OGA-Apriori: An Optimized Genetic Algorithm Approach for Enhanced Frequent Itemset Mining

Abstract

Frequent Itemset Mining (FIM) can be broadly categorized into two approaches: exact and metaheuristic-based methods. Exact approaches, such as the classical Apriori algorithm, are highly effective for small to medium-sized datasets. However, these methods face significant temporal complexity when applied to large-scale datasets. However, while capable of addressing larger datasets, metaheuristic-based approaches often struggle with precision. To overcome these challenges, researchers have explored hybrid methods that integrate the recursive properties of the Apriori algorithm with various metaheuristic algorithms, including Genetic Algorithms (GA) and Particle Swarm Optimization (PSO). This integration has led to the development of two prominent techniques: GA-Apriori and PSO-Apriori. Empirical evaluations across diverse datasets have consistently shown that these hybrid techniques outperform the traditional Apriori algorithm in both runtime and solution quality. Building upon this foundation, this study introduces an enhanced version of the GA-Apriori algorithm, Optimized GA-Apriori (OGA-Apriori), to improve runtime efficiency and solution accuracy. Comprehensive evaluations on multiple datasets demonstrate that the proposed OGA-Apriori approach achieves superior performance compared to the original GA-Apriori in both runtime and solution effectiveness.