Study Of Data Mining Technology For Traditional Chinese Medicine Compatibility

Traditional Chinese Medical Science(TCMS) Mining is the application of data mining in TCMS field, and it is the crucial informational technique to discover hidden TCMS knowledge. Traditional pure-informatics mining way includes association rules, classification and clustering, complex network analysis etc, which become an important part of TCMS studies. However, recent researchers take more and more TCMS-combined model such as molecular pharmacy network analysis and latent zheng-structure model in this multi-discipline field.For those reason, this paper advances a correlation metric about zheng in high-dimensional background and a generalized couplet drugs formulation structure model on which mining zheng-yao relations and discover those abnormal relations. The main work is as following:1) Considering zheng’s complex meaning and its high-dimensional feature, we take the Latent Factor Model to measure zheng’s similarities and get those high dimensional vectors. To the problem how to set the value of important K, we advance a way of valuate the whole vectors by choosing some special vectors and evaluating.2) Learning from the relations of Fangji, we define a generalized couplet drugs formulation structure(FormulaForest). First, we realized a simple formulation algorithm and compare it with the association methods. Second, combining zheng’s similarities and the simple structure we advance a collaborative optimization strategy which can optimize the structure and zheng-yao relations at the same time.3) Apply the high-dimensional outlier detection genetic algorithm to the discovering of abnormal zheng-yao relations, and advance a accelerating strategy based on FP-Tree to accelerate the enumerate part of previous algorithm.4) Combining some theoretic analysis, we prove our algorithm and result correct.