Visa Approval Prediction: End-to-End Project

Problem Statement

Predicting visa approval outcomes from applicant data is a classification problem where model performance needs ongoing monitoring once deployed, since real-world data drifts over time.

Solution

Built and compared KNN, Random Forest, and XGBoost models, tuned via grid search and cross-validation, reaching 96% accuracy with a strong ROC-AUC score.

Architecture

A Scikit-learn training pipeline feeds the best-performing model into a monitoring layer that logs predictions and tracks data drift in production, backed by MongoDB.

Results

• 96% accuracy with a strong ROC-AUC score after grid search and cross-validation across KNN, Random Forest, and XGBoost
• Model monitoring and logging in place to track prediction performance and data drift