Single Scan Polynomial Algorithms for Frequent Itemset Mining in Big Databases

This paper considers frequent itemset mining in big transactional databases. It first introduces a novel approach (Bio-SS) that combines the bio-inspired algorithms with the single scan algorithm (SSFIM). The proposed approach addresses the limitations of SSFIM by utilizing the bio-inspired operators in the generation process. This reduces the time complexity of SSFIM from exponential to polynomial, while taking advantage of the capacity to derive the frequent itemsets by performing a single database scan, independently from the minimum support value. This allows to considerably accelerate the scan procedure compared to existing approaches, especially when dealing with large scale databases. The numerical results show that the designed Bio-SS outperforms both accurate and metaheuristics baseline FIM approaches when dealing with big databases.