Sentiment Analysis of Covid-19 Bansos Issues on Twitter using Chi-Square and Naïve Bayes

Abstract

Social assistance or what is often called Bansos (Bantuan Sosial) is assistance in the form of goods or money from the government for the community that is temporary and selective. There were many reports from the public complaining that some had not received assistance or had not received it at all. This problematic social assistance has caused a stir in public reports on various social media, including Twitter. To find out the classification of public opinion related to social assistance on Twitter, it is necessary to do pre-processing and analysis of tweets. The dataset used is data crawled using the Twitter API and produces 702 tweet data in CSV format. Tweets retrieved based on the keyword 'bansos' in August 2021. The dataset is divided into two categories, positive and negative. Data with a total of 328 positive categories and 374 data of negative categories. The method applied in this study uses the Chi-Square feature selection method and the Naïve Bayes Classifier algorithm. The purpose of this research is to produce a website-based application that can classify tweets related to social assistance covid-19 into 2 categories, positive and negative by using the Chi-Square feature selection and the Naïve Bayes Classifier algorithm. The F1 score in the positive class is 85% and the negative class is 89% and produces an accuracy value of 87%. The results of the comparison between Naive Bayes and Naive Bayes using Chi-Square show that there is no difference in accuracy.