Sentiment analysis on social media against public policy using multinomial naive bayes

Purpose: The purpose of this study is to analyze text documents from Twitter about public policies in handling COVID-19 that are currently or have been determined. The text documents are classified into positive and negative sentiments by using Multinomial Naive Bayes. Methods/Study design/approach: In this research, CRISP-DM is used as a method for conducting sentiment analysis, starting from the business understanding process, data understanding, data preparation, modeling, and evaluation. Multinomial Naive Bayes has been applied in building classification based on text documents. The results of this study made a model that can be used in classifying texts with maximum accuracy. Result/Findings: The results of this research are focused on the model or pattern generated by the Multinomial Naive Bayes Algorithm. The classification results of social media users' tweets against the new normal policy obtained good results with an accuracy value of 90.25%. After classifying the tweets of social media users regarding the new normal policy, the results show that more than 70% agreed and supported the new normal policy. Novelty/Originality/Value: This study resulted in how classification can be done with Multinomial Naive Bayes and this algorithm can work well in recognizing text sentiments that generate positive or negative opinions regarding public policies handling COVID-19 . So, the research provided conclusions about the views of people around the world on new normal public policy.