Mining Distinct Representations of High-Utility Itemsets Using Particle Swarm Optimization

Abstract

Closed high-utility itemset mining (CHUIM) and top-k high-utility itemset mining (top-k HUIM) are techniques extensively applied in data analysis to discover patterns or relationships with potentially valuable information. Although these processes are helpful in various decision-making problems, it is challenging to retrieve the solutions in a short amount of time. One way of alleviating this is to find approximate results through heuristic models. However, few heuristics are currently available, and the previous works also suffer performance limitations that impede their overall usability. This thesis addresses these issues by introducing two distinct algorithms described in respective research papers---CHUI-PSO for CHUIM and TKU-PSO for top-k HUIM. Both algorithms are heuristic techniques based on particle swarm optimization and incorporate several unique strategies aimed at reducing the computational complexity associated with earlier methods. To evaluate the contribution of the models, we compare performance characteristics against the current state-of-the-art approaches. Experimental results indicate that the proposed algorithms outperform previous works in terms of precision and efficiency.