Post mining pruning methods

Abstract

Mining association rules has been an important topic in data mining research in recent years from the standpoint of supporting human-centered discovery of knowledge. The present day model of mining association rules suffers from the following shortcomings: (i) lack of user exploration and control, (ii) lack of focus, (iii) huge number of rules technically unreadable. All data mining researchers, have given a high importance on developing fast algorithms for rules discovery, and have applied different types of constraints in different algorithms to prune item sets, that do not occur frequently in the database, while generating association rules in order to speed up the mining process, depriving the user from guiding and tuning this process which in fact has been functioning like a black-box; hence, resulting in an exponential explosion of data and bypassing the user's area of interest. In this thesis we introduce the idea of allowing the user to concatenate multiple pruning methods and to apply them iteratively on the list of discovered association rules to reduce its size by removing redundant and uninteresting rules so that the analysis of these rules becomes easier, and more efficient.