Interpretable Medical Recommendations Based on SHAPs

In the past few years, chronic diseases of the elderly have gradually become the killer of the elderly, seriously affecting their physical health of the elderly. Doctors can extract relevant information from the data to make patient treatment recommendations. This study proposed a smart recommendation system for doctors by using deep learning traditional methods. Proposed method firstly use six traditional deep learning methods such as naive Bayes, logistic regression, decision tree etc for recommendation purpose. Next step to get the better understanding of the recommendation we apply an interpretable recommendation system based on SHapley Additive exPlanations. (SHAP). Last time is provide recommendations based on better and weak recommenations by graphical way. The proposed methods were applied to two chronic diseases of old age, including heart disease and diabetes. After the data is preprocessed, naive Bayes, decision tree, and other machine learning operations are carried out on the data. SHAP was used to calculate the importance of each feature for randomly selected patients. Finally, the contribution coefficient of the feature to the result is presented, which impacts the output. By analyzing the influence of different features of users on recommendation results, the proposed system explains why such results are recommended to users to improve users' trust in the recommendation results, which is of great significance for medical recommendation.