<![CDATA[Abstrak - Peningkatan aktivitas media sosial menyebabkan munculnya akun palsu (fake account) yang berpotensi menyebarkan informasi menyesatkan. Penelitian sebelumnya, seperti Jalal dan Ghafoor [1] serta Zahra et al. [2], menekankan pentingnya analisis perilaku dan penggunaan algoritma Machine Learning untuk mendeteksi akun bot. Namun, kebanyakan studi belum mengoptimalkan integrasi fitur perilaku dengan fitur Social Network Analysis (SNA). Penelitian ini menawarkan kontribusi ilmiah dengan menggabungkan fitur aktivitas pengguna dan fitur jaringan sosial ke dalam model deteksi akun palsu berbasis Machine Learning. Dua algoritma, yaitu Random Forest dan Support Vector Machine (SVM), digunakan untuk mengevaluasi efektivitas model. Hasil penelitian menunjukkan bahwa penambahan fitur SNA meningkatkan kinerja model secara signifikan. Algoritma Random Forest menghasilkan akurasi 91,1% dan F1-score 90,9%, lebih tinggi dibandingkan SVM. Temuan ini sejalan dengan pendekatan multimodal yang disarankan dalam Social Network Analysis and Mining (2025), namun penelitian ini memberikan kontribusi baru pada penggunaan kombinasi fitur jaringan dan perilaku dalam konteks data lokal. Penelitian ini diharapkan dapat mendukung pengembangan sistem deteksi akun palsu yang lebih akurat dan adaptif di berbagai platform media sosial.Kata kunci : Akun Palsu; Bot Media Sosial; Machine Learning; Social Network Analysis; SVM; Random Forest;Abstract - The rapid growth of social media has increased the spread of misleading information, often amplified by fake accounts and automated bots. Previous studies, such as Jalal and Ghafoor [1] and Zahra et al. [2], emphasized behavioral analysis and machine learning techniques for bot detection. However, many existing approaches have not fully optimized the integration of behavioral features with Social Network Analysis (SNA). This study provides a scientific contribution by combining user activity features and social network features in a machine-learning-based fake account detection model. Two algorithms, Random Forest and Support Vector Machine (SVM), were used to evaluate model performance. The results show that integrating SNA features significantly improves detection accuracy. The Random Forest model achieved 91.1% accuracy and a 90.9% F1-score, outperforming SVM. These findings align with the multimodal detection approach suggested by Social Network Analysis and Mining (2025), while offering a new contribution through the application of combined behavioral and network features on local datasets.This study is expected to support the development of more accurate and adaptive fake account detection systems across various social media platforms.Keywords: Fake Account Detection; Machine Learning; Social Network Analysis; Social Media Bots; Support Vector Machine; Random Forest;]]>
