Genre Intelligence Engine
Python
SQL
R
JavaScript
TypeScript
Problem
Music and event platforms often treat genres as static labels, which limits discovery and recommendation quality. Built a data mining system that learns real relationships between genres from artist and event behavior.
Approach
- Built a two-stage pipeline: preprocessing and mining.
- Cleaned and normalized genre signals from large JamBase-style artist and event datasets.
- Enriched missing metadata, built genre co-occurrence maps, and generated taxonomy paths.
- Trained embeddings and clustering outputs to support user preference scoring and feed ranking.
Outputs
- Genre taxonomy paths and cluster keys for explainable grouping.
- Event clusters and artist embedding clusters for recommendation candidate generation.
- User preference and affinity outputs for personalization.
- A ranked personalized feed proof of concept.