Design real-time data science project

Frame the loan default problem by defining business objectives, identifying key stakeholders, and translating it into a supervised classification task.

Establish an evaluation framework by selecting cost-sensitive metrics (e.g., considering false positives vs. negatives) and assessing a dataset's suitability for the business problem.

Develop a data preparation strategy for the loan dataset, justifying choices for handling missing values, outliers, and class imbalance.

Engineer and select domain-specific features for credit risk, justifying choices for encoding and scaling in preparation for modeling.

Implement a modeling pipeline that incorporates a technique to address class imbalance (e.g., SMOTE).

Train and compare a baseline model (Logistic Regression) with a complex model (XGBoost) to select the final model based on the pre-defined evaluation metrics.

Perform a comprehensive evaluation of the selected model, including classification threshold tuning based on business cost considerations.

Formulate a high-level deployment and monitoring plan, outlining strategies for detecting model drift and triggering retraining.