove redundant information. Secondly, SARIMA is used to capture the seasonal and linear trends of the time series, the nonlinear dynamic features are mined with the help of LSTM, and the output of the two models is coupled through linear regression to form an ensemble prediction result. Experimental results show that the root mean square error of the proposed combined model is lower than that of LSTM, and the coefficient of determination R2is increased from 0.35 to 0.876, which significantly improves the prediction accuracy and stability. The results show that the combined algorithm can effectively integrate the advantages of linear and nonlinear modeling, and provides an efficient solution for complex time series forecasting.

Fields of Study & Subjects

Forecasting Techniques and Applications Time Series Analysis and Forecasting Stock Market Forecasting Methods

Publication details

Journal / source
IEEE International Conference on Power, Electronics and Green Energy (ICPEGE)
Pages
922–932
Publication date
28 January 2026
Language
EN