Skip to content

Latest commit

 

History

History
26 lines (19 loc) · 2.35 KB

File metadata and controls

26 lines (19 loc) · 2.35 KB

Youtube-Trending-Scraper-and-EDA

The script is a modified version of this Youtube's trending video scraper. This script scrapes the most relevant information from videos for the selected countries specified in 'country_codes.txt' file. The example output files can be found in the output directory.

Country Codes

The script used 12 country codes for which it collected trending video data. These are 2 letter country abbreviations (LT, LV, EE, PL, FI, SE, NO, US, CA, GB, DE, FR), however the following EDA was done only on the Baltic countries (LT, LV, EE).

Scraping results

The script collected data between 2022-07-22 and 2022-11-03, and contains data for a total of 76 unique dates.

Data cleaning and EDA

After the data was scraped, it was then combined into a single dataset using the script in 'combine csv data.ipynb'. Additionally, as every country has different video category codes (i.e. 2 - Autos & Vehicles, 10 - Music, 17 - Sports, etc), categories for LT, LV and EE were additionally scraped and added to the dataset.

The Exploratory Data Analysis can be found in 'data cleaning and EDA.ipynb' file.

EDA conclusions

  • 40% of trending videos have less than 1 million views, and 66% have less than 2 million views
  • Some videos were trending for mulpitle days and therefore appeared in the dataset more than once. From the total of 11310 entries, 1861 are unique videos.
  • On average, videos are commented 0.3% of the time i.e. around every 307 views
  • Most of the videos trend for less than 5 days and similarly, they usually become trending after <= 5 days after being published. This suggests that Youtube's top trending list is fast-changing and newly published videos are more popular than older videos
  • Trending video titles vary a lot in lenght - although usually their lenght is between 20 and 100 charachter long
  • Various articles and symbols from the Cyrillic script like | , в, -, и, were common in trending video titles. Excluding these symbols, the words 'Alexey', 'Arestovych' and 'Game' were the most common words in the trending video titles
  • The category that has the largest number of trending videos is 'News & Politics' with 395 videos, followed by 'Entertainment' with 388 videos.

License

This project is licensed under the BSD 2-Clause License - see the LICENSE file for details