Evolutionary Multi-Objective Optimization Framework for Mining Association Rules
This work addresses the challenge of automating association rule mining for data analysis, particularly in domains like banking, but it is incremental as it adapts existing evolutionary algorithms to this task.
The paper tackles the problem of mining association rules from transactional datasets without requiring predefined minimum support and confidence thresholds, proposing two multi-objective optimization frameworks (NSGA-III-ARM and MOEAD-ARM) that achieve diverse, non-redundant rules, with NSGA-III-ARM performing better across most of the seven tested datasets.
In this paper, two multi-objective optimization frameworks in two variants (i.e., NSGA-III-ARM-V1, NSGA-III-ARM-V2; and MOEAD-ARM-V1, MOEAD-ARM-V2) are proposed to find association rules from transactional datasets. The first framework uses Non-dominated sorting genetic algorithm III (NSGA-III) and the second uses Decomposition based multi-objective evolutionary algorithm (MOEA/D) to find the association rules which are diverse, non-redundant and non-dominated (having high objective function values). In both these frameworks, there is no need to specify minimum support and minimum confidence. In the first variant, support, confidence, and lift are considered as objective functions while in second, confidence, lift, and interestingness are considered as objective functions. These frameworks are tested on seven different kinds of datasets including two real-life bank datasets. Our study suggests that NSGA-III-ARM framework works better than MOEAD-ARM framework in both the variants across majority of the datasets.