Recent Advances in Stock Market Prediction: Applications of Machine Learning and Deep Learning

This book provides a comprehensive assessment of forecasting models used in the stock market across both developed and emerging markets, utilising data from the UK, US, China, and India. The first section compares Particle Swarm Optimised Radial Basis Function Neural Networks (PSO-RBFNN) with standard RBFNN and two benchmark econometric models, ARIMA and Holt–Winters. The findings indicate that econometric models tend to perform better in developed markets, whereas neural networks show more evident advantages in emerging markets. PSO-RBFNN outperforms traditional RBFNN due to its improved parameter optimisation. The second section expands the analysis by examining Random Forest (RF), Support Vector Regression (SVR), and Long Short-Term Memory (LSTM) models, along with their ensemble models. SVR performs well across most datasets, while some ensemble models show mixed but notable improvements depending on market conditions. Stacking LASSO reduces extreme prediction deviations in the UK and US, whereas in China and India, it also demonstrates solid performance. Other ensemble models, such as simple average, weighted average, and short moving averages, sometimes perform better on certain error metrics. Overall, the findings highlight how market structure influences the strengths of machine learning and econometric forecasting techniques, offering valuable insights for researchers, practitioners, and policymakers interested in financial prediction.