An Efficient Visual Approach For Automatic Clustering And Validation

Abstract

Clustering or exploratory data analysis is a widely applied newlineunsupervised technique in the data mining domain The major concern of the newlinedomain is how the observed data can be categorized into meaningful newlinestructures However most of the existing clustering algorithms are not newlineadequate in dealing with arbitrarily shaped distribution of data such as data newlinesets of extremely large volume data visualization and data sets of highdimensional newlinefeatures The key limitations of Indexbased and Statisticalbased newlinecluster validation methods are that making unrealistic distributional newlineassumptions of the data and incurring high computational cost in cluster newlineanalysis which prevents the clustering algorithms from being efficiently used newlinein practiceThe count of clusters is considered as a key factor for clustering newlineoperations in most of clustering algorithms Therefore the quality of the newlineresultant clusters mainly depends on the assessment of cluster number The newlineClustering of unlabeled data set faces certain critical issues such as assessing newlinecluster tendency ie determining the number of clusters prior to clustering newlinegrouping the data into meaningful sets and validating the formed clusters newlineThe visual methods for various data analysis problems have been newlineextensively studied and the abstract data have been represented visually to newlineamplify cognition Visualization is considered to be one of the most newlineinstinctive methods for cluster detection and validation especially for newlineperforming well on the depiction of irregularly shaped clusters as a newlinepreclustering method The visual data mining allows the data miners and newlineanalysts to evaluate monitor guide the inputs products and process from the newlineresults of visualization techniques The Visual Clustering Analysis VCA is a newlinewide assortment of image processing techniques information visualization newlineand cluster analysis techniques The visualization used in the cluster analysis newlinemaps the highdimensional data with a 2Dimensional space and aids the newlineusers to have an intuitive and easily understood graph or image to newline newline