Using Naïve Bayes Classifier for Application Feedback Classification and Management in Bahasa Indonesia

The world keeps moving, software products too. An application's objectives, structures, requirements, and assumptions that have been elicited and analyzed previously may need to be reassessed and updated. In order to fully understand these requirements evolutions, what changes are necessary, and why those changes are needed, one essential source of requirements is user feedback. However, handling and analyzing so many user feedbacks can be time-consuming. Using natural language processing tools for Bahasa Indonesia and Naive Bayes classifier, this research aims to develop a tool to process natural language and classify user feedbacks. The developed tool is expected to make feedback classification less time-consuming so that developers can project their energy to more productive and creative works. The machine learning models are built using the feedback dataset taken from an up-and-running university e-learning system and show promising results. The highest confusion matrix scores are 92.5% for accuracy, 85.6% precision, 85.1% recall, and lastly, 85.4% for the F-measure score. The resulting web application for feedback management is then evaluated to the users, and even though it still needs to be further polished and improved for real industrial use, it is perceived to be useful and easy to use.