Melbourne housing prediction
Comparing property price models, with XGBoost achieving a held-out test R² of 0.87.
Concept illustration for the project. This is not a chart of the underlying dataset.
The question
How well can property features explain house prices, and which features contribute most to the predictions?
A data science project comparing machine learning approaches to Melbourne property price prediction.
My contribution
Built and compared XGBoost and Random Forest models, applying feature engineering and hyperparameter tuning.
The approach
- Prepared property data and engineered features for modelling.
- Compared XGBoost and Random Forest on held-out test data.
- Reviewed feature importance to interpret the models and explain their predictions.
What the work supported
Location and land size emerged as the strongest price drivers in the project analysis.
Scope and limitations
Model performance applies to the project dataset and evaluation setup. Predictive associations do not establish causation, and the model is not presented as a current property valuation service.