Logo Lanfrica

Imen-ABDELKADER/tunisia-apartment-price-predictor

Domaine:

socioeconomic

Type de record:

project
Créateur:
Ime
Hôte:
End-to-end MLOps pipeline for predicting apartment prices in Tunisia using web-scraped real estate data. Features automated training with MLflow, CI/CD via Jenkins, containerization with Docker, model monitoring (DeepChecks + Arize), and production deployment with FastAPI + Streamlit. Built with DagsHub (Nov 2025). # Tunisia Apartment Price Predictor – MLOps End-to-End **In progress – November 2025** Personal project (currently under active development) Predicting apartment prices in Tunisia using real estate data scraped from Tunisian websites. ## Current Progress (What I have already completed) - Web scraping of thousands of apartment listings from major Tunisian real estate websites - Exploratory Data Analysis (EDA) – price distribution by governorate, surface, rooms, neighborhoods, etc. - Data cleaning & feature engineering (location encoding, outlier treatment, missing values, etc.) - Preprocessing pipeline (scikit-learn ColumnTransformer + custom transformers) - Baseline modeling + advanced models (XGBoost, LightGBM, CatBoost, RandomForest, stacking) - Hyperparameter tuning and cross-validation ## Planned / In-Progress MLOps Architecture - Experiment tracking & model registry → MLflow + DagsHub - Data versioning → DagsHub - CI/CD → Jenkins - Containerization → Docker - Model monitoring & drift detection → DeepChecks + Arize AI - API serving → FastAPI - Interactive dashboard & predictions → Streamlit - Deployment → Render (or alternatives) ## Tech Stack Python • Pandas • Scikit-learn • XGBoost • LightGBM • MLflow • DagsHub • Docker • FastAPI • Streamlit • Jenkins • DeepChecks • Arize • Jupyter ## Project Goals - Build a robust, production-ready MLOps pipeline from A to Z - Deploy a live web app where anyone can estimate an apartment price in Tunisia - Showcase best practices: reproducibility, monitoring, automated retraining **Project still in active development** – new commits coming regularly (model serving, monitoring, CI/CD, and deployment phases are next). Feel free to star or watch the repo if you're interested in MLOps or real estate price prediction in Tunisia!