Association Rules Mining Based on Statistical Correlation

Association rules mining is an important data mining task, this paper emphatically analyzes realization skill and defects of existing algorithms. On the basis, a novel measure, named statistical correlation, which can indicate the correlation degree of multi-items in a rule, is put forward to cut association rules with independent or negative correlation, and its concept, calculating formulas and primary characteristics are defined. In order to conveniently use statistical correlation to cut redundancy rules, a fast association rules mining algorithm, called F-Fminer, is designed with two new data structures UFP-Tree and FP-Forest which use multi-trees structure to store data. F-Fminer adopts divide and conquer strategy to mine frequent itemsets for every UFP-Tree basing on depth-first searching. It can be seen from experimentation that the method in this paper has greatly enhanced mining efficiency and reduced a lot of redundant rules than other classical algorithms.