Partition based single scan approach for mining maximal itemsets

Abstract

In this paper, Frequent Itemset mining (FIM) limitations and compact representation of FIM that is Maximal Itemsets explored for extracting unknown redundant less frequent itemsets from the transactional database. The candidate generation and support calculations are the major tasks in FIM. FIM challenges with the following limitations for low support threshold: (i) huge frequent itemsets are generated as output (ii) difficulty in taking decision among the frequent itemsets due to redundancy. (iii) To find the cumulative support of itemsets, database scan is required for each length. The first issue can be resolved using maximal itemsets that are frequent who doesn’t have any superset is called as Maximal Itemset Mining (MIM). Maximal itemsets are useful in minimal key discovery kind of applications. Hence, we present a Single scan algorithm to address the above limitations. However, several unnecessary itemsets are being hashed in the buckets. To overcome the limitations, a partition-based approach is proposed. Empirical evaluation and results visualize that the PSS-MIM outperforms all candidate generate and other approaches.