1,177
#1 hits in training set
85%
Precision on weeks at #1
100+
Saved scenarios
Is chart longevity predictable from audio attributes? I trained a model on real Billboard Hot 100 history and wrapped it in an interface anyone can explore.
The What-If Studio is the core experience: adjust danceability or energy and watch predicted weeks at #1 shift, alongside the five most similar historical hits.
- Next.js
- React
- Tailwind CSS
- Python
- scikit-learn
- MongoDB
Predicts how long a track could hold the Billboard #1 spot, with a What-If Studio to tune audio features and see why.
Under the hood
- 01
Predictive ML Model
Random Forest on 1,177 Billboard #1 hits. Feature importance surfaces the top three driving attributes.
- 02
Similarity & Calibration Engine
k-NN matches tracks to five similar historical hits. Calibrated probabilities estimate weeks at #1 with 85% precision.
- 03
What-If Studio & Dashboard
Spotify-inspired dark UI, real-time parameter tuning, decadal chart analytics, JWT auth, MongoDB persistence for 100+ saved scenarios.
Rapids QR
Offline Link & Contact Generator
Open case