Abstract: Taking halide double perovskite materials as the research object,the machine learning method is used to predict the band gap and relative stability of halide double perovskite materials with high speed and high accuracy. Four distinct algorithms,namely Bayesian ridge regression,gradient boosting regression,support vector regression,and XGBoost,are employed to construct predictive models. The results show that gradient boosting regression can provide the highest performance prediction for relative stability（R~2=0.9161,MAE=0.2061）,XGBoost can provide the highest performance prediction for band gap（R~2=0.9899,MAE=0.0542）,and after using the SHAP method to explain the model,the new samples after element substitution are screened,and finally 18 halide double perovskites with ideal light absorption range and exceptional stability are obtained. These outcomes indicate that compared with traditional methods,data-driven machine learning can effectively accelerate functiona material discovery and improve design efficiency.