TCMPR: TCM Prescription Recommendation Based on Subnetwork Term Mapping and Deep Learning

Figure 1

An overview of our methods. First, we constructed the HSKG and symptom network (a). Second, comprehensive embedding of patient symptoms was formed with the SSTM and symptom network (b). Finally, the patient’s comprehensive embedding vector was used for TCM prescription recommendation (c); the predicted probability of each herb is the output, so as to obtain the recommended prescription.