Sentiment Analysis of Komisi Pemberantasan Korupsi (KPK) on Twitter Social Media by applying the Algorithm Naïve Bayes Classifier

Abstract

Komisi Pemberantasan Korupsi (KPK) is an official institution in Indonesia specifically assigned to handle corruption cases. Along with the rise of corruption cases in Indonesia, the public also expressed opinions on the performance of the KPK, which was conveyed through Twitter. However, this opinion is still vaguely positive or negative. Therefore, in this study, sentiment analysis was carried out on this opinion using the machine learning-based naïve bayes classifier algorithm.

Data comes from Twitter which is taken by crawling technique through API (Application Programming Interface). The data is processed through several stages, namely preprocessing which includes removing punctuation marks, removing repetitive words and words that often appear but do not really have meaning in sentences. The next stage is data labeling which is done manually by assigning a label or class to the data. Next is the modeling process, which is the process of building an appropriate model to predict the probability of incoming data and classifying them according to the previous probability calculations. The data used in the modeling process is 2055 tweet data which is divided into training sets and testing sets with a ratio of 80:20. Next, a system deployment with the chosen model was carried out to analyze sentiment towards the KPK on Twitter.

The results of this study indicate that using the multinomial Naïve Bayes Classifier model, the precision value is 0.69, the recall is 0.89, the F-1 Score is 0.74, and the accuracy is 64%. In this study, a website was also developed to retrieve new data which then automatically classified it into positive, or neutral labels. This website also displays the results in the form of tables and graphs.