Hybrid Multi Attribute Relation Method for Document Clustering for Information Mining

Abstract

Text clustering has been widely utilized with the aim of partitioning speci?c documents’ collection into different subsets using homogeneity/heterogeneity criteria. It has also become a very complicated area of research, including pattern recognition, information retrieval, and text mining. In the applications of enterprises, information mining faces challenges due to the complex distribution of data by an enormous number of different sources. Most of these information sources are from different domains which create difficulties in identifying the relationships among the information. In this case, a single method for clustering limits related information, while enhancing computational overheadsand processing times. Hence, identifying suitable clustering models for unsupervised learning is a challenge, specifically in the case of MultipleAttributesin data distributions. In recent works attribute relation based solutions are given significant importance to suggest the document clustering. To enhance further, in this paper, Hybrid Multi Attribute Relation Methods (HMARs) are presented for attribute selections and relation analyses of co-clustering of datasets. The proposed HMARs allowanalysis of distributed attributes in documents in the form of probabilistic attribute relations using modified Bayesian mechanisms. It also provides solutionsfor identifying most related attribute model for the multiple attribute documents clustering accurately. An experimental evaluation is performed to evaluate the clustering purity and normalization of the information utilizing UCI Data repository which shows 25% better when compared with the previous techniques.