Visualization for Explainable AI
Teamoverleg — 18 oktober
Dennis Collaris
PhD Visualization
Data
Black
Box
Subject
80% risk
Why?
ML is often applied as a black box.
Global
Local
Overview
Global
Local
Fraud Dashboard
Overview
ExplainExplore
Global
Local
Fraud Dashboard
Overview
ExplainExplore
Contribution-Value Plots
Global
Local
Fraud Dashboard
Overview
ExplainExplore
CV Plots
Global
Local
Fraud Dashboard
StrategyAtlas
Overview
How?
Strategy A
Strategy B
The basic principle
How?
The basic principle
| ID | Name | Age | Sex | Product | Branch | ... |
|---|---|---|---|---|---|---|
| 1 | 💤 | 💤 | 💤 | 🔥 | 💤 | ... |
| 2 | 🔥 | 💤 | 💤 | 💤 | 🔥 | ... |
| 3 | 💤 | 🔥 | 💤 | 🔥 | 🔥 | ... |
| ... | ... | ... | ... | ... | ... | ... |
| ID | Name | Age | Sex | Product | Branch | ... |
|---|---|---|---|---|---|---|
| 1 | Alice | 28 | F | Health | Zekur | ... |
| 2 | Bob | 57 | M | Car | FBTO | ... |
| 3 | Chad | 34 | M | Life | Intrpls | ... |
| ... | ... | ... | ... | ... | ... | ... |
2D projection
StrategyMap
feature contribution (LIME)
What?
Method 1: Heat map cluster analysis
Data
Model
What?
Method 2: Density plots
Data
Model
All data
Selection
What?
Method 3: Decision trees
Saved clusters →
DT for selected cluster →
Performance comparison →
What?
Demo of the system
Paper submitted to:
IEEE Transactions on Visualization and Computer Graphics
sklearn-pmml-model
bit.ly/
sklearn-pmml
or
Other projects
Importing models from R to Python
Other projects
Fundamental research on expert interpretation of feature importance
sklearn-pmml-model
bit.ly/
sklearn-pmml
or
explaining.ml
or
Links
Where you can find more information