Mining Condensed and Lossless Association Rules by Pruning Redundancy

There are excessive and disordered rules generated by traditional approaches of association rule mining, many of which are redundant, so that they are difficult for users to understand and make use of. Han et al pointed out the bottleneck of association rules mining is not on whether we can derive the complete set of rules under certain constraints efficiently but on whether we can derive a compact but high quality set of rules. To solve this problem, a new method was represented, which is based on statistics and probability to get a condensed rules set by removing redundant rules. Our set of rules is more concise, more intelligible and easier to manage and use than others. Especially, the condensed set is lossless so that its switch to original rules-set can be realized. Itpsilas important because the switch keeps the information complete. Experiments on some datasets show that the number of rules in rules-set has been reduced greatly.