Perbandingan Algoritma Machine Learning Umum Berbasis TF-IDF untuk Klasifikasi Artikel Bahasa Indonesia

Andi Gunawan (Universitas Kristen Maranatha Bandung)
Hendra Bunyamin (Universitas Kristen Maranatha Bandung)

Journal of Smart Technology and Engineering
Vol. 1 No. 1 (2025)

This study compares the performance of common machine learning algorithms in the classification of Indonesian news articles. A Dataset of 2160 articles from Detik.com was pre-processed and transformed into feature vectors using the Term Frequency-Inverse Document Frequency (TF-IDF) technique. The algorithms tested were Multinomial Naïve Bayes, Bernoulli Naïve Bayes, K-Nearest Neighbor, Random Forest and AdaBoost. Hyperparameter tuning was conducted using 5-fold cross-validation, and evaluation metrics included accuracy, precision, recall, and F1-score. The results indicate that Multinomial Naïve Bayes, with alpha set to 0.1, achieved the best overall performance with an accuracy of 0.8781, precision of 0.8138, recall of 0.8143, and F1-score of 0.814.