Prediction of one-year readmission risk in patients with unstable angina based on CatBoost

In order to solve the shortcomings of the readmission risk prediction and assessment model based on Chinese and Western medicine indicators for patients with unstable angina (UA) in clinical application, a risk prediction model incorporating tradition Chinese medicine (TCM) indicators for the readmission risk of UA patients within one year was constructed by introducing the CatBoost algorithm. First, data preprocessing, imbalance processing, and feature selection were performed on the dataset of UA patients collected from a tertiary hospital in Tianjin, China, to screen out the features that have an important impact on the model performance; next, the CatBoost model was used to train and test the dataset; finally the experimental results show that the CatBoost model exhibits superior performance compared with five mainstream machine learning; meanwhile, by interpreting the features extracted from the CatBoost model, it is found that adding TCM indicators in the risk prediction model has some practical value.