Generate Report

Generate a comprehensive model evaluation report by following these steps:

1. Locate artifacts in the directory specified by $ARGUMENTS (default: output/). Read evaluation_report.md if it exists to gather existing metrics.

2. Load the trained model (look for .joblib or .pkl files) and extract:

   • Model type and hyperparameters
   • Feature names and count
   • Feature importance scores (if the model supports them)

3. Load test data (look for x_test.parquet and y_test.parquet) and compute:

   • Prediction distribution statistics
   • Error distribution statistics
   • Any metrics not already in the evaluation report

4. Fill in the report template from templates/report_template.md. Replace each placeholder section with actual data:

   • Executive Summary: 2-3 sentence overview of model performance
   • Dataset Overview: number of samples, features, target variable description
   • Model Configuration: model type, all hyperparameters in a table
   • Performance Metrics: all computed metrics in a table
   • Feature Importance: top 5 features ranked by importance with scores
   • Recommendations: at least 3 actionable suggestions for improvement

5. Save the completed report to output/full_report.md.

6. Save the report generation code as a standalone Python script.

   • Collect all the code used in steps 1-5 into a single .py file.
   • The script should be runnable independently (include all imports, logging config, and a main() function).
   • Ask the user where they would like to save the script (suggest a default like scripts/generate_report.py).

7. Log a summary of what was generated using the project's logging format.

Follow the coding standards in CLAUDE.md. Use polars for any data loading.