Empirical Analysis of Broad-Based Index Stock Price Forecasting Using LSTM Models: A Comparative Study with ARIMA and Ensemble Learning Models

DOI

10.2991/978-94-6239-701-9_107How to use a DOI?

Keywords

LSTM model; Broad-based index; Stock price forecasting; Financial time series data; Multi-model comparison

Abstract

Financial time series data are typically nonlinear and non-stationary, which restricts the prediction accuracy of traditional models and hardly meets practical financial decision-making needs. Long Short-Term Memory (LSTM) networks, with their special gating mechanism, provide a new technical way for accurate forecasting of stock market indices. This study uses daily trading data of the CSI 300, ChiNext, and CSI 500 indices from January 2015 to September 2025. We construct an explanatory variable system including raw price features and common technical indicators, and establish LSTM, ARIMA, Random Forest, and XGBoost models. Using regression and classification evaluation with multi-period rolling tests, we compare model performance in price direction judgment and closing price fitting. Empirical results show that the LSTM model performs best: direction prediction accuracy reaches 51.74%, AUC is 0.52, MAPE is 0.87%, and RMSE is 50.50 yuan, significantly outperforming the other models. Its gating mechanism and nonlinear structure better match the characteristics of stock price data. This paper supports the application of LSTM in broad-based index forecasting and provides a reference for financial time series model selection.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Xinran Jia
PY  - 2026
DA  - 2026/07/30
TI  - Empirical Analysis of Broad-Based Index Stock Price Forecasting Using LSTM Models: A Comparative Study with ARIMA and Ensemble Learning Models
BT  - Proceedings of the 2026 11th International Conference on Social Sciences and Economic Development (ICSSED 2026)
PB  - Atlantis Press
SP  - 1043
EP  - 1050
SN  - 2352-5428
UR  - https://doi.org/10.2991/978-94-6239-701-9_107
DO  - 10.2991/978-94-6239-701-9_107
ID  - Jia2026
ER  -