DefinePK

Abstract

Pattern Deploying Methods performed good in discovering knowledge. These methods have given accurate results. Still it is observed that few of discovered patterns are holding noise knowledge instead of required and low frequency problem of long patterns. Hence we focused on perfect evaluation of discovered patterns by adapting the concepts of Deployed Pattern Evaluation (DPE) and Individual Pattern Evaluation (IPE). We used closed sequential algorithms to use the semantic information in the patterns to improve the performance and for accurate term weights we used d-patterns which use the evaluations of term weights based on the distribution of terms in documents. In this paper, terms are weighted according to their appearances in discovered closed patterns. Pattern Classification Models (PCM) Pattern Deploying Methods (PDS) resolved some extent the problems with low-frequency patterns. But still there is gap of pattern usage effectively can be resolved by our new approach. We also concentrated on ambiguous patterns influences in the documents. We made an analysis in comparison of other algorithms and methods hence our approach proved better.