Presidor/nigeria-house-price-prediction-ml

This project develops a machine learning model that predicts house prices in Nigeria based on property features such as location, size, and housing attributes. The project demonstrates an end-to-end data science workflow including data preprocessing, model training, evaluation, and deployment through an interactive prediction app. # 🏠 Nigeria House Price Prediction ## 📌 Overview This project predicts house prices across Nigeria using machine learning models. It leverages a cleaned dataset of over **11,000 property listings** across **23 states** and **185 towns**, with features such as: - Bedrooms - Bathrooms - Toilets - Parking spaces - Property type - Town - State The goal is to provide an interactive tool for estimating property values and exploring housing trends in Nigeria. --- ## 📂 Project Files - **Nigeria House Price Prediction.pdf** → Full project report with data exploration, cleaning, modeling, and evaluation. - **nigeria_houses_data.csv** → Raw dataset containing property listings. - **nigeria_houses_data_clean.csv** → Cleaned dataset after removing duplicates, renaming columns, and handling outliers. - **house_price_model.pkl** → Saved trained pipeline (XGBoost model). - **app.py** → Streamlit app for interactive price prediction. --- ## 🧹 Data Cleaning Steps performed: - Removed **10,438 duplicate entries**. - Renamed `title` → `property_type`. --- ## Feature Engineering **Created a new feature:** - restrooms = average of bathrooms & toilets - total_rooms = bedrooms + restrooms - bedrooms_per_restrooms ratio --- ## 🚀 Features - Data cleaning and preprocessing pipeline - Exploratory data analysis (EDA) - Supervised machine learning model training - Model evaluation and performance comparison - House price prediction mini app - Model serialization for deployment --- ## 🧠 Machine Learning Workflow 1. Data collection and loading 2. Data preprocessing and feature engineering 3. Exploratory data analysis 4. Model training 5. Model evaluation 6. Hyperparameter Tuning 7. Prediction interface --- ## Model Used 1. Linear Regression → **R² = 0.47** 2. Random Forest Regressor → **R² = 0.49** 3. XGBoost Regressor → **R² = 0.54 (Best performing model)** --- ## 📊 Insights - Most expensive towns: Ifako-Ijaiye, Maitama District, Guzape District, Asokoro District, Katampe - Cheapes …