An Efficient Algorithm for High-Utility Itemset Mining Using Genetic Algorithm

High-utility itemset mining (HUIM) is an expansion of frequent itemset mining (FIM) which has attracted much attention in recent years. The “frequent itemset mining” problem which was very important in the past has some shortcomings and it does not take into account the utility of items. The “high-utility itemset mining” problem was introduced to remove this shortage. One way to solve the HUIM problem is to investigate all possible states.However, checking the entire state space is timeconsuming when the number of items is high. As the result, efforts have been made to solve this problem using optimization algorithms. One of the first things to do was the HUPEumu-Ga algoritm based on genetic algorithm (GA). However, this technique is not applicable for transactions with a large number of items. In this paper, an adapted genetic algorithm is presented to mine highutility itemsets (HUIs). The proposed method consists of a preprocessing step which eliminates low utility itemsets and then uses a genetic algorithm (GA) equipped with alocal search mechanism to mine high utility itemsets. Also, in the proposed method, the probability of mutation of genes can vary at different times. Our proposed model was compared with three optimizationmethods of HUIM-BPSOsig, IHUPEumu-GA, and Binary PSO in order to evaluate the efficiency of theproposed method in terms of the number of HUIs.

Keywords:

High-Quality Itemset Mining , Genetic Algorithm , Local Search , Optimization