End-2-End Application of Machine Learning Models for Credit Acceptance Models
The slides shown at the event can be found here.
At the event Artur presented the entire development process of machine learning models for credit acceptance models, with a focus on methods, some of the topics were:
- Risk driver engineering from big data
- Risk driver selection using traditional and ML methods Examples
- Using resemblance modelling to spot data drifts
- Temporal cross validation for PSI calculation
- ML model training & tuning
- Controlling overfit
- Using Bayesian methods for faster tuning
- Monitoring & Deployment
Artur Usov is a principal data scientist in Retail Banking Analytics Tribe, with +10 years experience of working with data & analytics in various sectors, of which 6+ years are in banking. Currently focusing on enabling Instant Lending within ING by means of Machine Leaning & Analytics. Artur has academic background is in Statistics and Economics.