The repo for the project of DM ML course
Master: Information Systems - HEC Lausanne - 2019

Video: https://www.youtube.com/watch?v=UdqfCPRODcw&t=180s

Structure of the project

code

• 1. Download Spotify Data (don’t run this notebook)
• 2. Download Lyrics (don’t run this notebook)
• 3. Cleaning Lyrics (don’t run this notebook)
• 4. Exploratory Data Analysis - Spotify Audio Features
• 5. Exploratory Data Analysis - Lyrics
• 6. Classification
• 7. Clustering
• 8. Glove model

data

• top_hits.json
• songs.json
• top_hits_merged_clean_lyrics_audio_features.json
• not_hits_merged_clean_lyrics_audio_features.json

Our data come from Spotify using their API.

• We got the Top Hits from Spotify using their API from 1970 to 2018.
  link: https://open.spotify.com/playlist/3cS7RrbJgQPEiWPAyOnb4l?si=qbUDHQRGTHSgN99Yp6mBlA
• We also collected other songs produced in every decade from 1970 to 2018.

Spotify allows us to collect only information about a song.

More information about the data we collected:
https://developer.spotify.com/documentation/web-api/reference/tracks/get-audio-features/

To get the lyrics of a song, we scraped the follwoing websites:

• genius.com
• lyricsmode.com
• songlyrics.com
• metrolyrics.com

We would like to predict whether a song will be a hit or not, based on lyrics and audio features.

mkrsteska : Marija Krsteska
thil5 : Samuel Lew
Martynas2 : Martynas Savickas
yassinhediger: Yassin Hediger