🏡 Immo-Eliza — Full Deployment Project

A complete ML → API → Web App pipeline for real-estate price prediction

This project contains the end-to-end workflow for deploying a machine-learning model that predicts Belgian real-estate prices. It integrates:

1. Data cleaning & preprocessing
2. Model training
3. API deployment (FastAPI + Docker, hosted on Northflank)
4. Web application UI (Streamlit Cloud)
5. A fully automated pipeline for real-world estimation

---

📁 Repository Structure (Main Project)

IMMO-ELIZA-DEPLOYMENT/

├── best_price_model.joblib
├── clean_data.csv
├── data_preprocessing.py
├── Dockerfile
├── model_features.joblib
├── my_api.py # API code for FastAPI (development version)
├── price_scaler.joblib
├── README.md # ← you are here
├── requirements.txt
└── UI.py # Streamlit UI (development version)

This repository is the canonical source for:

• the training pipeline
• the preprocessing logic
• the inference artifacts (model + scaler + feature list)
• documentation about the full ML → API → UI workflow

Two additional repositories depend on it.

---

🚀 Project Components

1️⃣ Machine Learning Model (This Repository)

This is where the core model is built:

• Load & clean the real-estate dataset
• Encode categorical features + enrich with postal/province/region
• Normalize the target (price)
• Train multiple regression models
• Export:

best_price_model.joblib

model_features.joblib

price_scaler.joblib

This ensures that both the API and UI use the same preprocessing and feature space.

---

2️⃣ Prediction API (FastAPI + Docker + Northflank)

📂 Repository: IMMO-ELIZA-DEPLOYMENT-API
🌐 Live URL:
https://p01--immo-eliza-api--j9cqmtc6d89m.code.run

This repository contains:

IMMO-ELIZA-DEPLOYMENT-API/
├── best_price_model.joblib
├── clean_data.csv
├── data_preprocessing.py
├── Dockerfile
├── model_features.joblib
├── my_api.py # FastAPI application
├── price_scaler.joblib
└── requirements.txt

The API exposes a /predict endpoint that:

1. Accepts a JSON payload describing a property
2. Applies preprocessing
3. Runs the ML model
4. Inverse-transforms the predicted price
5. Returns the estimate to the UI

This API is deployed on Northflank and runs inside a Docker container.

---

3️⃣ Web User Interface (Streamlit Web App)

📂 Repository: IMMO-ELIZA-DEPLOYMENT-UI
🌐 Streamlit URL:
https://immo-eliza-ui.streamlit.app

This repository contains:

IMMO-ELIZA-DEPLOYMENT-UI/
├── best_price_model.joblib
├── clean_data.csv
├── data_preprocessing.py
├── model_features.joblib
├── price_scaler.joblib
├── requirements.txt
└── UI.py # full multi-step interface

The UI allows users to:

• enter property information step-by-step
• dynamically update subtypes based on category
• validate postal codes
• call the backend API
• display the final predicted price

---

🔗 How Everything Fits Together

[ Training Repo ]
↓ exports
best_price_model.joblib
model_features.joblib
price_scaler.joblib
↓ copied into
──────────────────────────────────
[ API Repo (Northflank) ]
|
| JSON request
↓
/predict endpoint
|
↓ JSON response
──────────────────────────────────
[ UI Repo (Streamlit Cloud) ]

The training repo is where the models get trained and compared.
The API repo uses the trained artifacts to expose predictions.
The UI repo calls the API and presents results to the user.

---

🛠 Local Development

1. Create a virtual environment

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

2. Run the API locally

uvicorn my_api:app --reload

3. Run the Streamlit UI locally

streamlit run UI.py