Nada-elkamel/Aqari-Tunisia-Property-Estimator
Domain:
socioeconomicRecord type:
software
Creator:
Nad
Host:
# 🏠 Aqari-Tunisia-Property-Estimator
🔍 Overview
Aqari is a machine learning pipeline designed to predict real estate prices in Tunisia. Built with a philosophy of "Zero rows dropped", the system handles missing values through advanced imputation techniques and manages outliers via Winsorization rather than deletion.
Key Objectives
✅ Preserve all 8,129 property listings without dropping rows
✅ Handle missing values using multivariate imputation strategies
✅ Engineer meaningful features from raw data
✅ Train and compare multiple regression models
✅ Deploy an interactive web application for instant price estimation
📊 Dataset
| **Property** | **Value** |
| ------------------- | ----------------------------------- |
| Source | dataSetFull.csv |
| Total Rows | 8,129 listings |
| Original Features | 26 columns |
| Target Variable | price_tnd (Price in Tunisian Dinar) |
| Geographic Coverage | 24 Governorates across Tunisia |
Feature Categories
📍 Location Features
├── governorate, city, location
├── latt, long, distance_to_capital
🏗️ Structural Features
├── Area, pieces, room, bathroom
├── age, state, garage
✨ Amenities (Binary Flags)
├── garden, concierge, beach_view, mountain_view
├── pool, elevator, furnished, equipped_kitchen
├── central_heating, air_conditioning
💰 Target
└── price_tnd, price_eur
Missing Value Summary (Before Imputation)
| **Feature** | **Missing Count** | **% Missing** | **Imputation Strategy** |
| ------------------ | ----------------- | ------------- | -------------------------------- |
| age | 4,145 | 50.99% | Median imputation |
| price_tnd | 1,708 | 21.01% | IterativeImputer (BayesianRidge) |
| city | 1,316 | 16.19% | Mode per governorate |
| pieces …