A classification report summarizes how well a classification model predicts each class. It combines precision, recall, F1-score, and accuracy so performance can be judged beyond a single headline number. section 12.4.1 section 12.5.1
Read each measure as a different operational question:
- Precision asks how many articles or alerts labeled as a class were correct. section 12.4.1
- Recall asks how many actual high-risk articles or threats the model identified. section 12.4.1 section 12.5.1
- F1-score balances precision and recall, which is particularly useful when classes are imbalanced. section 12.4.1
- Accuracy reports the percentage of all predictions that were correct, but it can conceal weak performance for an important class. section 12.4.1 section 12.5.1
Interpreting a cybersecurity report
A threat classifier can show 63% accuracy while achieving zero recall for threats. That result may look acceptable overall but means the model missed every threat in the reported test output. Compare class-level metrics and inspect the confusion matrix before trusting the model for alerting or response decisions. section 12.5.1
