Mossaabjelliti/tunisia-real-estate-ml

A web app that predicts real estate prices in Tunisia # Tunisia Real Estate Predictor 🇹🇳 > An end-to-end machine learning project for estimating residential property prices in Tunisia from scraped real-estate listings. This project explores the complete data-science workflow: collecting real-world property listings, cleaning and transforming noisy data, performing exploratory analysis, engineering predictive features, training regression models, and exposing predictions through an API. ## Project goals The objective is not only to train a model, but to build a reproducible pipeline from raw web data to a usable prediction service. ```text Property listings ↓ Data collection / scraping ↓ Cleaning & validation ↓ Exploratory data analysis ↓ Feature engineering ↓ Model training & evaluation ↓ Best model ↓ FastAPI prediction service ``` ## What the project demonstrates - Web scraping and data acquisition - Data cleaning and preprocessing - Exploratory data analysis (EDA) - Feature engineering for tabular regression - Machine-learning model comparison - Model evaluation with regression metrics - Model serialization and reuse - FastAPI model serving - Separation between data collection, ML logic, and API layers ## Project structure ```text ├── scraper/ # Property listing collection ├── data/ # Raw and processed datasets ├── notebooks/ # EDA and model development ├── src/ # Reusable preprocessing, features and training logic ├── api/ # FastAPI prediction service ├── app/ # Optional frontend / prediction interface ├── models/ # Serialized trained models ├── requirements.txt └── README.md ``` ## Data pipeline ### 1. Collection Property listings are collected from real-estate listing sources and converted into structured records. ### 2. Cleaning The pipeline handles typical web-data issues such as missing values, inconsistent representations, duplicated listings, and numerical fields embedded in text. ### 3. Feature engineering Potential …