Model Training: Feature Importance, Lasso Selection, Final Evaluation

[Anadetjur]

This PR adds the final notebook ana_model_experiment_recovered.ipynb, which contains the complete model training pipeline and evaluation.
Key additions:
Feature importance analysis using Random Forest
Feature selection using Lasso Regression
Final model using Gradient Boosting Regressor with Lasso-selected features
Model performance comparison table
Notes
The final model balances interpretability and performance and is ready for review and integration.