There has been a noticeable surge in spam and illegitimate comments with the booming growth of digital footprints on social media platforms leading to a great inconvenience to moderators and content administrators. Here, we present a comprehensive study of YouTube comment classification using machine learning models. Our dataset comprises 50,744 comments extracted from 21 Music Videos. With selective feature vectors and manually curated datasets, we use SVM, Logistic Regression, Random Forest, and KNN Classifier models and Hybrid metric to assess each model performance. The intention of this research is to provide a view of large-scale data with intuitive insights and knowledge-based analysis when selective features are used.