SoccerPredictAI — Reports Overview

Quarto analysis reports generated from the DVC pipeline

This site contains reproducible analysis reports for the SoccerPredictAI end-to-end MLOps system. Each report is generated automatically after the corresponding DVC stage completes. Re-run make docs-build to refresh all rendered outputs.

Reports

#	Report	DVC stage(s)	Description
01	EDA & Data Profiling	load_data_from_sources, preprocessing	Raw data schema, temporal coverage, target distribution, preprocessing deltas
02	Feature Engineering	feature_engineering	Feature inventory, ELO rating distributions, rolling-stat correlations, completeness
03	Temporal Validation Strategy	split_data	Walk-forward fold structure, train/test Gantt, leakage assertion
04	Model Comparison & Selection	classification_models	CV metrics across candidate models, calibration comparison
05	Data Quality Gates	validate_raw, validate_interim, validate_features	Great Expectations dashboard, failed expectations drill-down
06	Feature Ablation Study	ablation_study	Log-loss by feature set, ELO contribution, delta heatmap
07	Final Model Evaluation	final_train	Test-set metrics, calibration curves, SHAP / feature importances, segment performance
08	Hyperparameter Tuning	tune_xgb	Optuna history, best vs default params, parameter importance
09	Batch Inference	batch_inference	Distribution shift (KS), prediction distribution, entropy, confident picks

How to regenerate

# Run the full pipeline then rebuild all docs
dvc repro
make docs-build

Reports use freeze: auto — cells are re-executed only when their source changes. Individual reports can be rendered with:

quarto render reports/qmd/01_eda_and_preprocessing.qmd