A Fast Algorithm Combining FP-Tree and TID-List for Frequent Pattern Mining

Finding frequent patterns plays an essential role in mining associations, correlations, and many other interesting relationships among variables in transactional databases. The performance of a frequent pattern mining algorithm depends on many factors. One important factor is the characteristics of databases being analyzed. In this paper we propose FEM (FP-growth & Eclat Mining), a new algorithm that utilizes both FP-tree (frequent-pattern tree) and TID-list (transaction ID list) data structures to discover frequent patterns. FEM can adapt its behavior to the dataset properties to efficiently mine short and long patterns from both sparse and dense datasets. We also suggest a combination of several optimization techniques for effectively implementing FEM to speed up the mining process. The experimental results show that a significant improvement in performance is achieved.