Network Analysis for Core Herbal Combination Knowledge Discovery from Clinical Chinese Medical Formulae

Abstract

Chinese medical formula (CMF) is the main therapies in traditional Chinese medicine (TCM) clinical practices. Discovery of the CMF empirical knowledge of the famous TCM physicians in clincal data is significant. This paper proposes an effective data mining approach to analyze and uncover the clinical CMF empirical knowledge of famous TCM physicians. Complex network is a current hot research topic in complex systems field. We construct the combination network of CMFs by graph method and have a statistical analysis of the network based on large-size CMF data set. As we known, CMF has a number of constituent herbs, which are prescribed by TCM physicians. We consider any CMF with same element herb has correlation to each other. That is, we construct a connected sub-graph with the CMF as node and the edge weight is computed by an appropriate similarity method. With CMF dataset, we can build CMF network with dense connected edges. We implement an algorithm based on the 'hub' node features of the CMF relation network to extract the core CMFs in a large size CMFs. Thereafter, the maximum frequent itemset algorithm is applied to discover the core herbal combinations from the core CMFs. We have utilized the network-based method to discover several core CMFs from the outpatient data of the famous TCM physicians. The clinical CMFs for disharmony between the liver and spleen syndrome are analyzed. The preliminary results show that we propose an effective approach for core herbal combination knowledge discovery in large size CMF data sets.