Simultaneous Support Vector selection and parameter optimization using Support Vector Machines for sentiment classification

Abstract

Sentiment classification is widely used in some areas, such as product reviews, movie reviews, and micro-blogging reviews. Sentiment classification method is mainly bag of words model, Naive Bayes and Support Vector Machine. In recent years, the machine learning method represented by support vector machine (SVM) is widely used in the field of sentiment classification. There are more and more experiments show that support vector machine (SVM) performs better than the traditional bag of words model in the field of sentiment classification. However, more researches mainly focus on semantic analysis and feature extraction on sentiment, but also did not consider the case of sample imbalance. The purpose of this study was to test the feasibility of sentiment classification based on the genetic algorithm to optimize SVM model. Genetic algorithm is an optimization algorithm, which often used for selecting the feature subset and the optimization of the SVM parameters. This paper presents a novel optimization method, which select the optimal support vector subset by genetic algorithm and optimize SVM parameters. We construct the experiment show that the proposed method has improved significantly on sentiment classification than the traditional SVM modeling capabilities.