Automated Property Valuation Model with Machine Learning

Create a comprehensive Python script that integrates pandas and scikit-learn to build a dynamic property valuation model pulling live data from multiple Excel sheets. The script should incorporate hedonic pricing methodology, using features like square footage, location coordinates, neighborhood crime rates, and recent sales data to generate real-time property value estimates with 95% confidence intervals. Include error handling for missing data and generate a detailed Jupyter notebook with model performance metrics and feature importance visualization.