M-CAT FEATURE EXTRACTION AND SVM CLASSIFICATION BASED OPINION MINING

Abstract

Online social media is now an online discourse in which people participate at an impressive rate to build content, post it, bookmark it, and network. Opinion Mining (OM), often referred to as sentiment analysis, is the area of research that analyzes the thoughts, assessments, behaviors, and emotions of people regarding various individuals conveyed in textual input. This is done by categorizing a viewpoint into categories, such as positive, negative, or neutral. In e-commerce websites, opinion mining is very important, and also beneficial to individuals. As a result of user feedback an ever growing amount of results are stored on the web as well as the number of individuals who purchase products from the web increases. Reviews of shipper locations share their thoughts. For example, every organization, web forums, discourse groups, blogs, etc., would have a comprehensive knowledge add-up. Records those are functional for both suppliers and consumers of products on the Site. The method of seeking user opinion on the subject or product or issue is referred to as mining opinion. It can also be described as the process of automatic information extraction, which is called opinion mining, by means of opinions expressed by the consumer who is currently using the software with regard to a certain product. The analysis of the emotions from the opinions extracted is defined as Sentiment Analysis. The goal of opinion mining and Sentiment Analysis is to make computers capable of understanding and expressing feelings. This work concentrates on mining reviews from the websites like Amazon, which allows user to freely write the view. To do this, in this research work Cat Swarm Optimization algorithm, mRMR and Support Vector Machine algorithms are used. To improve the performance of CAT algorithm, it is modified by combining mRMR (Minimum Redundancy and Maximum Relevance) and proposed a new algorithm called M-CAT. Opinion mining is a three step process. In first step pre-processing work is done by Word Stemming, Spelling Check, Letter Replacement, Dialect Replacement. In Second step M-CAT algorithm is used for feature extraction. In third step by using SVM algorithm the result is classified as positive, negative and neutral. At the end we have used quality metric parameters to measure the performance of M-CAT algorithm compared with CAT algorithm.