An Enhanced Spectral Clustering for Overlapping Data in Multiple Task Clustering

Clustering is one of the most widely used approaches for exploratory data analysis in data mining. In large scale data sources, multitask clustering is an important research work to handle overlapping data, negative and non-negative values among clustering of multiple tasks which is used to improve the learning relationship among related tasks and sharing of information across the tasks. Recent past spectral clustering has become the well accepted clustering algorithm to perform multitask clustering which relies on the eigenstructure of a similarity matrix and outperforms partitioning of data with more complicated structures than traditional clustering algorithms such as the Kmeans and Fuzzy C-means. This research work proposes an enhanced multitask spectral clustering (EMTSC) technique to perform multitask clustering without overlapping in data points among clustering of multiple tasks in large data sets for improving clustering performance.