Renovation Cost Prediction at an Early Stage: an Experiment Using Machine Learning Regression and Classification Techniques

Abstract: Accurate early-stage cost forecasting in renovation projects helps property owners make informed decisions, conserving resources and enhancing efficiency. However, estimates are challenging due to project complexity. Traditional methods, often intuitive and dataset-limited, lack precision. This study analyzes 104 projects using decision trees, finding that regression achieves a lower mean absolute percentage error (MAPE) of 7 %, versus 15 % for classification. A combined approach, integrating external data from Google Earth and Google Solar API, improves accuracy and traceability, emphasizing the potential for reliable, data-driven decisions in renovation projects.