Skip to content

Spotify Pattern Mining Recommendations

🌐 View Project

Market basket analysis on 1M+ Spotify listening records to surface high-confidence music recommendation rules

Preview

Key Features

  • Built three transaction sets (artist, track, artist+track) from 742K cleaned Spotify records using user-level market basket framing
  • Compared Apriori and ECLAT algorithms across speed, memory, and rule quality — Apriori won both benchmarks
  • Surfaced top rules by lift (e.g. Daft Punk catalog co-listens, genre clusters like hip-hop and pop trios) with actionable recommendation pitch

Technology Used

  • R · arules · arulesViz · dplyr · ggplot2 · plotly · Quarto
  • Apriori & ECLAT with support/confidence/lift tuning
  • Memory profiling via Rprofmem and process timing via proc.time