Looking into the possibility for designing normal distribution based dissimilarity measure to discover time profiled association patterns

Abstract

This research addresses the design of a novel dissimilarity measure for mining similar patterns from time stamped temporal databases applying the concept of standard score and normal distribution. The basic idea behind the design of dissimilarity measure is to use and transform supports to z-space and compute the probability of z-score of temporal patterns. The probability is obtained using normal distribution chart. The objective has been to design a normal distribution based dissimilarity measure which can be used to discover all valid similarity-profiled temporal association patterns.