Kaala vrksha

Abstract

Discovery of frequent itemsets from snapshot databases is most addressed widely in the literature. The support value of itemsets for frequent itemset mining is a numeric value of one dimension. In contrast to traditional frequent pattern mining the discovery of similar item sets from time stamped transaction datasets is recent research interest that is gaining immediate attention and interest from academia and industry. In similarity profiled temporal pattern mining, support of temporal items and temporal item sets is multi-dimension support time sequence. The idea is to obtain set of similar temporal item sets whose support values at different timeslots vary similar to the reference sequence. The challenge is to obtain set of all temporal item sets with minimum computation time and computation space. This research proposes a tree structure, KAALA VRKSHA for temporal pattern mining that finds the similarity between temporal patterns by applying Gaussian based distance function. Experiments are conducted to compare performance of proposed method to existing approaches NAIVE, SEQUENTIAL, SPAMINE and G-SPAMINE. Results proved that proposed method outperformed existing approaches and comparatively required lesser computational time and space.