Several Data Mining Methods Research In Traditional Chinese Medicine Prescription Knowledge Discovery

TCM(Traditional Chinese Medicine) is an import part of traditional Chinese culture and the crystallization of human wisdom, and it plays an import role in human history especially ancient people struggled with the disease. Chinese medicine formulae is an important subject of TCM, and it’s compatibility regularity has an important significance. After thousands of years of development, TCM accumulate abundant materials and a lot of classic books, in the face of such vast amounts of data, ordinary manual processing method has been difficult to meet the needs of the people on the research of Chinese medicine theory. Therefore, people research on TCM with information technology is urgent.This paper mainly through data mining techniques to study traditional Chinese medicine prescription, the main thesis work are as follows:1) A drugs contribution-based algorithm to discover the core drugs in the prescription is proposed, this algorithm can get the low frequency herb which frequency method ignore.2) This paper gives the definition of drugs distance and combines the distance and normal point mutual information to get an evaluation of the correlation degree of drug indicators DPMI, based on DPMI, a drug combination mining algorithms which is similar to Apriori is proposed. This algorithm not only drugs the common herb pair in traditional medicine, but also has the mulberry bark, Tuckahoe and Cortex Lycii combination which is effective for disease.3) This paper first introduces the concept of interest degree, then combines support and interest, then an FP-Growth algorithm based on interest degree to analysis Chinese medicine formulate is proposed. This algorithm has achieved good effect not only on reducing the scale of frequent item sets, but also reducing the algorithm time.4) This paper uses the LDA model on the prescription text mining, and gets the drug clustering based on the drugs distribution under the LDA theme model, then based on the theme distribution of the prescription, hierarchical clustering on formulate is used. By the experimental results, the topic-based LDA model on the consumptive lung prescription drugs and prescription clustering has very good results.