PercyMayaba/South-African-Health-Insurance-Premium-Prediction-Neural-Network-

This project builds a synthetic South African health insurance dataset and trains a neural network using TensorFlow/Keras to predict insurance premiums. # South-African-Health-Insurance-Premium-Prediction-Neural-Network- This project builds a synthetic South African health insurance dataset and trains a neural network using TensorFlow/Keras to predict insurance premiums. It demonstrates how machine learning can support actuarial pricing, risk-based underwriting, and premium optimization. The project is built to run on Google Colab or deploy easily to Hugging Face Spaces (Gradio app). ⚙️ Tech Stack 🧠 TensorFlow / Keras — Neural network for regression 🐍 Python (NumPy, Pandas, Scikit-learn) — ETL and preprocessing 📊 Matplotlib / Seaborn — Data visualization 💻 Gradio — Web app interface for Hugging Face Spaces ☁️ Google Colab — Primary development environment 🧩 Key Features Generates synthetic health insurance data resembling South African demographics. Predicts insurance premium prices based on age, BMI, province, and chronic conditions. Uses StandardScaler preprocessing and a fully-connected neural network. Supports Gradio app deployment on Hugging Face Spaces. Includes ideas for model fairness, explainability, and production deployment. 🧪 Dataset Description Synthetic dataset includes: Feature Description Age Age of the individual Gender Male / Female BMI Body Mass Index Smoker 1 if smoker, else 0 ChronicCondition 1 if has chronic disease Province South African province (categorical) Dependents Number of dependents Income Monthly income (ZAR) Premium Target variable – annual health insurance premium (ZAR)