Advancing Stock Market Return Forecasting with LSTM Models and Financial Indicators

Stock market return forecasting is critical for portfolio management and market efficiency. This study evaluates the predictive ability of the LSTM model using S&P500 index data from 1871 to 2019. The predictors include 12 financial indicators, such as dividend-price ratio, earnings-price ratio, and volatility, among others. The study results show that the LSTM model exhibits a consistent decrease in training and validation losses, indicating its ability to learn from the data. However, the low training set and negative test set show significant overfitting and fail to generalize to unseen data. Therefore, the LSTM model has potential applications and needs to be further improved to make it more valuable and reliable. Despite some limitations of LSTM models for stock market prediction, this study shows the promise of deep learning for financial prediction. Future research should focus on extending features, optimizing models, and integrating ensemble methods to improve the accuracy and robustness of predictions. Insights are provided to improve the predictability of stock market returns in the future using advanced modelling techniques.