Exploratory Data Analysis - Youtube Trending Videos

by Peter Hontaru

Introduction

Problem statement:

Are there differences between YouTube Trending videos in
English-speaking countries (Canada, Great Britain, United States)? If
yes, which aspects of the videos cause these differences?

All throughout this analysis we will generate questions, answer them
with data and then further refine these questions based on what was
discovered.

Who is this project intended for?

• Content Creators / Marketing Agencies who want to better understand
  their audience and tailor their content effectively
• Data Science enthusiasts, as they might get new ideas for how to use
  R (for beginners) or see other people’s analyses (for intermediates)
• those generally interested in YouTube

Why this dataset?

YouTube has been an influential source of knowledge both in my
professional life as well in my personal goals/ambitions. Thus, I
thought it would be interesting to explore this dataset and gather
insights in order to better understand the platform.

Key Insights:

• there were medium to high correlations between our main variables

• despite a similar number of total trending videos, Canada has more
  unique videos (60%) versus Great Britain (8%) and The United
  States of America (16%)

• most videos in CA only trend for 1-2 days (80%) versus GB and US
  where it isn’t uncommon to trend for up to 10 days (highest 38
  days), showing a need for consistent innovation

• generally, 75% of videos trend within their first day and 95%
  of videos trend within the first 5 days. In CA, videos usually
  take 1-2 days to reach trending, while in the US and UK it could
  take up to 5-7 days
• only 25% of all videos receive over 1 million views and 5%
  receive over 5 million

• Music and Entertainment are the most common trending
  categories
• News & Politics is the most controversial category (as judged by
  % of dislikes), while Pets & Animals and Comedy are the
  least disliked

What are Trending Videos and why are they important?

Trending videos work alongside the home page to provide users with
content to watch. While the home page is highly personalised (via
the YouTube algorithm) based on previous views, what the user watched
longest, engagement, subscriptions, the trending page is very broad
and identical across all accounts. Since it shows this feed to
millions of users, it serves as a great source of views for content
creators (think viral videos).

Where did the data come from?:

• contains >120,000 videos across three countries (Canada, Great
  Britain, United States of America)
• 8 months of Daily Trending data between 2017-11-14 and
  2018-06-14 (approx 200 videos/day/country)
• all the data is downloaded from
  https://www.kaggle.com/datasnaek/youtube-new - Raw data files are
  available within the “raw data” folder

3. Extended analysis:

Full analysis available:

• at the following
  link,
  in HTML format (recommended)
• in the Exploratory-Data-Analysis.md of this repo (however, I
  recommend previewing it at the above link since it was originally
  designed as a html document)