college-tuition-prediction

This project predicts tuition costs for 600+ U.S. colleges using engineered institutional and geographic features.

Highlights

• Dataset: 600+ U.S. college records
• Models: Ridge Regression, Gradient Boosting Regressor, Neural Network (PyTorch)
• Evaluation: 7-fold cross-validation
• Team rank: 2nd out of 13 teams

Performance

• Best RMSE: 5,860.88
• R2 score: 0.799
• Baseline RMSE: 13,086.66
• Improvement over baseline: 55%

Learnings

• Engineered features (region and institution type) were high impact.
• Ridge regression helped regularize high-dimensional feature sets.
• Gradient boosting outperformed linear baselines by capturing nonlinear tuition patterns.