Efficiently mining special patterns

Zhongmei Zhou, Xuejun Wang, Guiying Pan

Abstract: One of the main tasks of KDTCM (knowledge discovery in Traditional Chinese Medicine) is discovering paired or grouped drugs in Chinese Medical Formula database. Paired or grouped drugs are special combinations of two or more kinds of drugs. They not only have strong efficacy, but also can be used to construct formula classification rules. Associated and correlated pattern mining is effective in a certain extent because of the large number of correlation and association relationships among various kinds of drugs. However, both independent associated patterns and mutually positively correlated simultaneous associated patterns have the downward closure property. They might be more special and have probability greater than associated and correlated patterns to be paired or grouped drugs. Therefore, in this paper, we develop an algorithm for generating all such kinds of patterns synchronously. Experimental results show that the proposed techniques in the paper are effective and feasible.