This project analyzes over 3 million trending YouTube videos from 113 countries to explore the factors behind video virality. It consists of three main parts:

• Exploratory Data Analysis (EDA):

  • Cleaned and explored a large-scale dataset to understand view counts, engagement metrics, and distribution patterns over time and across countries.

• Supervised Learning (Trend Prediction):

  • Built classification models (Random Forest and Logistic Regression) to predict whether a video will become trending based on features such as view count, like count, comment count, and movement in rankings. Achieved up to 93% accuracy with the Random Forest model.

• Unsupervised Learning (Content Clustering and Topic Modeling):

  • Used KMeans to cluster video titles (TF-IDF) and identify content styles with higher trend potential.

  • Applied LDA topic modeling on video tags to extract popular content themes like comedy, music, and sports, and analyzed which topics are more likely to trend.

  • Combined clustering results with prediction outcomes to validate meaningful content groupings.