Improved association rule mining method based on T statistical
基于T统计量的一种改进关联规则挖掘方法*

One major purpose of data mining is to discover interesting association rules. And traditional data mining methods are generally based on support-confidence system, which might frequently dig out false rules or neglect useful ones. At terms of this problem, referring to the mind of control experiment, we develop T statistical-based association rule mining methodology, in which we use significance to replace confidence in order to obtain results with more statistical significance. Accounting case analysis and data mining results show that this new method can efficiently solve problems underlying in traditional association rules methods, and thus improve the validity of the rules.