Efficiently Mining Positive Correlation Rules

One of the main tasks of knowledge discovery in Traditional Chinese Medicine is discovering novel paired or grouped drugs from the Chinese Medical Formula database. Novel paired or grouped drugs, which are special combinations of two or more kinds of drugs, have strong efficacy in clinical care. We use positive correlation rule mining to analyze the large number of complex correlation relationships among various kinds of drugs in order effectively to find novel paired or grouped drugs in the Chinese Medical Formula database. We firstly give some new related definitions and then develop an efficient algorithm for discovering all positive correlation rules based on frequency patterns from a large database. Experimental results on the Chinese Medical Formula database and the mushroom database show that all techniques developed in this paper are feasible.