Forecasting Stock Prices Changes Using Long-Short Term Memory Neural Network with Symbolic Genetic Algorithm

Abstract

This paper presents an enhanced Long-Short Term Memory Neural Network (LSTM) framework that combines Symbolic Genetic Algorithm (SGA) to predict cross-sectional price returns for 4500 listed stock in China from 2014 to 2022. Using the S&P Alpha Pool Dataset for China, the framework incorporates data augmentation and feature selection techniques. The study demonstrates significant improvements in Rank Information coefficient (Rank IC) and IC information ratio (ICIR) by 1128% and 5360% respectively when applied to fundamental indicators. For technical indicators, the hybrid model achieves a 206% increase in Rank IC and an impressive surge of 2752% in ICIR. Furthermore, a simple rule-based strategy based on the proposed hybrid SGA-LSTM model outperforms major Chinese stock indexes, generating average annualized excess returns of 31.00%, 24.48%, and 17.38% compared to the CSI 300 index, CSI 500 index, and the average portfolio, respectively. These findings highlight the effectiveness of LSTM with SGA in optimizing the accuracy of cross-sectional stock return predictions and provide valuable insights for fund managers, traders, and financial analysts.