Presentation Type
Invited
Track
Finance/Insurance Application
Abstract
Predicting the probability a consumer will not repay their loan is a complicated, but important challenge. In the credit card industry, it is especially important because there is no collateral. Various models can be used to predict this risk of a consumer not paying, but the decision as to which model to use can be a big challenge itself. In general, the types of models range from simple models, like logistic regression, to complex models, like machine learning. There are pros and cons in using either a simple or complex type of model. Typically, complex models perform better in terms of accuracy, but complex models can be hard to explain and even harder to implement. Vigorous research should be done before deciding which type of model to implement, keeping in mind the benefits of using a simpler model with fewer complexities.
Start Date
2-8-2022 9:50 AM
End Date
2-8-2022 10:50 AM
Session 1: Modeling Consumer Risk: A Comparison of Logistic Regression, Scorecard, and Machine Learning Models
Pasque 255
Predicting the probability a consumer will not repay their loan is a complicated, but important challenge. In the credit card industry, it is especially important because there is no collateral. Various models can be used to predict this risk of a consumer not paying, but the decision as to which model to use can be a big challenge itself. In general, the types of models range from simple models, like logistic regression, to complex models, like machine learning. There are pros and cons in using either a simple or complex type of model. Typically, complex models perform better in terms of accuracy, but complex models can be hard to explain and even harder to implement. Vigorous research should be done before deciding which type of model to implement, keeping in mind the benefits of using a simpler model with fewer complexities.