A Unified Framework of Sentimental Analysis for Online Product Reviews Using Genetic Fuzzy Clustering with Classification

N. Vijayalakshmi and Dr.A. Senthilrajan

The speedy growth of online technology there is a large amount of data plays in the web for internet consumers. Where the numerous of people express their views in their daily interaction which can be their sentiments or opinions about a particular thing. Bulky amount of data also found in the forms of reviews and ratings in lot of online market run websites stores such as, club factory, mantra, Amazon etc., In order to programme the enquiry of such data the area of SA is used. From a researcher‟s perspective, lot of social media sites release their APIs, prompting data gathering and analysis by researchers and developers. Therefore, this study recommends a novel metaheuristic cluster based classification model for sentimental analysis. The proposed method involves 4 major processes namely, (i) pre-processing of the tweets, (ii) feature extraction (FE), and (iii) clustering and (iv) classification. At the pre-processing stage, the raw tweets are gather from Twitter, have undesirable and fuzzy words, URLs, stop words etc., which will be completely removed before FE. In next level is put on the pre-processing, tweets are totally transformed into the feature vector (FV) by computing a set of 11 features from the Twitter dataset. The normalized FV is given input to the genetic algorithm with fuzzy clustering (GAFC) which uses genetic algorithm (GA) and fuzzy clustering technique to cluster the data. Finally, parameter tuning grey wolf optimization (GWO) using new kernel extreme learning machine (KELM) model is applied for classification purposes. To ensure the efficiency of the projected ideal, it is applied to a same set of four datasets namely Canon dataset, iPod dataset, DVD and Nokia dataset collected from Amazon website. Results gained obviously approve the supremacy of the established model in terms of different measures.

Volume 12 | 01-Special Issue

Pages: 301-310

DOI: 10.5373/JARDCS/V12SP1/20201076