Article Preview
Top1. Introduction
Now a day’s many applications generating huge volumes of data (Vazirgiannis et al., 2003). Many techniques have been developed to extract knowledge, especially in the context of data mining. The results of such operations are abstract and compact representations of the original data, which called patterns (Catania, 2004). The pattern gives the semantic representation of raw data. There is required raw data for pattern and vice-versa mapping (Zdenka, 2012). The volume of extracting patterns from various knowledge discovery applications is increasing rapidly, so there is need for effective and efficient pattern management system (Fernández et al.,2011; Mohammad et al., 2009; Vazirgiannis et al., 2003). Patterns can come from clustering, association rules, classification, probabilistic rules and many other ways. Moreover, patterns can be found in images, signals, text and music and of course in the World Wide Web (Barbara and Anna, 2005). Theses all kinds of patterns are stored in pattern warehouse through PBMS (Pattern Base Management system) (Catania, 2004; Jaesoon et al., 2002; Manolis et al., 2007). There are no tools or systems that are designed to deal exclusively with pattern retrieval. All current technologies just try to extend the existing query system to get patterns, but this is not effective because of a special characteristic of patterns and its diversity (Evangelos et al., 2005; Rizzi, et al., 2003; Rizzi, 2004). Pattern retrieval can be improved through applying data mining techniques itself. Problem of diversity or complex structure can be solved by grouping relevant patterns. It is called classification of patterns. This task is not easy because patterns are semantically rich and have different and specialized structure. The main issue of this paper is revealing the difficulties with classification of association type patterns in the context of pattern retrieval. Association rule mining is a popular method for discovering interesting relations between variables in large databases (Thakur et al., 2006; Tiwari and vipin, 2010). Patterns can be a set of association rules discovered from databases using different measures of interestingness (Riccardo et al., 2011).
The objectives of this paper are to discuss the need of pattern management, benefit of pattern classification in context of retrieval and discuss the problems and challenges with pattern classification. The aim only point out the problems which may occur in applying data mining techniques for pattern retrieval on pattern warehouse rather than generating patterns itself. Classification is one of the powerful data mining techniques. For simplicity, research work revolves around the challenges with the classification of patterns.