A Novel TCM Prescription Recommendation Method based on Attention Factorization Machines

Abstract

A prescription recommendation algorithm for attention factorization machines is proposed in this study. This algorithm leverages the pair-wise interaction of factorization machines to capture the multi-category attributes of patients and prescriptions. Additionally, an attention network is incorporated into the factorization machine to assign higher weights to the effective features within the prescription. This enables the algorithm to discern the importance of different combinations of features in the prescription, thereby enhancing the recommendation performance of the model. Through extensive experimentation, it is observed that the prescription recommendation model based on attention factorization machines does not rely on manual features and exhibits commendable recommendation performance. Furthermore, it achieves a certain degree of individualized recommendation effect.