Sentimental Analysis Using Least Squares Twin Support Vector Machine

Abstract

Sentiment analysis is field of text mining in which reviews are in form of unstructured data so opinions can be extracted from overall opinion. This paper works on finding approaches that generate output with good accuracy. Least squares twin support vector machine (LSTSVM) is a quite new version of support vector machine (SVM) based on non-parallel twin hyperplanes. LSTSVM is an extremely efficient and fast algorithm for binary classification and its parameters depend on the nature of the problem. The goal of this paper is to improve the accuracy through LSTSVM. A result on several benchmark datasets is applied to train a sentiment classifier inorder to demonstrate the accuracy of the proposed algorithm. N-grams and different weighting scheme were used to take out the most classical features. It also analyzes Chi-Square weight features to select informative features for the classification. Experimental analysis reveals that by using Chi-Square feature selection in LSTSVM may provide significant improvement on classification accuracy.