9. backstage (live)
Calls setlist.fm api
Gets historical data on the
queried band
Create a panel of songs based
on the data
Song Prediction
10. backstage
Calls setlist.fm api
Gets historical data on the
queried band
Machine Learning
Create a panel of songs based
on the data
Predictions
Song Prediction
11. real-time backstage
Calls setlist.fm api
Gets historical data on the
queried band
Machine Learning
Calls last.fm api
Scrapes json to get user
listening history
Calculate User’s Top Songs
Create a panel of songs based
on the data
Predictions
Song Prediction Personalization
12. backstage
Calls setlist.fm api
Gets historical data on the
queried band
Machine Learning
Calls last.fm api
Scrapes json to get user
listening history
Calculate User’s Top Songs
Create a panel of songs based
on the data
Predictions
Song Prediction Personalization
Actionable Insight
Match user’s top songs to
predictions
13. algorithm
• A random-forests model is trained at the band level
• Observations based on the last 50,000 songs the band has played
• 7 categories one hot-encoded as ~700 features (current Iron
Maiden model: 767)
• song names, event id, venues, tours, cities, countries, new
song
Accuracy 0.92
Precision 0.70
Recall 0.54
ROC Score 0.75
14. about Doğa
• Economics and Sociology B.A, 2008, McGill
University, Montreal, QC, Canada
• Sociology PhD (expected) Fall 2016, University of
Pennsylvania Sociology Department
• Research in economic and political sociology:
• Diffusion of financial and sociopolitical-crises
across international borders
• Struggles between secularist and religious
authorities during state-formation of Turkey,
France, and Mexico