XAI Stories 3.0
Case studies for retail banking
Preface
Foreword
0.1
Why?
0.2
What?
0.3
How?
0.4
About academic partners
0.5
About business partner
1
Predicting loan applications
1.1
Introduction
1.1.1
The problem
1.1.2
The data
1.2
Preparing the data
1.2.1
Data preprocessing
1.2.2
Forward selection
1.3
Models
1.3.1
Choosing the model
1.3.2
Metrics
1.3.3
Linear models
1.3.4
TabNet
1.3.5
Trees
1.4
Explanations
1.4.1
Variable importance
1.4.2
Partial dependence profiles
1.4.3
Break down
1.4.4
Profile of a typical loan taker
1.5
Bussiness value
1.6
Summary
2
Mincer earnings equation estimation
2.1
Introduction
2.2
Data
2.3
Models
2.3.1
Basic Models Comprehension
2.4
Models Performance
2.4.1
Residual Diagnostics
2.5
Explanations
2.5.1
Local Interpretable Model-agnostic Explanations
2.5.2
Ceteris Paribus Profiles
2.5.3
SHAP
2.5.4
Partial Dependence Profiles
2.5.5
Variable importance
2.6
Explanations validation
2.6.1
Survey description
2.6.2
Survey results
2.7
Summary and conclusions
2.8
Authors contributions
3
In search of factors influencing loan pay off
3.1
Introduction
3.2
Problem
3.3
Data analysis
3.4
Data engineering
3.5
Pipeline
3.6
Explanations
3.6.1
Permutation Feature Importance
3.6.2
SHAP
3.7
Verification of explanations
4
Selecting Clients of Bank that may want to apply for a loan.
4.1
Introduction
4.2
Model
4.2.1
Data:
4.2.2
Preprocessing
4.2.3
Model selection
4.2.4
Selecting features
4.2.5
Using thresholds
4.3
Explanations
4.3.1
Feature Importance
4.3.2
SHAP Beeswarm
4.3.3
PDP
4.3.4
Individual
4.3.5
Conclusions
4.4
Validations
4.4.1
Mentors
4.4.2
Second meeting
4.4.3
We compared them to the report of ‘National Register of Debtors’ (KDR).
4.4.4
Survey
4.5
Summary and conclusions
5
True income prediction based on client expenses
5.1
Introduction
5.2
Dataset
5.2.1
Groups
5.2.2
Preprocessing
5.3
Models
5.3.1
Models description
5.3.2
Business metrics description
5.3.3
Model performance validation methodology
5.3.4
Results
5.4
Explanations
5.4.1
Model level
5.4.2
Instance level
5.5
Usage and user experience
5.5.1
Results
5.6
Summary
6
Why you won’t pay your loan back
6.1
Executive summary
6.2
Introduction
6.3
Data
6.3.1
Exploratory Data Analysis (EDA)
6.3.2
Feature Engineering
6.4
Models and results
6.5
Explanations
6.6
User Study
6.6.1
Construction of the survey
6.6.2
Survey’s result
6.7
Summary
Acknowledgements
References
Published with bookdown
XAI Stories 3.0
References
R Core Team. 2018.
R: A Language and Environment for Statistical Computing
. Vienna, Austria: R Foundation for Statistical Computing.
https://www.R-project.org/
.