This project explores one of the world’s most popular streaming platforms—Spotify—using a structured database built from scratch. With over 10,000 records on the most-streamed songs, we dove deep into the data to clean, model, and query meaningful musical trends.
We designed a relational database using our custom ER diagram featuring:
Track: Core song info (name, cover, artist count)Artist: All unique artist namesTrackProfile: Musical features like BPM, energy, speechiness, etc.StreamingInfo: Cross-platform performance metricsReleased_By: A junction table linking artists and tracks
Built in PostgreSQL, and populated with Python data cleaning & preprocessing code.
We cleaned messy data and handled:
- Missing values (replaced or randomly filled when appropriate)
- Strange Unicode characters
- Duplicate rows (via string normalization + sorting by streams)
- Inconsistent numerical values (e.g., "in_deezer_playlists" with commas)
The result? A clean and query-ready dataset.
We wrote several exploratory SQL queries to uncover interesting trends:
- Query 1: Tracks how average BPM and danceability changed over the years
- Query 4: Analyzes how tempo (BPM) affects popularity based on streams
- Query 5: Examines whether high-liveliness tracks also have album cover art, linking visual and sonic appeal
These queries tell stories about the intersection of music, data, and culture.
Nikol Tushaj
Rajla Culli
Chloe Monique Quevedo
Lina Kolevska