A predictive analytics approach for forecasting global stock index returns using deep learning techniques

Accurately predicting stock index returns remains a critical yet complex task due to the inherent volatility of financial markets and the intricate temporal dependencies within financial time series. This study presents a robust machine learning framework to forecast the relative returns of major global stock indices, including the Standard & Poor’s 500 Index (S&P 500), Financial Times Stock Exchange 100 Index (FTSE 100), Nikkei 225, Deutscher Aktienindex 30 (DAX 30), and Cotation Assistée en Continu 40 (CAC 40). The framework employs advanced deep learning models, including Long Short-Term Memory (LSTM), Dual-Layer Long Short-Term Memory (DL-LSTM), and Transformer architectures with Multi-Head Self-Attention, trained on both technical and fundamental indicators. The technical indicators include Exponential Moving Average, Relative Strength Index, Moving Average Convergence Divergence, and Bollinger Bands. In contrast, the fundamental indicators comprise Earnings Per Share, Price-to-Earnings ratio, Net Profit Margin, Return on Assets, and Dividend Yield. Experimental results show that the Dual-Layer Long Short-Term Memory model consistently outperforms baseline methods, particularly on the Standard & Poor’s 500 Index, achieving up to 78 percent accuracy, 81 percent precision, 76 percent recall, an F1 score of 80 percent, and a Brier Score of 0.46. These findings underscore the potential of combining advanced machine learning techniques with a comprehensive set of market indicators to enhance financial forecasting and support data-driven investment decision-making. • Forecast global stock index returns using deep learning and structured market indicators. • Combine technical and fundamental signals to support informed investment strategies. • Validate forecasting accuracy with macroeconomic benchmarks and performance metrics. • Apply sliding window methods and grid search to optimize time-series model training. • Demonstrate improved predictive power in key global indices using deep learning tools.