Spam Detection in YouTube Comments: A Machine Learning Approach
Sarita Yadav, Junali Jasmine Jena, Jay Prakash Singh, Mahendra Kumar Gourisaria, Sonal Kumar Jain, Vivek Kumar · 2025
Due to the broad spectrum of content and evolving tactics of spammers, identifying spam in YouTube comments has become a challenging task. If it goes undetected it may hamper the security and integrity of the application. Automated spam detection techniques have found profound application in this field. This study offers an extensive approach for enhancing spam classification through the integration of cutting-edge machine learning algorithms. A multi-pronged paradigm is investigated, including SMOTE to address class imbalance, and TF-IDF vectorization for feature extraction and used for spam classification. A wide range of models is scrutinized in the study such as Gradient Boosting, AdaBoost, XGBoost, etc. The best-performing models, based on cross-validation scores, were the Support Vector Classifier (SVC) with an average accuracy of 96.42%, followed by Gradient Boosting, Voting Classifier and LightGBM, achieving an average accuracy of 94.63%, followed by Logistic Regression and Random Forest achieving an accuracy of 94.37%.