Analysis of Pre-Weighted and Post-Weighted Association Rule Mining

Association Rule Mining (ARM) is an important task of data mining that finds interesting patterns and relationships among items in the huge amount of data stored. While widely used in many different fields, the classical ARM assumes that all items have the same significance. Weighted Association Rule Mining (WARM) overcomes this problem by assigning weights to items with the goal of reflecting their importance to the mining results. In this case, the main question is whether weighting process should be applied before ARM or after it. To answer this question, this paper analyzes and compares two alternative approaches: Pre-Weighted Association Rule Mining (PreWARM) and Post-Weighted Association Rule Mining (PostWARM). It is shown by experimental studies that PostWARM produces more compact rules with higher information content and PreWARM finds more meaningful rules than standard rule generation methods.