Automating Public Complaint Classification Through JakLapor Channel: A Case Study of Jakarta, Indonesia

Abstract

The DKI Jakarta provincial government is ready to support the digital transformation program with a series of digitally integrated policies. Residents of DKI Jakarta can now easily submit complaints about problems in their surrounding environment through the JakLapor service feature on the JAKI application. However, incoming reports are still manually classified. As a result, many citizens still report unsuitable complaints based on their category. This research aims to compare and find the best complaint classification model by applying multiple machine learning models to classify texts automatically. We also use feature extraction to see which model performs the best. This study employed Support Vector Machine (SVM), Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Adaptive Boosting (AdaBoost) algorithms as the machine learning model. Meanwhile, we use Count Vectorizer, Terms Frequency-Inverse Document Frequency (TF-IDF), N-Gram, and Latent Semantic Analysis (LSA) as the feature extraction algorithms. The classification results show that the Random Forest method model with TFIDF feature extraction is the most accurate and optimal model among the others, with a 90% accuracy rate.