Performance metrics

updated: 2023-01-07

Binary vs Continuous

• AUC-ROC quantifies the overlap of two distribution of metrics, one for positive and the other for negative classes.
  • sklearn.metrics.roc_auc_score — scikit-learn 1.2.0 documentation
• Average Precision is useful when we are interested in a few positive examples out of a sea of data samples. One useful domain is information retrieval, where one aims to identify relevant webpages out of all web pages in the Internet.
  - sklearn.metrics.average_precision_score — scikit-learn 1.2.0 documentation

Multiple objectives

• Performance profile:
  • Performance Profiling: Theoretical Foundations, Applied Implementations and Practitioner Reflections: Journal of Sport Psychology in Action: Vol 12, No 4