Automated Property Valuation Model with Machine Learning Regression

Create a comprehensive Jupyter notebook that develops a predictive property valuation model using scikit-learn regression techniques. The script should integrate historical real estate transaction data from a Google Sheet, perform feature engineering with pandas, train multiple regression models (Random Forest, Gradient Boosting, Linear Regression), and generate a comparative performance dashboard with cross-validation metrics. Include geospatial feature extraction, handling of categorical variables like property type, and a method to export prediction intervals directly back to the source spreadsheet.