How to evaluate machine learning model performance

Short answer

Split data into training and test sets, then compute relevant metrics. Use accuracy for balanced data, precision and recall for imbalanced data, and cross-validation for robust estimates. For handling class imbalance, see how to handle imbalanced datasets in classification. For foundational ML concepts, see what is the difference between supervised and unsupervised learning.

Steps

1. Split data using train_test_split or cross-validation
2. Train your model on the training set
3. Predict on the test set
4. Compute metrics: accuracy, precision, recall, F1
5. Plot ROC curves and check calibration

Tips

• Never evaluate on training data alone
• Use stratified sampling for imbalanced datasets
• Understand the bias-variance tradeoff in machine learning to interpret errors