A prototype system for the analysis of sentiment regarding drug and food issues in Indonesia

Background: Drug and food policies or issues can give rise to various opinions in the community. In Indonesia, such perspectives can be quickly identified with the use of big data analytics software that can perform a sentiment analysis of Twitter content and complaints received by the country’s National Agency of Drug and Food Control. These tools are employed to examine positive or negative views regarding the aforementioned policies or issues. In consideration of these matters, this study was aimed at developing a prototype system for the analysis of sentiment regarding drug and food issues and integrating itwith Indonesia’s information and complaints service network. Method: To the above-mentioned ends, the research adopted a supervised machine learning classification method called a naive Bayes classifier (NBC), which was incorporated with evaluation parameters, namely, accuracy, precision, recall, and the F-measure. The systems development life cycle was used as a framework. The system was then tested using 10-fold cross-validation. Results: The validation results showed that the highest accuracy exhibited by the system was 88%. It registered a precision of 81%, a recall of 100%, and an F-measure of 90% on 540 tweets that served as training data and 60 tweets that served as testing data. Sentiment analysis results are displayed on a dashboard. Conclusion: An NBC exhibits reasonably good performance in the examination of sentiment toward drug and food issues in Indonesia. Sentiment data can serve as input that facilitates rapid response to these problems and subsequently clears the way for formulating appropriate education strategies for the community.