Penerapan Information Gain Untuk Seleksi Fitur Pada Algoritma Naïve Bayes Untuk Analisis Sentimen Identitas Kependudukan Digital

Yuliana Nogo Welan (Institut Keguruan dan Teknologi Larantuka)
Alfian Nara Weking (Institut Keguruan dan Teknologi Larantuka)
Dominikus Boli Watomakin (Institut Keguruan dan Teknologi Larantuka)

Teknik: Jurnal Teknik Mesin, Elektro dan Ilmu Komputer
Vol. 5 No. 2 (2025): Juli : Jurnal Teknik Mesin, Elektro dan Ilmu Komputer

The rise of digital technology has driven the Indonesian government to implement Digital Population Identity (IKD) as a solution to enhance public services. However, user reviews on Google Play Store show diverse responses, requiring sentiment analysis to understand public perception. This study aims to improve sentiment classification accuracy on IKD app reviews using the Naïve Bayes algorithm optimized with Information Gain feature selection. The dataset consists of 1,000 Indonesian-language reviews manually labeled and preprocessed using text cleaning and TF-IDF feature representation. To address class imbalance, the SMOTE technique was applied. Experiments were conducted by comparing models without feature selection and balancing against those using Information Gain and SMOTE. Results indicate that the combination of Information Gain and SMOTE significantly enhances model performance, achieving 68,5% accuracy and 53,0% positive F1-Score. These findings confirm that Information Gain is effective in improving sentiment classification efficiency and accuracy. This study provides valuable insights for developing strategies to improve digital service quality. Kata kunci: Analisis Sentimen, Information Gain, Naïve Bayes, Seleksi Fitur, SMOTE.