Enhanced Facet Ranking and Text Classifier for Opinion Mining

The enormous growth and usage of social networks offer positive ways to any business by sharing the emotions, feelings and experiences. Web users are benefited with valuable online reviews. To utilize the reviews effectively, researchers are working on necessary methods and ideas such as classification of positive and negative sense of reviews, ranking the facet in the reviews to make the effective classification etc. This study aims to propose a novel facet identification namely Facet Based Adjective identification method (FBAI) for efficient feature selection of reviews. The next algorithm FacetRank marks facet of each opinion from review set with positive, negative and neutral polarity. To classify the ranked facets, a novel Cluster based k Nearest Neighbor (C-kNN) algorithm is proposed. Constrained single pass clustering algorithm is combined with existing kNN classification algorithm to classify the review set as positive or negative. C-kNN reduces the resemblance checking calculation and can process high dimensional data which enable dynamic classification. This analysis takes household product reviews as input data set. The ranked review set (using FBAI+FacetRank) is given to kNN and C-kNN for classification. F1 score of C-kNN 2.43 % higher than kNN. Linear time complexity of C-kNN achieved is 68% of kNN.