ayush1997/YouTube-Like-predictor

Trains a Random Forest regressor on ~350,000 YouTube videos with derived features from video metadata and channel attributes, achieving predictions via scikit-learn with grid search hyperparameter tuning. Data pipeline leverages the YouTube API to extract video IDs, attributes, and channel statistics across categories, followed by feature engineering and cleaning in Jupyter notebooks to produce the final training dataset. Supports batch predictions on up to 40 video IDs at a time, with a pretrained model available to avoid the ~18-minute training overhead.

147 stars. No commits in the last 6 months.