DefinePK

Abstract

In knowledge discovery and data mining, clustering is an aggressive concept to explore different attributes with different relations, because each data type has its own and unique challenge to achieve relative data based on partitioning of homogeneous data. In knowledge discovery categorical data clustering is an essential and challenging task because of special characteristics. So to arrange attributes in systematic manner for uncertain categorical data indexing approach is required. In this paper we propose and introduce A Novel Fuzzy based Partitioned Genetic Algorithm (NFPGA) for uncertain categorical data. This novel approach consists two phases to explore and process categorical data. In first stage partition data set with maximum number of clusters then combine all the clusters generated in first phase. This procedure repeated until number of clusters equal to pre-defined clusters present in data set. This proposed approach i.e. NFPGA is implemented on synthetic data sets which are available UCI repository, novel fitness function; cross-over and mutation operations are evaluated on categorical data based on parallel partitioning procedure. Performance of proposed approach has been crossover with different existing clustering related approaches with objective functionalities and similarity index measures, from this proposed approach gives better and excellence performance.