Modified FP-Growth: An Efficient Frequent Pattern Mining Approach from FP-Tree

摘要

Prefix-tree based FP-growth algorithm is a two step process: construction of frequent pattern tree (FP-tree) and then generates the frequent patterns from the tree. After constructing the FP-tree, if we merely use the conditional FP-trees (CFP-tree) to generate the patterns of frequent items, we may encounter the problem of recursive CFP-tree construction and a huge number of redundant itemset generation. Which also leads to huge search space and massive memory requirement. In this paper, we have proposed a new data structure layout called Modified Conditional FP-tree (MCFP-tree). Moreover, we have proposed a new pattern growth algorithm called Modified FP-Growth (MFP-Growth), which uses both top-down and bottom-up approaches to efficiently generate the frequent patterns without recursively constructing the MCFP-tree. During mining phase only one MCFP-tree is maintained in main memory at any instance and immediately deleted or discarded from the memory after performing the mining. From the experimental analysis, it is noticed that the proposed MFP-Growth algorithm requires less memory to construct the MCFP-tree as compared to conditional FP-tree. Moreover, the execution of the MFP-Growth method is found significantly faster than the traditional FP-Growth as it does not generate redundant patterns.