Multilevel Association Mining with Particle Swarm Optimization: A Comprehensive Approach for High-Utility Itemset Discovery

In various domains, the application of data mining algorithms has become indispensable for uncovering meaningful insights from extensive datasets. These algorithms play pivotal roles in engineering, medicine, and business, facilitating informed decision-making and pattern recognition. High-utility pattern mining, an essential aspect of knowledge discovery, has emerged to address limitations in traditional frequent pattern mining by considering the relative importance of individual items. This paper introduces a multilevel association approach using particle swarm optimization (PSO) for mining frequent itemsets, accommodating both negative and positive thresholds. The approach encompasses comprehensive data preprocessing, association rule mining, and data pruning. Experimental results on diverse datasets, including T10I4D100K and Mushroom, showcase the effectiveness of the proposed methodology.