Spam and Sentiment Analysis Model for Twitter Data using Statistical Learning

The past empirical work of twitter spam detection and sentiment analysis is based on random selection of features for the generation of classification models. This paper focus on the selection of model by applying multiple linear regression using stat models for fitting n dimensional hyper plane predictor (i.e. Twitter features) to our response variable (i.e. Spam and sentiments). This paper includes following parts: 1) Spam Detection Classifier 2) Bayesian and Log-Likelihood based sentiment classifier 3) Evaluation of classification system using different machine learning algorithms (i.e. Binomial, CART and Random Forest). Our experimental evaluation demonstrates that the efficiency of Random Forest is higher compared to other algorithms of the proposed classification system.