Chi-squared computation for association rules: preliminary results
Sergio A. Álvarez · 2003
Chi squared analysis is useful in determining the sta-tistical significance level of association rules. We show that the chi squared statistic of a rule may be computed directly from the values of confidence, support, and lift (interest) of the rule in question. Our results facilitate pruning of rule sets obtained using standard association rule mining techniques, allow identification of statisti-cally significant rules that may have been overlooked by the mining algorithm, and provide an analytical descrip-tion of the relationship between confidence and support in terms of chi–squared and lift. 1.