K-Means Clustering Method Based on Node Similarity in Traditional Chinese Medicine Efficacy

A typical traditional Chinese medicine (TCM) prescription (or formula) is composed of one or more kinds of herbs, which can play a certain role according to the appropriate dosage and production method. The number of possible TCM prescription is nearly as large as that of chemical structures, so the development of quantitative prescription-efficacy relationship models is as appealing as to build a quantitative structure-efficacy relationship model. In this paper, we first construct a binary network of prescriptions and herbs based on the real TCM data. Through the analysis of the network, we propose a method to calculate the expected value of herbs based on the node similarity, which represents the probability that two kinds of herbs exist in the same prescription. Moreover, we use the hierarchical clusters to classify the herb pairs whose expected value is greater than a certain value, and observe which herbs belong to the same category. In addition, we predict the efficacy of herbs by both classic K-means algorithm and improved K-means algorithm which is based on the node similarity. After verification, the results show that the improved algorithm can predict the efficacy of traditional Chinese medicine more accurately comparing with the classic K-means algorithm. Therefore the model can be used for predicting the potential efficacy of traditional Chinese medicine.