Mazen Asag

Diamond Price Prediction 💎📈

This project builds a regression model to accurately predict diamond prices based on various features such as carat, cut, color, and clarity.

📌 Problem Statement

Predict the price of diamonds using multiple features for improved valuation and pricing decisions.

📊 Data Collection

Dataset: Kaggle - Playground Series Season 3 Episode 8

🧬 Features

• carat: Weight of the diamond
• cut: Quality of the diamond cut
• color: Diamond color grade (D–J)
• clarity: Purity level based on internal flaws
• depth, table, x, y, z: Physical dimensions
• price: Target variable

📈 Model Performance Comparison

Model	MSE	RMSE	MAE	Train R²	Test R²	Adj. R²
CatBoost Regressor	3.36e+05	579.96	295.48	0.9827	0.9792	0.9792
XGBRegressor	3.42e+05	585.44	296.96	0.9838	0.9788	0.9788
Random Forest	3.71e+05	609.35	310.42	0.9968	0.9770	0.9770
KNN Regressor	4.49e+05	670.78	350.55	0.9815	0.9722	0.9721
Decision Tree	7.00e+05	837.25	422.66	1.0000	0.9566	0.9566
Linear Regression	1.01e+06	1006.60	671.58	0.9366	0.9373	0.9373
Ridge	1.01e+06	1006.60	671.61	0.9366	0.9373	0.9373
Lasso	1.01e+06	1006.87	672.99	0.9366	0.9373	0.9372
AdaBoost	1.94e+06	1393.41	971.55	0.8827	0.8798	0.8798

🏆 Best Model: CatBoost Regressor

Achieved highest R² and lowest error metrics across all models.

🔍 Insights & Findings

• Tree-based models outperform linear models significantly.
• CatBoost and XGBoost showed the best test-time performance.
• Decision Tree overfitted; Lasso/Ridge failed to capture nonlinearities.

📓 Notebooks

• EDA Notebook
• Model Training
• Model Explainability with LIME

🛠️ How to Run

git clone https://github.com/Mazenasag/Gemstone_Price_Prediction_.git
cd diamond-price-prediction
# Run the Flask app
python app.py
        

Then open your browser and visit: http://127.0.0.1:5000/

✍️ Author

Developed by Mazen Asag