An improving mining algorithm aiming at a kind of specific function of degree of interest

Association rule mining is an important research area in data mining, which has a broad application background. There exist two short-comings in the classical association rule mining problem, namely every itemset is treated equivalently and use a uniform minimum support and minimum conference as weighting standard. Through introducing the function of the degree of interest on itemsets, φ, a direct generation of the problem of association rule mining, called φ-association rule mining, was introduced in Li (2001). It can solve the two short-comings at the same time. Based on an FP-tree, a universal algorithm for mining a T-frequent closed itemset was proposed in Li et al. (2001). Because this algorithm is based on general definition of φ, it will not perform well for all φ. In this paper, aiming at a kind of specific φ, namely, the degree of interest of every item is given, we present an improved algorithm. The experimental and performance studies show that our algorithm is more efficient than previous algorithm.