Enhanced Cluster Ensemble Approach Using Multiple Attributes in Unreliable Categorical Data
Abstract
Cluster analysis is an efficient tool to identify useful and user-preferable data patterns from categorical data streams. Conventional clustering approaches focused on numerical with single attribute relations from categorical data. Existing approaches perform poorly and have low complexity to combine relative attributes, whether information is present or hidden. Therefore, our proposed Enhanced Categorical Cluster Ensemble Approach (ECCEA) to classify data depends on various different attributes from multidimensional data sources. ECCEA creates a matrix and then converts this matrix into attribute groups with the help of the graph method. Practical outcomes show an effective clustering result with multi-attribute relations with respect to associated attributes from categorical data sets. Further improvement of our proposed approach is to perform well on their corresponding types of attributes to improve the performance with respect to multi-attribute similarity determined for feature-based data exploration using clustering.