Grandual Information and Industry Return Predictability from the Perspective of Big Data 大数据视角下的公司离散颗粒信息与行业收益率预测

This project uses high-frequency data to deeply explore the relationship between granular information sectoral movements—whether the returns of all individual stocks in the market can effectively predict industry returns. The study of industry return predictability is related to the optimal decision-making of investors, financial institutions, and policymakers. The complex economic network of interconnected firms highlights the relation between sectoral movements and heterogeneity shocks to individual stocks. The advancement of big data and machine learning technology allows one to study the micro-driving force of industry returns. To extract meaningful information, this project uses the LASSO (Least Absolute Shrinkage and Selection Operator) shrinkage method to reduce the dimensionality and integrates the economic structure into machine learning methods to improve predictability, stability, and economic interpretability. This project will serve as the basis for more realistic asset pricing theories and help identify and prevent financial risks. Specifically, the research objectives are as follows: 1) investigate the relation between granular information and sectoral movements; 2) study structural machine learning by incorporating economic links between firms; 3) apply the developed approach to the Chinese market to provide a new perspective for the development of digital economy in our country.