Efektifitas Hasil Analisis Sentimen Aplikasi SIGNAL Berbasis Lexicon-Based dan Random Forest

Abstract

SIGNAL (Samsat Digital Nasional) is a digital innovation developed by the Indonesian National Police to simplify vehicle tax payments, STNK validation, and other administrative services online. As the number of users grows, various user opinions are reflected in the form of reviews on the Google Play Store. The research adopts a lexicon-based approach by extracting positive and negative keywords directly from the dataset to classify sentiments in user-generated reviews. A sentiment label is assigned based on the frequency and dominance of positive or negative terms within each review. To evaluate the effectiveness of this lexicon-based classification, the Random Forest machine learning algorithm is employed as a benchmark. These findings indicate that the lexicon-based approach, when built from domain-specific vocabulary, can effectively classify sentiment with minimal computational resources while maintaining competitive performance. This research contributes to the development of lightweight sentiment analysis systems and highlights the potential of hybrid methods for enhancing accuracy.