Special Issue Article
Product Segmentation for Opinion Mining Using Probabilistic Principle Component Analysisin Customer Behaviors
|P.Saravanakumar1, Dr. A. Vijaya2
|Related article at Pubmed, Scholar Google|
Opinion mining plays a significant aspect in data mining to obtain the opinion of the user with regard to a product. Product is reviewed by the users to collect additional information about the product before they purchase that provides a strong decision to the users while purchasing a product. Works conducted on multiple reviewer-level features identified the measures for reviewers with a certain extent to subjectivity. At the same time the method Random Forest predicted the impact of reviews but did not worked with segmentation on the basis of different user opinions. The existing Variable Clustering (VC) algorithm, works on with the market segmentation for retailing based on customers’ lifestyle. But VC algorithm provided with the segmentation method did not guide full-proof method for different product decision. To guide different users with variety of products, Opinion Pattern Mining Segmentation (OPMS) based on the Probabilistic Principle Component Analysis (PPCA) report is proposed in this paper. OPMS segments the pattern based on different user opinion (i.e.,) behavior where the opinion is obtained using the result of PPCA report. PPCA report determines the maximum likelihood for the user estimation on the product reviews. PPCA report usage in opinion pattern mining reduces the dimensionality on the segmentation process using the covariance matrix. Efficient segmenting of user profiles obtains the users behavioral patterns (i.e.,) opinion pattern mining with increased threshold rate and decrease the false positive. Threshold and false positive rate are examined through factor analysis in the PPCA report. Probabilistic PCA in proposed work update the product reviews based on the user behavioral reviews. Experimental work uses the OpinRank Review Dataset information for Opinion Pattern Mining and improves the segmentation efficiency up to 8 % when compared with VC algorithm. OPMS is experimented on the factors such as Opinion Decision Threshold, False Positive Rate, Segmentation efficiency and User’s Product Trend Ratio Level.