Classification Rule Learning with APRIORI-C

Mining of association rules became one of the strongest fields of data mining. This paper presents a classification rule learning algorithm APRIORI-C, upgrading APRIORI to dealing with classification problems, decreasing its memory consumption and time complexity, further decreasing its time-complexity by feature subset selection, and improving the understandability of results by rule post-processing. This step also improved accuracy when dealing with unbalanced class distributions. The algorithm was applied to UCI domains as well as to the COIL challenge data.