AI/ML Task 3 - Model Validation and Hyperparameter Tuning

Author: Nice Chetiwal

Project Overview

This project focuses on model validation, overfitting control, cross-validation, and hyperparameter tuning using the California Housing Dataset.

The goal is to compare multiple regression models and select the best final model based on RMSE, R2 score, and cross-validation performance.

Dataset

• Dataset: California Housing Dataset
• Target variable: Median house value
• Problem type: Regression

Models Used

• Linear Regression
• Ridge Regression
• Uncontrolled Decision Tree Regressor
• Tuned Decision Tree Regressor

Key Concepts Covered

• Train-test split
• Overfitting detection
• Cross-validation
• GridSearchCV
• Hyperparameter tuning
• RMSE evaluation
• R2 score evaluation
• Final model selection

Final Model

The final selected model is the Tuned Decision Tree Regressor.

Best parameters:

• max_depth = 12
• min_samples_leaf = 10
• min_samples_split = 2

Performance:

• Test RMSE: 0.6026
• Test R2 Score: 0.7229
• Cross-validation RMSE: 0.6098

Files

• Nice_Chetiwal_AI_ML_Task_3.ipynb - Main Jupyter Notebook
• Nice_Chetiwal_AI_ML_Task_3_Report.pdf - Project report

Conclusion

The tuned Decision Tree performed better than the linear models and the uncontrolled Decision Tree. Hyperparameter tuning helped reduce overfitting and improved model generalization.