Deep generative models for reject inference in credit scoring

@article{Mancisidor2020DeepGM,
  title={Deep generative models for reject inference in credit scoring},
  author={R. A. Mancisidor and Michael C. Kampffmeyer and K. Aas and R. Jenssen},
  journal={Knowl. Based Syst.},
  year={2020},
  volume={196},
  pages={105758}
}
Credit scoring models based on accepted applications may be biased and their consequences can have a statistical and economic impact. Reject inference is the process of attempting to infer the creditworthiness status of the rejected applications. In this research, we use deep generative models to develop two new semi-supervised Bayesian models for reject inference in credit scoring, in which we model the data generating process to be dependent on a Gaussian mixture. The goal is to improve the… Expand
Deep learning for credit scoring: Do or don't?
TLDR
Deep learning algorithms do not seem to be appropriate models for credit scoring based on this comparison and XGBoost should be preferred over the other credit scoring methods considered here when classification performance is the main objective of credit scoring activities. Expand
Three-stage reject inference learning framework for credit scoring using unsupervised transfer learning and three-way decision theory
TLDR
A novel three-stage reject inference learning framework using unsupervised transfer learning and three-way decision theory that integrates rejected credit sample selection using three- way decision theory, higher-level representations to transfer learning from both accepted and selected rejected credit samples, and credit scoring using the reconstructed accepted credit samples is proposed. Expand
An Explanation Framework for Interpretable Credit Scoring
TLDR
This work presents a credit scoring model that is both accurate and interpretable and enhanced with a 360-degree explanation framework, which provides different explanations required by different people in different situations. Expand
Explainable AI for Interpretable Credit Scoring
TLDR
A credit scoring model that is both accurate and interpretable and able to satisfy the six predetermined hypotheses testing for correctness, effectiveness, easy understanding, detail sufficiency and trustworthiness is presented. Expand
Differential Replication for Credit Scoring in Regulated Environments
TLDR
It is shown how copying can be used to replicate the decision behavior not only of a model, but also of a full pipeline, and can ensure the decomposability of the attributes used to provide explanations for credit scoring models and reduce the time-to-market delivery of these solutions. Expand
A machine learning approach combining expert knowledge with genetic algorithms in feature selection for credit risk assessment
TLDR
The ability of interpretation of the predictive power of each feature in the credit dataset is strengthened by the engagement of experts in theCredit scoring process, and the proposed wrapper-based feature selection approach which explores how the features contributing most towards the classification of borrowers are explored. Expand
Forecasting Probability of Default for Consumer Loan Management with Gaussian Mixture Models
TLDR
A new method based on Gaussian Mixture Model (GMM) to forecast the probability of default for individual loan applicants, which gives a probabilistic view of credit standing for each individual applicant instead of a binary classification and therefore provides more information for financial decision makers. Expand
Credit Rating Based on Hybrid Sampling and Dynamic Ensemble
TLDR
Experiments on three credit data sets prove that the combination of hybrid sampling and dynamic ensemble can effectively improve the performance of the classification. Expand
Characterizing Fairness Over the Set of Good Models Under Selective Labels
TLDR
A framework for characterizing predictive fairness properties over the set of models that deliver similar overall performance, or “the set of good models,” is developed, which addresses the empirically relevant challenge of selectively labelled data in the setting where the selection decision and outcome are unconfounded given the observed data features. Expand
Credit Scoring for Micro-Loans
TLDR
This paper showcases how micro-loan credit system can be developed in real setting and shows what challenges arise, and introduces semi-supervised algorithm that aids model development and evaluates its performance. Expand
...
1
2
...

References

SHOWING 1-10 OF 71 REFERENCES
Bound and collapse Bayesian reject inference for credit scoring
TLDR
Results show that the proposed flexible method to generate the probability of missingness within a model-based bound and collapse Bayesian technique improves the classification power of credit scoring models under MNAR conditions. Expand
Reject inference in application scorecards: evidence from France
Credit scoring models are commonly developed using only accepted Known Good/Bad (G/B) applications, called KGB model, because we only know the performance of those accepted in the past. Obviously,Expand
Credit scoring and reject inference with mixture models
  • A. Feelders
  • Computer Science
  • Intell. Syst. Account. Finance Manag.
  • 2000
TLDR
This work proposes a new reject inference method based on mixture modeling, that allows the meaningful inclusion of the rejects in the estimation process, and describes how such a model can be estimated using the EM-algorithm. Expand
Credit scoring, augmentation and lean models
TLDR
The extent to which, given the previous cutoff score applied to decide on accepted applicants, the number of included variables influences the efficacy of a commonly used reject inference technique, reweighting is explored. Expand
Modified logistic regression using the EM algorithm for reject inference
TLDR
This study provides a novel reject inference technique in which the rejected applicants are included in the model estimation process and the extrapolation problem is avoided using the methodology. Expand
A Semi-supervised Approach for Reject Inference in Credit Scoring Using SVMs
TLDR
A novel semi-supervised approach that determines a linear predictor using Support Vector Machines (SVMs) and incorporates information on rejected loans, assuming that the labeled data and unlabeled data are not drawn from the same distribution. Expand
Reject inference in survival analysis by augmentation
TLDR
The conclusion is essentially that augmentation achieves negative benefits only and that the scope for reject inference in this context pertains mainly to circumstances where a high proportion of applicants have been rejected. Expand
Reject inference in credit operations based on survival analysis
TLDR
The proposed method has an advantage of predicting the time to delayed repayment for an applicant with associated characteristics so that the proper loan duration can be set. Expand
An econometric analysis of the bank credit scoring problem
Abstract Most credit assessment models used in practice are based on simple credit scoring functions estimated by discriminant analysis. These functions are designed to distinguish whether or notExpand
Sample selection in credit-scoring models1
Abstract We examine three models for sample selection that are relevant for modeling credit scoring by commercial banks. A binary choice model is used to examine the decision of whether or not toExpand
...
1
2
3
4
5
...