SENTIMENT ANALYSIS USING NATURE-INSPIRED ALGORITHM

Abstract

The tremendous growth of Web 2.0 has changed the way people express their views and opinions. With the increasing amount of data and information on Web, feature selection is highly essential. As Selecting and extracting feature is itself a exhaustive task that it need to have some automated algorithms to reduce time and space complexity. Traditional techniques for feature selection help reducing feature subset but are of NP hard polynomial nature due to which we need to have some optimized solution. From the past few decades, swarm intelligence is used as optimization techniques for reducing feature subset by decreasing dimensionality and computational complexity resulting in increased accuracy. In this thesis, we have used Bat Algorithm with SVM for improvement in feature subset with increased accuracy. The algorithm is verified on two different sizes of datasets. Bat algorithm significantly outperformed other algorithms in selecting lower number of features by removing irrelevant, redundant and noisy feature maintaining the accuracy.