Cumulative Summary List Driven Lightweight Frequent Closed High Utility Itemset Mining

Pattern mining has always been a dominant approach in data mining to identify the object(s) of interest from a humongous large data space to make real-time decisions. In sync with diverse business intelligence or other enterprise demands, pattern mining methods have gained wide-spread attention to serve purposes like high-utility itemset identification, segmentation and top-(itemsets) recommendation. In this paper, a robust cumulative summary list-structure driven lightweight frequent closed high-utility itemsets mining model (CSLFC-HUIM) is proposed that uses an improved cumulative summary list structure (CULS) to ensure optimal set of candidate itemsets in search space. Unlike classical support value driven strict-thresholding, CSLFCHUIM applied lift-score as minimum threshold called cumulative support/utility threshold to prune the search space. The strategic amalgamation of cumulative support value, utility threshold with CULS and cumulative summary list structure (CSLS) concept guarantee optimal set of itemsets in nested-list structure. It reduces unwanted database-scanning, execution-time and memory exhaustion, which makes it a potential solution for different pattern mining and business intelligence tasks.