Rejected Inference in Credit Scoring

The video explains rejected inference in credit scoring and how it can be applied [00:00].

Here's a summary of the video:

It covers what rejected inferences are, using the example of Victor and Chris applying for a loan [00:09].
It explains the concepts of bias population, good borrowers (non-events), and bad borrowers (events) [00:32].
It shows how to identify the most predictive group variables and use them to build a score part for training the model [01:27].
The video defines rejected inference as a method to improve the quality of a scorecard using data from rejected loan applications [02:10].
It describes different methods used in rejected inference, including fuzzy methods, hard cutoff method, and parceling [02:28].
The video provides a practical example of the reject inference process using data from Lending Club and the documentation technique in Python [04:36].
It demonstrates training a model with the biased sample of accepted applicants and using it to infer the class of rejected applicants [05:36].
It explains how to combine the data of accepted and rejected applicants to compute a final model [06:00].
The video shows how to predict the label/target variable for the rejected datasets using the model computed on the biased sample [09:20].
It covers evaluating the model's performance on test sets of biased data and then on the whole dataset [10:20].
#RejectedInference
#StatisticalModeling
#PredictiveModeling
#MachineLearning
#DataAnalysis
#ScorecardDevelopment
#ModelTraining
#CreditScoring
#RiskManagement
#Python
#DataSciencePython

Видео Rejected Inference in Credit Scoring канала Prof. Phd. Manoel Gadi