🏠 Century21 ML Estimation Project

🏠 Machine Learning for Real Estate Price Estimation

Vision: Combine machine learning and real estate expertise to provide transparent, reliable and explainable property price estimations.

This project was developed as part of the Machine Learning Engineer program at Liora (ex-DataScientest) and explores how data science can improve the real estate valuation process.

🎯 Context & Challenges

The real estate market is becoming increasingly complex due to:

• rising interest rates affecting purchasing power
• strong territorial price disparities
• increasing environmental regulations (DPE impact)
• growing demand for data transparency in property valuation

Today, property estimation often relies on manual comparisons and subjective expertise, which can lead to inconsistencies between different agents or platforms.

The challenge is therefore to develop a data-driven estimation approach capable of supporting real estate professionals.

💡 Project Objective

The goal of this project is to design a Machine Learning model able to estimate the fair market price per square meter of a property.

The system aims to:

• support real estate agents during valuation
• reduce the time required for market analysis
• provide explainable pricing arguments
• increase trust for both buyers and sellers

🧠 Machine Learning Approach

The estimation model relies on supervised regression techniques applied to real estate transaction data.

Data sources

• DVF (historical property transactions)
• INSEE socio-economic indicators
• geospatial context (transport, services)
• property characteristics

Modeling pipeline

1. Data cleaning and preprocessing
2. Feature engineering (spatial and contextual variables)
3. Model training using gradient boosting algorithms
4. Evaluation and validation using regression metrics

Candidate models include:

• Linear Regression (baseline)
• Random Forest
• Gradient Boosting
• XGBoost

🔍 Explainable AI

A key requirement of the system is model transparency.

Explainable AI techniques are used to highlight the main factors influencing property prices, allowing agents to clearly justify their estimations.

Methods explored include:

• SHAP values
• feature importance analysis
• partial dependence plots

📈 Expected Impact

The integration of Machine Learning into the valuation workflow could provide several benefits:

• more reliable property pricing
• faster estimation process
• improved negotiation support
• increased conversion from estimate → signed mandate

Ultimately, the goal is to combine data-driven insights with human expertise in real estate decision-making.

🛠️ Technologies