Recommendation system for constructing T2DM patient drug regimen based on deep learning and reinforcement learning

The invention discloses a recommendation system for constructing a T2DM patient drug regimen based on deep learning and reinforcement learning, and relates to the technical field of digital medical treatment. A reinforcement learning method is adopted, a drug recommendation scheme is obtained according to basic information of a patient and an examination index set before drug use, then examination indexes after drug use are obtained through an index prediction module, and scores of the examination indexes after drug use serve as values of the drug recommendation scheme; and carrying out multiple recommendations until the number of the obtained values reaches a preset threshold value, and calculating to obtain a total value according to all the values. The total value reflects the long-term value of the drug recommendation scheme. The method not only considers the accuracy of the drug recommendation scheme, but also comprehensively considers the side effect and drug tolerance of the drug and the influence of t