SIBA: A fast frequent item sets mining algorithm based on sampling and improved bat algorithm

The adaptability of the classical top-k frequent item sets mining algorithms is unsatisfied when being applied to some large datasets with both large scales of transactions and items. In this work, an algorithm called SIBA (sampling, improved bat algorithm (BA)) is proposed to solve this problem. SIBA applies BA and improves it by cloud model to search frequent item sets from a large number of items rapidly, and builds the sub-dataset in every iteration to reduce the scanning cost. Based on four open access datasets, it is compared with Apriori, FP-growth, and other heuristic algorithm such as PSO (Particle swarm optimization) and GA (Genetic algorithm). The experimental results show that when being applied to frequent item sets mining, SIBA is faster than Apriori and FP-growth, and meanwhile, it is more robust than PSO and GA.