FEATURE SELECTION OPTIMIZATION USING EVOLUTIONARY ALGORITHMS FOR SENTIMENT ANALYSIS

Abstract

With the growth of web 2.0 the data present online has grown tremendously. People express their views and opinions about various products and policies and these views are very important for gauging the reaction of people towards the products and policies through sentiment analysis. Selecting and extracting feature is a vital step in sentiment analysis and greatly influences the accuracy of sentiment classification. The statistical techniques of feature selection like document frequency thresholding produce sub optimal feature subset due to the Non Polynomial(NP) hard nature of the problem. Evolutionary algorithms are used extensively in optimization problems. Optimization techniques could be applied to feature selection problem to produce Optimum feature subset. They render feature subset selection by reducing feature subset dimensionality and computational complexity thereby increasing the classification accuracy. Firefly algorithm, an evolution based optimization algorithm is used in various optimization problems to produce an optimum solution. Here firefly algorithm is used for producing optimum feature set on four different datasets. Also firefly algorithm optimization results are compared with feature selection using genetic algorithm. Firefly algorithm here increases the performance of the classification considerably in terms of accuracy and training time required to train the classifier. It produces superior results as compared to the baseline model and feature selection using genetic algorithm.