PERBANDINGAN ALGORITMA NAÏVE BAYES DAN SUPPORT VECTOR MACHINE UNTUK DETEKSI SPAM KOMENTAR PADA PLATFORM TIKTOK

mochamad andriansyah (Universitas Muhammadiyah Sukabumi)

Jurnal Informatika dan Teknik Elektro Terapan
Vol. 14 No. 3 (2026)

Abstrak. Perkembangan platform media sosial TikTok yang pesat menyebabkan meningkatnya interaksi pengguna dalam bentuk komentar, namun juga memunculkan permasalahan spam berupa promosi tidak relevan, tautan mencurigakan, konten penipuan, dan pesan bot. Penelitian ini bertujuan membangun model klasifikasi komentar spam dan non-spam pada TikTok menggunakan algoritma Naïve Bayes serta membandingkan performanya dengan Support Vector Machine (SVM). Metode penelitian menggunakan kerangka kerja CRISP-DM yang meliputi Business Understanding, Data Understanding, Data Preparation, Modeling, Evaluation, dan Deployment. Dataset terdiri dari 1.300 komentar TikTok yang terbagi seimbang menjadi 650 komentar spam dan 650 komentar non-spam. Preprocessing teks meliputi case folding, cleansing, tokenizing, stopword removal, dan stemming, dilanjutkan pembobotan fitur TF-IDF (max_features=2000, ngram_range=(1,2)). Hasil pengujian menunjukkan Naïve Bayes mencapai akurasi, precision, recall, dan F1-score sebesar 98,46%, sedangkan SVM mencapai performa sempurna 100% pada seluruh metrik. Perbandingan menunjukkan SVM lebih unggul dibandingkan Naïve Bayes dalam mendeteksi komentar spam pada TikTok. Penelitian ini diharapkan menjadi referensi pengembangan sistem moderasi komentar otomatis berbasis machine learning pada media sosial.   Abstract. The rapid growth of the TikTok platform has increased user interaction through comments, but has also given rise to spam problems such as irrelevant promotions, suspicious links, fraudulent content, and bot messages. This study aims to build a classification model for spam and non-spam comments on TikTok using the Naïve Bayes algorithm and to compare its performance with Support Vector Machine (SVM). The research method refers to the CRISP-DM framework, covering Business Understanding, Data Understanding, Data Preparation, Modeling, Evaluation, and Deployment. The dataset consists of 1,300 TikTok comments evenly divided into 650 spam and 650 non-spam comments. Text preprocessing includes case folding, cleansing, tokenizing, stopword removal, and stemming, followed by TF-IDF feature weighting (max_features=2000, ngram_range=(1,2)). Results show that Naïve Bayes achieved 98.46% accuracy, precision, recall, and F1-score, while SVM achieved a perfect 100% on all metrics. The comparison shows that SVM outperforms Naïve Bayes in detecting spam comments on TikTok. This research is expected to serve as a reference for developing automated, machine-learning-based comment moderation systems on social media.