Feature Selection Approach Based on a Novel Variant of Hybrid Differential Evolution and PSO for Sentiment Classification

Over the last few years, digital text data is increasing with the advancement in use of technologies. Therefore, sentiment analysis is an important step for improving the performance of any classifier. But with the increase in data, this problem has become more complex. To improve the overall performance of classifier and reducing the data-dimensionality, Feature selection is a key sub-stage for the sentiment analysis. In this paper, we propose a feature selection approach, which is based on a novel variant of hybrid differential evolution algorithm and PSO for sentiment classification. Experimental results show that our approach achieves 97.56% and 98.54% of average accuracy by using Naive Bayes (NB) classifier for the two Twitter datasets, viz., TestData_2015 and twitter sanders, respectively. The results also show that our proposed approach is showing better results with state-of-art algorithms on Twitter datasets. Statistical results also showcase the effectiveness of our proposed model.