DefinePK

Abstract

Generally, data mining concept is used to gather information from various data repository. Frequent pattern mining is to be designed for displaying repetitions in the transactional database. Patterns are defined by predefined format. In this evolutionary work, the proposed concept to mine the transactional database using the combination of recommendations and prediction by the help of software simulation. In hardware side, this process is explained in pattern mining using systolic tree creation. This will handle pattern mining to configure the frequent pattern while generating systolic tree structure. But it can handle certain size of dataset only, but also generate more candidate item set when implementing the item set matching by tree projection algorithm. This will occupy more and more memory, each time reevaluation done from the scratch of dataset in hardware side. It is required more time to process. To overcome this problem, in software side to implement HI-Growth tree technique to analyze large scale of dataset. The new concept is introduced based on recommendation approach to avoid candidate set generation. This method is achieved to reduce the internal memory and mining time is by dividing the frequent pattern into the dense and the sparse patterns. In this paper, investigate the mining speed of the HI-Growth tree is fast-paced than original software side algorithm of FP-Growth tree, and also it consumes less amount of memory for analyzing dense and sparse pattern through recommendation technique to improve the mining efficiency, while achieving the higher throughput to overcome the defect of hardware approach.