Given the video title, category, publish time, tags, views, likes, dislikes, and comment count, explore the factors that contribute to videos trending.
The YouTube trending videos page provides a platform for small and large content creators to have their videos reach a broader audience. Having your video listed on this page provides higher visibility and potential for channel growth. This ultimately allows the content creator to generate more revenue and strengthen their brand value.
Trending aims to surface videos that:
- Are appealing to a wide range of viewers
- Are not misleading, clickbaity or sensational
- Capture the breadth of what’s happening on YouTube and in the world
- Showcase a diversity of creators
- Ideally, are surprising or novel
- From the YouTube Support Page
The dataset was found on Kaggle , it includes seven months of data on daily trending YouTube vidoes for the US and other countries. The original csv includes video title, channel title, publish time, tags, views, likes and dislikes, description, and comment count. I loaded the data into a SQLite3 database (using sql was a requirement) and normalized with the following schema:
Embedded videos are not supported, here is a link to the video on streamable .
The barchart of the top videos from the final day in the dataset:
Thus far, I've focused mainly on visualization and some EDA. Further work could attempt clustering to search for interesting structures in the data for the videos that become trending.