mo7medSa3d/Real-Estate-Price-Prediction-Project-in-Egypt-Dubizzle-

This project focuses on building a high-accuracy Machine Learning Regression Model to predict residential unit prices within the Egyptian real estate market, utilizing data aggregated from an online property classifieds platform. # 🏠 Real Estate Price Analysis & Prediction in Egypt (Dubizzle Properties) ## 🌟 Project Overview This project aims to build a high-accuracy **Machine Learning model** to predict residential property prices in the Egyptian market, using data collected from real-estate listing platforms. * **Main Objective:** Identify the key factors influencing property prices and build a reliable **Regression model**. * **Languages & Tools:** Python (Pandas, NumPy, Matplotlib/Seaborn, Scikit-learn, XGBoost). * **Current Model Status:** Achieved **moderate accuracy** (**R² ≈ 0.51** after feature engineering), indicating that the most critical improvements depend on **increasing dataset size** and **adding missing features**. --- ## 💾 Dataset and Features The dataset was collected from real-estate listings and contains **4,500 rows** (before cleaning). | Feature (English) | Description | Example | |-------------------|-------------|---------| | `price` | Property price (the **target** variable) | 4,492,000 EGP | | `area` | Property area (m²) | 151 m² | | `beds` | Number of bedrooms | 3 | | `baths` | Number of bathrooms | 3 | | `compound` / `location` | Compound or neighborhood name | First Settlement / Diar1 | --- ## 🧹 Data Preprocessing Methodology Several advanced data cleaning and feature engineering steps were applied to improve model stability: 1. **Initial Cleaning:** Removed symbols, normalized values, and converted numeric columns using `Numeric` types. 2. **Outlier Removal (IQR):** Applied to all numerical features (`price`, `area`, `baths`, `beds`). 3. **Advanced Location Encoding (Target Encoding):** * The categorical `compound` feature was replaced with a new numeric feature: **`Compound_Value`**. * This represents the **average property price per compound**, making it the strongest predictor. --- ## 🤖 Modeling and Performance Three regression models were tested: | Model | R² Score | Notes | |-------|----------|-------| | **Linear Regression** | 0.27 | Weak …