YouTube Video Metadata

This dataset contains metadata collected from publicly accessible YouTube videos using the YouTube Data API v3
for research in trend prediction, engagement analysis, machine learning, and natural language processing applications.
The dataset consists of 3951 YouTube video records collected across multiple content categories including music, gaming, sports, education, technology, finance, podcasts, comedy, news, and entertainment. Each record contains structured metadata attributes such as video title, description, channel information, publication timestamp, duration, tags count, view count, like count, comment count, and thumbnail information.
In addition to raw metadata, several engineered features were generated to support predictive analytics and NLP-based research. These include title length, description length, word count, uppercase ratio, punctuation indicators, sentiment polarity scores, temporal features (year, month, weekday, hour), and engagement metrics such as like-to-view ratio and comment-to-view ratio.
The dataset was preprocessed to remove duplicate entries and handle missing values. Textual attributes were cleaned using regular expression-based preprocessing techniques. The final dataset is stored in CSV format and is suitable for machine learning workflows, trend prediction studies, recommendation systems, social media analytics, engagement forecasting, and benchmarking experiments.