This document discusses predicting spam videos on social media platforms using machine learning. It proposes using attributes like number of likes, comments, and view count to classify videos as spam or not spam. A predictive algorithm is developed that uses threshold values for attributes and natural language processing of comments to classify videos. Testing of the algorithm on a dataset achieved a spam prediction precision of 93.6%. Issues with small datasets decreasing accuracy are also discussed, along with continuing work to address this issue.