Comparison of Naive Bayes smoothing methods for Twitter sentiment analysis

In sentiment analysis, the absence of sample features in the training data will lead to misclassification. Smoothing is used to overcome this problem. Previous studies show that there are differences in performance obtained by the various smoothing techniques against various types of data. In this paper, we compare the performance of Naive Bayes smoothing methods in improving the performance of sentiment analysis of tweets. The results indicated that Laplace smoothing is superior to Dirichlet smoothing and Absolute Discounting with the micro-average value of F1-Score 0.7234 and macro-average F1-Score 0.7182.