nourhen369/House-Price-Prediction

tunisian house price prediction ML project using data scrapd from tayara.tn using BeautifulSoup. # 🏠 House Price Prediction in Tunisia This project focuses on building a machine learning pipeline to predict real estate prices in Tunisia using data scraped from Tayara.tn. The goal is to estimate property prices (in TND) based on features such as property type, surface area, number of bedrooms and bathrooms, and location. --- ## 🕸️ Web Scraping - Data was scraped from Tayara.tn using **BeautifulSoup** in Python. - Extracted fields included: - `Title` - `Price` - `Type` - `Area` - `Localisation` - `#Bedrooms` - `#Bathrooms` --- ## 🧼 Data Cleaning - Removed unrealistic price entries. - Imputed missing values using a combination of: - **KNN Imputer** (for categorical and numerical fields) - **Median imputation** (for area and bedrooms) - **Mode imputation** (for bathrooms based on type and location) - Dropped outliers and irrelevant columns. 📝 You can find **data visualizations before and after cleaning** in the following reports generated with `ydata_profiling`: - `data/immobiliers_report_before_cleaning.html` - `data/immobiliers_report_after_cleaning.html` --- ## 🧠 Feature Engineering Several new features were engineered to improve model performance: - `Area_x_Bathrooms` (interaction term: Area × Bathrooms) - `Log_Area` and `Log_Price` (log-transformed features to handle skewed distributions) - `Bathrooms_per_Bedroom` (ratio feature) Features derived directly from the target, such as `Price_per_m2`, were excluded to prevent data leakage. --- ## 🔧 Modeling Pipeline The machine learning pipeline was built using **PyCaret**, which automates: - Imputation - Encoding of categorical variables - Feature scaling - Dimensionality reduction (PCA) - Feature selection - Model training and evaluation Final model was built using **XGBoost** and **scikit-learn**, independently of PyCaret and persisted with `joblib`. --- ## 🚀 Deployment - **API:** FastAPI serves the trained model for prediction. - **Web Interface:** Streamlit provides a simple user interface …