This study aims to compare the effectiveness of different machine learning models in stock selection for the A-share market. Daily trading data of 50 A-share stocks from November to December 2023 were selected

and a feature system incorporating price momentum

intraday characteristics

and volume-price relationships was constructed. Three machine learning models—logistic regression

random forest

and XGBoost—were trained and tested. Empirical results demonstrate: (1) All models effectively predict cross-sectional relative stock performance

achieving test set accuracy rates of 75.71%-77.14%

significantly higher than random guessing; (2) The XGBoost model performed best with an accuracy rate of 77.14%; (3) Logistic regression model exhibited the highest recall rate (84.76%) in identifying outperforming stocks. This study provides empirical evidence for applying machine learning in quantitative stock selection for A-shares

validating the effectiveness of short-term technical factors in cross-sectional prediction.