Sentiment Score Analysis and Topic Modelling for GST Implementation in India

Abstract

Sentiment analysis has been widely used as a powerful tool in the era of predictive mining. However, combining sentiment analysis with social network analytics enhances the predictability power of the same. This research work attempts to provide the mining of the sentiments extracted from Twitter social application for analysis of the current trending topic in India, i.e. Goods and Services Tax (GST) and its impact on different sectors of Indian economy. This work is carried out to gain a bigger perspective of the current sentiment based on the live reactions and opinions of the people instead of smaller, restricted polls typically done by media corporations. A variety of classifiers are implemented to get the best possible accuracy on the dataset. A novel method is proposed to analyse the sentiment of the tweets and its impact on various sectors. Further, the sector trend is also analysed through the stock market analyses and the mapping between the two is made. Furthermore, the accuracy of stated approach is compared with state-of-the-art classifiers like SVM, naive Bayes and random forest and the results demonstrate accuracy of stated approach outperformed all the other three techniques. Along with this, topic modelling was also done to get a picture of trending topics that are linked to GST. LDA and text ranking algorithms were applied to get connected topics.