Exploring the International Application of Machine Learning in Asset Pricing: An Empirical Study

Abstract

This thesis delves into the application of machine learning models for predicting cross-sectional returns in diverse markets. Chapter One explores the predictive abilities of XG-Boost, Random Forest, and neural network models in relation to fund performance and fund manager information characteristics. The findings indicate that fund performance characteristics prove to be more informative of future fund performance than the characteristics of fund managers. Chapter Two probes the presence of bimodality in momentum stocks and examines the profitability of deep momentum, a machine learning return prediction model, in the UK, Japan, and South Korea. The findings demonstrate that bimodality is a phenomenon linked to developed markets and can cause losses for JT strategy investors. However, the deep momentum model generates substantial profits in all markets by relieving bimodality in long-short portfolios. Chapter Three investigates the efficacy of the momentum factor in Chinese stock markets. We compare the performance of the traditional linear JT model, the XG-Boost model, the neural network model, and neural network reclassification models as developed by Han (2022). The study finds that machine learning models based on the momentum factor outperform the traditional JT linear regression model, indicating a non-linear relationship between the momentum factor and stock returns in China. Han's reclassification models perform the most strongly after reclassification of the true target distribution within high-return deciles moves from a bimodal shape to a right-skewed distribution. The study also observes a significant positive correlation between the return of the long-only portfolio developed using the momentum factor in the machine learning framework and the size and sentiment index. Overall, this thesis attests to the practicality of machine learning models for predicting cross-sectional returns in various markets, with potentially gainful implications for investors and policymakers.