An Interpretable Artificial Intelligence Model of Chinese Medicine Treatment Based on XGBoost Algorithm

Abstract

TCM treatment model is an effective tool to provide correct guidance and decision-making in clinical practice. Recently, traditional Chinese medicine (TCM) has become an increasingly concerned problem in the world, and it is still a hot research topic. However, most machine learning researches pursue the performance of the model, but ignore the trust mechanism of decision-making process. The interpretable TCM treatment model based on XGBoost integration is constructed in this paper, and the interpretability of the model is taken into account when the performance is good. AUC is selected as the main evaluation index of model performance, and other commonly used evaluation indexes are added in the comparative experiment: accuracy. The results show that the average performance of the proposed model is better than that of the traditional logistic regression algorithm. The interpretability of the model is considered in the selection of base classifier, feature selection and model integration. Finally, the whole model and the decision explanation to the specific samples are provided.