RATE Analytics
26 juni
Dennis Collaris
PhD Visualization
Fraud detection explanations
sick-leave insurances
PREVIOUS WORK
My solution
PREVIOUS WORK
Decision support
Diagnostics
Refinement
Justification
Applications
PREVIOUS WORK
Paper presented at:
Workshop on Human Interpretability in Machine Learning
Stockholm, Sweden
PREVIOUS WORK
DEBTOR MANAGEMENT
Effectiveness of debt collection strategies
Decision support
Diagnostics
DEBTOR MANAGEMENT
Design decisions
DEBTOR MANAGEMENT
Design decisions
Global
Local
DEBTOR MANAGEMENT
Design decisions
Surrogate learning
0 1 2 3 x
y
2
1
Feature 1
Feature 2
Feature 3
Feature 1
Feature 2
Feature 3
Feature 1
Feature 2
Feature 3
DEBTOR MANAGEMENT
Help data scientists to create and tune explanatory surrogate models.
DEBTOR MANAGEMENT
Configuration view
-
← Any tabular data set
-
← Any Python classifier, or PMML
-
← Different surrogate models
DEBTOR MANAGEMENT
Global overview
R²
Every line = an explanation
More information than traditional feature importance
Selecting subgroups
R²
DEBTOR MANAGEMENT
Local explanation view
← quality
← explanation
DEBTOR MANAGEMENT
Context view
DEBTOR MANAGEMENT
Context view
DEBTOR MANAGEMENT
Context view
Demo
Paper submitted to:
IEEE Visual Analytics Systems and Technology
Vancouver, Canada
Applicable to any machine learning model
LIME
OTHER ACTIVITIES
LIME
0 1 2 3 x
y
2
1
-2 -1 0 1 2 d
OTHER ACTIVITIES
0 1 2 3 x
y
2
1
0 1 2 3 x
y
2
1
LEMON
LIME
LEMON
OTHER ACTIVITIES
LIME / LEMON
Can be used for any Python model...
sklearn-pmml-model
OTHER ACTIVITIES
Can be used for any model...
Events
OTHER ACTIVITIES
NEXT STEPS
Opportunities
-
Achmea
-
Overlijdens Risico Verzekeringen (ORV) (Senna van Iersel)
-
Health insurance churn prediction (Lizan Kops)
-
Pricing GLMs (Joost van Bruggen)
-
Recruitment analytics (Silke Lhoëst)
-
Team Data Science (Schade Particulier) (Wouter Slot)
-
-
Collaboration
-
RATE colleagues
-
Lorentz Grant Workshop @ Leiden
-
IBM & Hardvard @ Boston
-
NEXT STEPS
Topic
Global
Local
NWO RATE meeting 26 june
By iamdecode
