Discovering Knowledge From Traditional Chinese Medicine Data With Complex Network Model Based On Mapreduce

Data mining technology is an important application of traditional Chinese medicine (TCM) for knowledge discovery. Commonly used data mining model is based on the transaction item. It considers Chinese medical formula (CMF), comprised of a few herbs, as a transaction stored in the transaction database. The transaction model can be really used to find much compatibility knowledge of TCM, but it cannot explicitly analysis relationships between herbs and against deeply mining.This paper, on the view of complex network model, tries to build a TCM network with CMF dataset. With this network, on which applying many complex network analysis algorithms, we can clearly find relationships between herbs and deeply discover the compatibility knowledge of TCM.The main content of the thesis includes:1) Propose a measure of similarity between herbs and construct the network with its help. Then we explore the network in much aspect and find out that there are some characteristics in this kind of network. Most of them comply with the characteristics of complex network, for example, the node degree accords with power-low distribution.2) For digging out core herbs in network and computing the dependence of specify herb, this paper implements PageRank algorithm and uses measure of vertex dependency which defined in Betweenness Centrality Computing algorithm.3) Apply improved Label Propagation Algorithm for discovering community structure in TCM network, which can make similar herbs gathering together to form lots of herb communities. Herbs in the same community always comprise all kinds of formula together frequently.4) Finally, focusing on building network and mining quickly in large dataset, we rewrite the algorithms based on MapReduce model, and all of them have been tested in Hadoop platform.