Quantitative Association Rule Mining using Multi- objective Particle Swarm Optimization
Jayashree Piri, Raghunath Dey · 2014
Abstract--Association rule mining is a technique of discovering interesting correlation among items present in a dataset. To discover such interesting association rules, more than one objective need to be optimized rather than exploiting a single objective. This motivated to pose the association rule mining algorithm as a multi objective problem and use particle swarm optimization based multi objective metaheuristics to solve this problem as they tend to explore the global search space effectively in less time. This paper considers confidence, comprehensibility, interestingness as three objective for mining association rule and use a pareto based Particle swarm optimization to extract useful and interesting rules from quantitative database. The results of these algorithms are evaluated on various quality measures and are found to be suitable. Keywords- Association rule mining, Multi objective Particle swarm optimization, repairing operator, complete overlap, partial overlap, Join operator, quality measures. 1