Enhanced Reconfigurable Weighted Association Rule Mining for Frequent Patterns of Web Logs

Abstract

Systolic tree structure is a reconfigurable architecture in Field-programmable gate arrays (FPGA) which provide performance advantages. It is used for frequent pattern mining operations. High throughput and cost effective performance are the highlights of the systolic tree based reconfigurable architecture. Frequent pattern mining algorithms are used to find frequently occurring item sets in databases. However, space and computational time requirements are very high in frequent pattern mining algorithms. In the proposed system, systolic tree based hardware mechanism is employed with Weighted Association Rule Mining (WARM) for frequent item set extraction process of the Web access logs. Weighted rule mining is to mine the items which are assigned with weights based on user’s interest and the importance of the items. In the proposed system, weights are assigned automatically to Web pages that are visited by the users. Hence, systolic tree based rule mining scheme is enhanced for WARM process, which fetches the frequently accessed Web pages with weight values. The dynamic Web page weight assignment scheme uses the page request count and span time values. The proposed system improves the weight estimation process with span time, request count and access sequence details. The user interest based page weight is used to extract the frequent item sets. The proposed system will also improve the mining efficiency on sparse patterns. The goal is to drive the mining focus to those significant relationships involving items with significant weights.