SR-Mine: Adaptive Transaction Compression Method for Frequent Itemsets Mining

Extraction of frequent itemsets is a key step in association rule mining. Frequent Pattern (FP) mining from a very large dataset is still a challenging research problem. The basic frequent itemset algorithms are Apriori and FP-growth. FP-growth uses Frequent Pattern Tree (FP-tree) to store the database information in a compressed form. A large number of research papers have been proposed as an improvement of the basic frequent itemset mining algorithms. Several researchers have proposed modifications to existing data structures as well as new data structures to improve the mining process. A new method, Size Reduced Mining (SR-Mine), is proposed to speed up the FP-tree creation. The proposed work is implemented with the basic FP-growth algorithm and with the other two recent algorithms based on FP-tree. The three modified algorithms have been tested with standard datasets and compared with the original algorithms. The proposed method can be applied with the frequent itemset mining algorithms which consider each transaction one by one to construct a data structure for mining. The experimental results show that the proposed method can improve the performance of the mining.