Using XGBoost to Rank Chinese Stocks by Expected Return Class

Summary

This research summary describes applying XGBoost, a gradient-boosted tree method, to Chinese stock selection and index enhancement. The model uses CART trees and classifies each stock into one of three groups formed by sorting market returns. This turns stock selection into a classification task, with factor data as predictors and the return group as the target. Preprocessing includes outlier treatment, missing-value imputation, industry and market-cap neutralization, and standardization; the study separates training and test data.

Evaluation combines classification measures such as precision, recall, and F1 score with portfolio returns and risk measures. The summary reports results for a 2011–2019 sample, including a long portfolio’s annual return and excess return over an equal-weight benchmark, as well as index-enhancement figures for the CSI 500 and CSI 300. These are reported findings, not guarantees. The available text does not provide full model specifications, transaction-cost assumptions, or enough detail to independently assess robustness and out-of-sample performance.

Key ideas

• The study predicts which of three market-wide return groups each stock will enter.
• XGBoost combines CART trees through boosting to model relationships between factors and return classes.
• Factor preparation includes outlier handling, imputation, neutralization, and standardization.
• The evaluation pairs classification scores with portfolio and benchmark-relative performance measures.
• Reported historical returns do not establish future performance, and the summary omits details needed for independent validation.

Tags

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