Analysis of metastereotypes on gender differences in empathy selection: a classification model based on machine learning

This study uses a classification model based on machine learning to explore the impact of metastereotypes on gender differences in empathy selection. The study uses a random forest model and optimizes it through feature engineering techniques such as mutual information method and principal component analysis, achieving excellent performance with an F1 score of 85.2% and an AUC of 0.91. Empathy scores and metastereotype scores were identified as key predictive features, and gender-specific SHAP value analysis showed that male decisions were more driven by empathy, while females had greater variability in the influence of metastereotypes. The gender difference coefficient (GDC) was 0.12, indicating that the gender effect was statistically significant but limited in magnitude. Visualization tools such as heat maps and box plots effectively demonstrated feature importance and gender-specific patterns, providing a solid framework for understanding the complex interaction between metastereotypes and gender differences in empathy selection.