BIG DATA IN ACTION: ANALYZING TWITTER SENTIMENTS THROUGH MACHINE LEARNING AND NATURAL LANGUAGE PROCESSING

Sentiment analysis (SA) is essentially a subfield of data mining and natural language processing that is primarily concerned with the problem of extracting useful knowledge from comments posted by users of the web. However, academic studies of various topics in South Africa have been underway for over than ten years. The method of present study incorporates a machine learning and natural language processing of sentiments on data of twitter. Synthetic Minority Over-sampling Technique (SMOTE) has been employed in the identification of samples that are from the minority class to create synthesized samples. Bag of Words (BoW) is applied here for the extraction of features since textual data require a good representation before further analysis. The evaluation metrics employed are accuracy, precision, recall and F1 score which are applied on several classification models such as MLP, LR and PSO. From the data it emerges that the LR model is the best one in this case with 88% accuracy, 83% precision, 81% recall and 82% F1-score. This paper suggests directions for further research in the subsequent social media analytics by proving the efficiency of machine learning approaches in handling sentiment analysis.