Performance MetricsIt is an evaluation criterion for measuring the generalization ability of a model, which is used to determine the quality of machine learning results.

When comparing the capabilities of different models, using different performance metrics will lead to different evaluation results. Judging the quality of a model is relative and depends on what performance metric is used. The type of performance metric depends on the actual task requirements.

Common performance metrics

• Performance Metrics for Regression Tasks: Mean Square Error
• Performance measures in classification tasks: error rate and accuracy; precision, recall and F1 measure; receiver operating characteristic curve ROC and AUC (Area under ROC Curve); cost-sensitive error rate and cost curve.