🏠 Belgian Immo Price Prediction App

This is a machine learning web application that predicts the price of real estate in Belgium based on property features such as type, location, surface area, and energy efficiency.

🚀 Technologies Used

• Python
• scikit-learn (for model training)
• Streamlit (for the web app)
• pandas, numpy, Pillow (for data manipulation and display)

📁 Project Structure

.
├── kangaroo_clean.csv         # Cleaned dataset used for training
├── pred_model.ipynb.ipynb     # Jupyter Notebook for training the model
├── streamlit_app.py           # Streamlit app for predictions
├── kangaroo_clean.sav         # Trained model (generated from notebook)
└── ImmoElizaDeploy.png        # Image used in the app UI

🔧 How It Works

1. Model Training

• Load kangaroo_clean.csv
• Encode and scale features
• Train a regression model (e.g. RandomForestRegressor)
• Export model using pickle as kangaroo_clean.sav

2. Web App (Streamlit)

• Load the trained model and dataset
• Display dropdowns and sliders for user input
• Dynamically filter localities based on province
• Format user input and predict property price
• Display the predicted price

✅ Features Used

• type_subtype
• province
• locality_normalized
• kitchenType
• bedroomCount
• bathroomCount
• habitableSurface
• gardenSurface
• epcScoreMapping
• facedeCount
• roomCount
• version

🛠 How to Run

1. Install dependencies:

pip install -r requirements.txt

2. Make sure all files are in the same folder.

3. Run the Streamlit app:

streamlit run streamlit_app.py

📌 Notes

• Ensure column names in user input match those expected by the model (e.g., epcScoreMapping).
• The dataset represents real Belgian housing data and can be updated with more recent listings.