Pattern Retrieval through Classification from Pattern Warehouse

Abstract

The pattern is special kinds of data which are created through various data mining techniques and stored in the pattern warehouse through a specialized pattern management system (PMS). Pattern warehouse makes the pattern non-volatile or persists. Now a day's persistent pattern retrieval is a very new and important issue. This paper focuses on problems and challenges with pattern retrieval. One can see the applicability of classification in pattern retrieval as an opportunity and trying to bring attention to probable issues and challenges behind the physical implementation of this concept. This paper concluded that the applicability of classification in pattern retrieval is well feasible. It has also discussed that how of pattern classification is different with data's classification. Classification method should be defined in such a way that it can handle pattern efficiently. So far, little emphasis has been posed on developing an overall classification system for pattern retrieval. This paper concerns only association kinds of patterns. It has presented some issues regarding (i) Decision boundary of pattern classes. (ii) Problem of calculating a reliable estimate of pattern classes. (iii) How to define class boundary (iv) How to handle overlapping of pattern classes (v) Parameter selection for pattern classes estimation (v) Preprocessing of patterns (vi) How to handle classification on demand. (vii) Updating of pattern classes (vii) Finding optimal test conditions.