Multistrategy Optimized Marine Predator Algorithm for Customized Association Rule Mining

Note that current association rule mining algorithms are lacking in customized mining interested rules and uninterested rules for users. This paper proposed a novel heuristic optimization algorithm named Differential Variability (DV) and Oscillatory Functions (OF) based Marine Predators Algorithm (DVOFMPA) to solve the customized association rule mining (CARM) problem. First, differential variability is utilized in the high-velocity ratio phase to improve the quality of the prey. Then, the oscillatory function is applied in the unit velocity ratio phase to enhance the sensitivity of the prey. Finally, a customized value function based on the user’s interested items, uninterested items, and the values of optimization weight factors is designed to evaluate the filter rules associated with the user-specified directions.The effectiveness of DVOFMPA is tested on 7 public datasets and 2 augmented datasets with different sizes and sparseness. Compared to the baseline algorithms, the number of interested rules improves at least 7:2% and at the same time the number of uninterested rules reduces 12:7%, which indicates the advantages of the proposed DVOFMPA in the customized association rule mining.