Classification

[GernotMaier]

No description provided.

[Copilot]

Pull request overview

This PR introduces gamma/hadron classification functionality and significantly refactors the codebase to reduce duplication and improve maintainability. The changes consolidate common patterns for both stereo reconstruction and classification tasks.

Key Changes:

• Adds new classification training and application scripts for gamma/hadron separation
• Introduces energy and zenith binning for classification models
• Refactors data processing, model loading, and evaluation code into shared modules
• Adds pre-training quality cuts for data filtering
• Removes all existing unit tests without replacement

Reviewed changes

Copilot reviewed 23 out of 23 changed files in this pull request and generated 11 comments.

Show a summary per file

File	Description
tests/unit_tests/test_*.py	All unit tests deleted - significant test coverage reduction
tests/resources/classify-parameter.json	New configuration file defining energy and zenith bins for classification
tests/conftest.py	Updated with shared fixture for telescope data
src/eventdisplay_ml/features.py	New module centralizing feature definitions for both analysis types
src/eventdisplay_ml/hyper_parameters.py	New module defining hyperparameters and pre-cuts
src/eventdisplay_ml/config.py	New configuration module handling CLI arguments for both training and application
src/eventdisplay_ml/models.py	New module consolidating model loading and application logic
src/eventdisplay_ml/scripts/train_xgb_classify.py	New classification training script
src/eventdisplay_ml/scripts/apply_xgb_classify.py	New classification application script
src/eventdisplay_ml/scripts/train_xgb_stereo.py	Refactored to use new config and model modules
src/eventdisplay_ml/scripts/apply_xgb_stereo.py	Refactored to use new model application infrastructure
src/eventdisplay_ml/evaluate.py	Enhanced with classification evaluation and improved feature importance handling
src/eventdisplay_ml/data_processing.py	Significantly refactored with support for both analysis types and energy/zenith binning
src/eventdisplay_ml/utils.py	Added model parameter loading and output filename generation utilities
src/eventdisplay_ml/training_variables.py	Deleted - functionality moved to features.py
pyproject.toml	Added new CLI entry points for classification scripts
docs/changes/*.md	Added changelog entries for new features and refactoring

Comments suppressed due to low confidence (1)

tests/unit_tests/test_utils.py:1

• This PR deletes all existing unit tests (test_utils.py, test_training_variables.py, test_evaluate.py, test_data_processing.py, test_train_xgb_stereo.py, test_apply_xgb_stereo.py) without providing replacement tests. This represents a significant reduction in test coverage for the codebase. The new functionality introduced (classification routines, refactored data processing, new model loading, etc.) also lacks test coverage. Consider adding comprehensive unit tests to maintain code quality and prevent regressions.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The apply_regression_models function uses group_df.index to index into the preds array on line 211. However, if the DataFrame has been reset with a non-default index or has gaps in indices, this could lead to IndexError or incorrect assignment. The function should use positional indices or ensure the index is properly aligned with the predictions array.