Scribd
Upload
5 views1 page

Predictive Model for Credit Delinquency

The predictive model plan outlines a logistic regression approach to forecast credit delinquency, emphasizing preprocessing, feature selection, model training, thresholding, and evaluation metrics. Logistic regression is chosen for its simplicity and interpretability, crucial for regulatory compliance and stakeholder trust. The evaluation strategy includes precision, recall, F1 score, AUC-ROC, and fairness audits to ensure ethical considerations are met.

Uploaded by

faseehazulaika
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
5 views1 page

Predictive Model for Credit Delinquency

The predictive model plan outlines a logistic regression approach to forecast credit delinquency, emphasizing preprocessing, feature selection, model training, thresholding, and evaluation metrics. Logistic regression is chosen for its simplicity and interpretability, crucial for regulatory compliance and stakeholder trust. The evaluation strategy includes precision, recall, F1 score, AUC-ROC, and fairness audits to ensure ethical considerations are met.

Uploaded by

faseehazulaika
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd

Predictive Model Plan – Student Template

1. Model Logic (Generated with GenAI)

Prompt used:
"Outline a predictive modeling pipeline to forecast credit delinquency."

Output (paraphrased):
The model uses logistic regression to predict the likelihood of a customer becoming delinquent. Key
steps include:

 Preprocessing: Median/mode imputation, capping outliers, encoding categories.

 Feature selection: Based on EDA insights (e.g., missed payments, DTI).

 Model training: Logistic regression to estimate delinquency probability.

 Thresholding: Classify customers as high-risk if probability > 0.5.

 Evaluation: Use metrics like precision, recall, and AUC.

2. Justification for Model Choice

Logistic regression was chosen for its simplicity, speed, and interpretability—crucial for regulatory
compliance and stakeholder trust in financial services. It outputs clear probabilities and coefficients,
helping the Collections team understand why a customer was flagged. It also performs well with
structured data and aligns with Geldium’s need for transparent, ethical AI.

3. Evaluation Strategy

Metrics:

 Precision & Recall: To balance identifying true delinquents and avoiding false alarms.

 F1 Score: For overall model effectiveness on imbalanced data.

 AUC-ROC: To evaluate how well the model ranks risk.

Fairness:

 Use SHAP for explainability.

 Check group-wise performance (e.g., by income, location).

 Apply fairness audits (e.g., disparate impact analysis) and exclude biased proxies.

Ethical consideration:
Ensure the model doesn’t unintentionally penalize certain demographics and supports responsible,
explainable intervention.

You might also like