glen-bert/SolarForecastingWithML ? reverse-engineered prompt

Build me a clean Jupyter notebook that forecasts solar power generation from weather data using machine learning.

I want the notebook to load a Kaggle style dataset, clean missing values and outliers, do basic exploration, and create useful time based features like day and year sine and cosine values. Then train and compare a few regression models, including linear regression, decision tree, random forest, gradient boosting, SVR, and XGBoost. Please evaluate them with R squared, MSE, RMSE, and MAE, and show which model performs best.

Also include feature importance so I can see what matters most, especially radiation, sunshine, and air temperature if they stand out. Add a simple hyperparameter tuning step for the stronger models and make the notebook easy to read with charts and short explanations for each section. If you need to look up current library docs online, go ahead.