Abstract
High utility itemset mining is an active research problem in data mining. Because traditional high utility itemset mining algorithms cannot cope with the exponential growth of search space, the heuristic high utility itemset mining algorithms have been widely studied. To solve the problem of itemset loss caused by the early convergence of heuristic high utility itemset mining algorithms, a new algorithm is designed to discovering more high utility itemsets within fewer iterations. In this paper, the proposed strategy of positional evolution based on the female elephant factor is proposed to reduce effectively the search space and improve the execution efficiency of the algorithm. Moreover, in order to prevent the algorithm from converging too quickly and falling into local optimum, the proposed strategy of two-phase population diversity maintenance which keeps a balance between population diversity and convergence. Extensive experiments on real datasets show that the proposed algorithm outperforms the advanced heuristic high utility mining algorithms.