This project aims to predict housing prices across different regions of Aotearoa (New Zealand) using machine learning models and historical data. We incorporate regional economic indicators, property attributes, and time-series data to build a model that forecasts housing prices. The final output is an interactive demo where regions are color-coded based on predicted price ranges, offering a visual and intuitive way to explore housing market trends.
- Time-Series Forecasting: Predicts housing prices based on historical data trends.
- Regional Breakdown: Offers insights into housing prices across different regions, including urban and rural areas.
- Economic Indicators: Utilizes economic data like interest rates, unemployment, and inflation as features.
- Interactive Visualization: Visualizes regional price predictions on a map of Aotearoa, color-coded from green (less expensive) to red (more expensive).
- Model Evaluation: Provides model performance metrics such as Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE).
- Real Estate New Zealand (REINZ): Property sales and housing price data.
- Geospatial Data: Regional and district-level geographic features.
- Economic Indicators: Data on interest rates, inflation, unemployment, etc.
- Government and Open Data Portals: Various economic and property datasets.
Our machine learning model predicts housing prices using:
- Linear Regression: As a baseline model to assess performance.
- Random Forest Regressor / XGBoost: To capture complex relationships in the data.
- Gradient Boosting: For advanced modelling
The final project delivers an interactive app built with Dash or Streamlit that presents predicted housing prices across Aotearoa. Key features of the app include:
- A map of New Zealand, with regions color-coded to indicate housing price predictions.
- Green: More affordable regions.
- Red: More expensive regions.
- Users can click on each region to see detailed price trends and breakdowns of contributing factors.