DT-FNN based effective hybrid classification scheme for twitter sentiment analysis

Abstract

Sentiment analysis refers to the interpretation and computational study of emotions, opinions and appraisals within the text data using text analysis methods. A basic aim of sentiment analysis is to categorize the sentiment polarity of the sentences, document or aspects. Product manufacturers use the knowledge from sentiment analysis for improving their services & products. Hence, there is an atrocious need of an efficient technique that can accurately identify the sentiment polarity of the content. The supervised classification algorithm has been proved favourable for most of the sentiment analysis task and is widely used in opinion mining. This study presents a novel method for sentiment analysis by combining two supervised classification algorithms viz. Decision Tree (DT) and Feed Forward Neural Network (FNN). Pre-processing of data is carried out by using Independent Component Analysis (ICA) and Windowed Multivariate Autoregressive Model (WMAR) is introduced for extraction of potential features. Then highest scores are extracted using Improved Bat Algorithm (IBA) technique and finally, the experimental results are compared with existing algorithms i.e. ID3, J48 and Random forest classifier. The proposed method significantly outperforms the existing sentiment classification methods with accuracy of 97.84%.