Maximal itemsets mining algorithm based on Bees' Algorithm

Abstract

Maximal frequent itemset is the largest frequent itemset in a database which is not covered by other itemsets. All frequent itemsets can be built up from maximal one. Moreover, it is possible to focus on any part of the maximal frequent itemset to supervise Data Mining. Bees' Algorithm is simple, robust and population-based stochastic optimization algorithm which is based on bees' natural foraging habits. It performs a neighborhood search combined with random search. This paper produces a turning on by pairing of maximal frequent itemset and Bees'Algorithm; it presents maximal frequent itemset — oriented Bees'Algorithm to mine maximal frequent itemsets from transactional databases which has been named Mining Maximal Itemsets Bees' Algorithm (MMIBA). The fitness, coding, scout bees' duties, type of harvested information, and termination criteria have been oriented to the problem of maximal frequent itemset mining. MMIBA was applied on real life databases available publicly on the Internet which are chess, Mushroom, Cancer Cells, Census data, and Dense Census. The experiments were accomplished depending on three levels of minimum support threshold, which are twenty five, fifty, and seventy five percentages of the databases' sizes, to validate the efficiency of MMIBA. The level twenty five percentage was depicted to elucidate the ability of MMIBA in mining the MFIs with low values of minimum support, while the level seventy five percentage to validate its ability in high minimum support values.