Machine learning in the Australian equity market

This study explores the application of advanced machine learning techniques to enhance understanding of Australian equity markets. We assess the predictive power of ten models, including traditional linear approaches and sophisticated machine learning methods such as random forests, gradient boosting, and neural networks with varying layers. Our findings demonstrate the superior performance of tree-based methods and neural networks in capturing the complex dynamics of stock returns in Australia, consistently outperforming linear models in both forecasting accuracy and economic gains. By analysing a diverse set of predictors collectively, including firm-specific characteristics and macroeconomic variables, we uncover that factors such as firm size, volatility, and trading frictions are crucial in influencing Australian stock returns. Contrary to expectations, these models often perform well across various market segments, including large, liquid stocks, challenging conventional assumptions about machine learning's efficacy in different market contexts.