Multi-objective particle swarm optimization algorithm using adaptive archive grid for numerical association rule mining

The most challenging issues in association rule mining are dealing with numerical attributes and accommodating several criteria to discover optimal rules without any preprocessing activities or predefined parameter values. In order to deal with these problems, this paper proposes a multi-objective particle swarm optimization using an adaptive archive grid based on Pareto optimal strategy for numerical association rule mining. The proposed method aims to optimize confidence, comprehensibility and interestingness for rule discovery. By implementing this method, the numerical association rule does not require any major preprocessing activities such as discretization. Moreover, minimum support and confidence are not prerequisites. The proposed method is evaluated using three benchmark datasets containing numerical attributes. Furthermore, it is applied to a real case dataset taken from a weight loss application in order to discover association rules in terms of the behavior of customer page usage.