Predicting the degree of virality of trending pop songs

By Rayyan Shaik, Esther Amao, Helen Nguyen

Our Goal: Predict how viral a trending song will be based on its lyrics

Our Neural Network's results:

• Accuracy: 68%
• F1 Scores: 48%, 77%, 0% for classes 0, 1 and 2 respectively

Link to project presentation slides

https://docs.google.com/presentation/d/1LOISIGj1PhIdwS4g1bRxUfjqIPT69dUa0159Eut3wM0/edit#slide=id.g2dc43245a29_0_255

About this Respository

This repository contains the code required to scrape the following:

• List of pops songs (names & artists) by year
• Lyrics and lyrics meta-data per song
• Weekly streaming data (global & US) per song

How to use

Required libraries

• selenium
• lyricsgenius
• python-dotenv

Required API Keys

• Genius API Key required
  • https://docs.genius.com/

Running Files

• Simplying running main.py via py main.py or python3 main.py should be sufficient to begin a general scrape

• In main.py you can adjust the years variable range to change the range of years queried (if playlists within the range exist).

• All scraping functions run their own data-cleaning utilities ensure all the data is standardized and usable

Scrape Songs List

• Run the file song_scraper.py
  • Adjust the variable year_range to scrape the desired range of playlists for songs

Scrape Lyrics

• Run the file lyrics_scraper.py
  • config['TOKEN'] needs to be set to your Genius.com API key

Spotify Streams

• Run the file spotify_plays_scraper.py