🚗 Car Price Prediction App

This project is a Machine Learning-powered Web App built with Streamlit that predicts the price of a car based on important features such as brand, model, mileage, fuel type, body type, year, and vehicle condition.

The project includes:

• Data preprocessing (cleaning, encoding)
• Model training with hyperparameter tuning (GridSearchCV, 10-fold CV)
• Multiple regressors (Random Forest, XGBoost, Decision Tree, Voting Regressor)
• Automatic selection of the best model
• Model compression for efficient deployment
• Streamlit frontend for interactive user prediction
• Public demo hosted on Streamlit Cloud 🚀

---

📸 Demo Screenshot

🌐 Live Demo (Try it yourself!)

👉 🔗 Click here to open the Streamlit app

✅ You can test different car combinations and see the predicted price instantly!

---

📂 Project Structure

Car_Price_App/
├── car_price_app/
│   ├── generate_dataset.py
│   ├── preprocess_data.py
│   ├── model_training.py
├── data/
│   ├── AutoTrader.csv  # real or synthetic dataset
├── models/
│   ├── *.pkl  # downloaded separately
├── app.py  # Streamlit app
├── download_models.py  # Google Drive downloader for models
├── requirements.txt  # Python dependencies
├── .gitignore
├── README.md

---

🚀 How to Run Locally

Install dependencies:

pip install -r requirements.txt

Download pretrained models:

python download_models.py

Run the Streamlit app:

streamlit run app.py

✅ The app will open automatically at http://localhost:8501

---

🎯 Features Used for Prediction

• Standard Make (e.g., TOYOTA, BMW)
• Standard Model (e.g., COROLLA, 3 SERIES)
• Body Type (e.g., SUV, Sedan, Coupe)
• Fuel Type (e.g., Petrol, Diesel, Electric)
• Vehicle Condition (New or Used)
• Year of Registration
• Mileage
• Standard Colour

---

🔥 Machine Learning Models

Model	Description
Random Forest	Ensemble of decision trees
XGBoost	Gradient boosting decision trees
Decision Tree	Single interpretable tree
Voting Regressor	Combined voting of best models

✅ All models are hyperparameter-tuned with GridSearchCV and 10-fold Cross Validation.

✅ Models are compressed with joblib to optimize memory usage.

---

📈 Evaluation Metrics

Each model is evaluated based on:

• Mean Absolute Error (MAE)
• Root Mean Squared Error (RMSE)
• R² Score

✅ The best model is automatically selected based on R² Score and saved for deployment.

---

📥 Model Download

The trained models are stored separately on Google Drive.

✅ Run the provided script:

python download_models.py

to automatically download and extract models into /models/ before starting the app.

---

⚡ Tech Stack

• Python 3.10
• Pandas
• Scikit-learn
• XGBoost
• Streamlit
• Joblib
• GDown (for model download)

---

✨ Future Enhancements

• Add SHAP explainability to interpret model predictions
• Allow uploading a CSV of multiple cars for batch prediction
• Add model monitoring and retraining
• Deploy on a faster and better host like Snowflakes for user ease

---

👨‍💻 Author

Bolaji Adeyemi
Passionate about AI, ML, and Building Scalable Intelligent Applications.

Connect with me on LinkedIn