Back to projects

Project field note

Spotify Billboard Analytics

A project pairing Spotify audio features with Billboard Hot 100 chart data, backed by PostgreSQL 16 on AWS RDS with a Node/Express API, a React + Recharts frontend, and a Python/pandas cleaning pipeline.

Interactive artifact

Audio twin finder

Pick a song and the demo finds its closest "audio twins" by measuring distance across five audio features, the same k-nearest-neighbor idea the app runs across 447k tracks with a GiST-indexed cube. Illustrative 20-track sample; feature values are approximate.

Audio fingerprint

Closest twins k-NN · 5-D

    A music analytics application built as a database systems course project, joining Spotify audio features with Billboard Hot 100 chart history to explore how a song's characteristics relate to its chart performance.

    The backend runs PostgreSQL 16 on AWS RDS with a Node/Express API, and a Python and pandas pipeline cleans and loads the raw data. The schema normalizes four core tables to 3NF and uses GiST/GIN indexes, pg_trgm for fuzzy text search, materialized views for heavy aggregates, and a k-NN similarity query for finding songs with similar audio profiles.

    The frontend is built in React with Recharts. I owned the Trends and Analytics visualizations, turning the query results into interactive charts.

    As a graded course assignment, the source code cannot be shared publicly under course policy.