In this paper, we investigated deve

In this paper, we investigated developing a classification model for sentiment in micro-blogging posts from Twitter. Using various preprocessing techniques, and applying various feature selection techniques to the Naïve Bayes classifier, we were able to achieve reasonably good performance for the training set used. Ultimately we also noticed that all the classifiers trained were performing slightly better for classifying the positive class compared to the negative class. We show that Naïve Bayes algorithm with application of Information Gain measured using Chi square with minimum value of 3 to select high information features, gives accuracy above 89%. This study can further be extending by applying other techniques such as feature selection and aggregation using n-gram collocations and Parts-Of-Speech tagging to improve the feature model by eliminating word disambiguation.