Implementation of Predictive Models and Text Classification on Twitter Data

Abstract

Twitter is the social news and network blogging service, “what happening now in the world and what people are talking about right now”.  It has Monthly on average 330 million active users(1).It is  one of the best platform to share views in the form of tweets and re-tweets.  If we process the Twitter data set we can get sentiment analysis of the people who are participating. People are very interesting about the future predictions of the current leading issues and talks in the twitter regarding any issues.  In this paper we prepared a prediction model for sentiment analysis using machine learning model, that prediction model generates number of positives, negatives and neutrals from the data set. In this way more data sets are predicted for the sentiments. New model is designed for future predictions on the sentimental numerical data by applying the regression models, and that model is applicable to test data in getting the future prediction. In this paper we design a sentiment prediction model for GST and future prediction.  Here the data set is Goods and service tax (GST) of the India, collected from the twitter for one month, clean the GST data set and then predict the sentiment using naive bakes supervised prediction algorithm. Finally we applied regression machine learning algorithms to find and asses the near future sentiments of the public on GST, if we compare the regression methods, we can observe that linear regression model is the best fit model for the GST twitter data to predict near future sentiments of the twitter users.