Chinese A-Share Stock Selection Results Across AI Models

Summary

This weekly report compares artificial intelligence stock-selection portfolios across broad A-share universes, industry-neutral portfolios, and portfolios restricted to the CSI 300 or CSI 500. It names XGBoost, support vector machines, random forests, logistic regression, stacking, and naive Bayes as the best-performing models in different universes and periods. The comparisons cover the reported week, recent three-month period, and recent year, using absolute and benchmark-relative returns where provided.

The results vary with the selection universe and evaluation window: no single model leads across all cases. For example, the report identifies XGBoost as the weekly leader for broad non-neutral selection and CSI 500 constituents, while other models lead in other settings or periods. Some reported excess-return figures are absent or negative, so the excerpt does not provide a complete, consistent performance table or enough detail to independently compare the strategies. It gives no model inputs, portfolio construction rules, transaction-cost assumptions, or testing methodology. The source cautions that strategies based on historical experience may stop working.

Key ideas

• The report compares several machine-learning models across broad, industry-neutral, and index-constituent A-share portfolios.
• The leading model changes across stock universes and measurement periods.
• Reported evaluation windows include the week, recent three months, and recent year.
• The excerpt omits some excess-return values and does not explain model inputs or portfolio construction.
• The report warns that historical model performance may not persist.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.