<![CDATA[The rapid expansion of social media usage in Indonesia has increased the spread of harmful online communication, including hate speech, which may contribute to social conflict and discrimination. As a result, automated hate speech identification has become an important research area in Indonesian natural language processing. Although many studies have applied machine learning and deep learning techniques for this task, comprehensive comparisons between conventional algorithms and transformer-based models in the Indonesian context remain limited. This study evaluates several machine learning algorithms, namely Naïve Bayes (NB), Support Vector Machine (SVM), Logistic Regression (LR), and Random Forest (RF), alongside the transformer-based IndoBERT model for Indonesian hate speech classification. All models were trained and evaluated using the same dataset, identical preprocessing stages, and consistent evaluation metrics consisting of accuracy, precision, recall, and F1-score to ensure fair comparison. Experimental findings show that IndoBERT achieved the strongest overall performance, reaching an accuracy of 87.45% and an F1-score of 84.92%. Among the classical machine learning approaches, Logistic Regression produced the highest result with an accuracy of 84.49% and an F1-score of 84.32%. While several machine learning models obtained relatively competitive recall values, IndoBERT demonstrated more stable performance across evaluation metrics and showed stronger capability in understanding contextual language patterns commonly found in Indonesian social media content. Overall, the study highlights the advantages and trade-offs between conventional machine learning and transformer-based deep learning approaches in Indonesian hate speech detection, while also providing practical insights for developing automated content moderation systems.]]>
