Interestingness measures for association rules based on statistical validity

摘要: Assessing rules with interestingness measures is the pillar of successful application association discovery. However, discovered are normally large in number, some which not considered as interesting or significant for at hand. In this paper, we present a systematic approach to ascertain rules, and provide precise statistical supporting framework. The proposed strategy combines data mining measurement techniques, including redundancy analysis, sampling multivariate discard non- rules. Moreover, consider real world datasets characterized by uniform non-uniform data/items distribution mixture levels throughout data/items. unified framework applied on these demonstrate its effectiveness discarding many redundant non-significant while still preserving high accuracy rule set whole.