Precision and recall
Think of it like a fishing net: precision is about how many of the fish you caught are actually the type you wanted, while recall is about how many of the fish you wanted you actually managed to catch.
Precision measures the accuracy of positive predictions (true positives divided by total predicted positives), while recall measures the model's ability to find all relevant instances (true positives divided by total actual positives).
Precision is defined as the ratio of true positives to the sum of true positives and false positives (TP / (TP + FP)), representing the positive predictive value; recall is the ratio of true positives to the sum of true positives and false negatives (TP / (TP + FN)), representing the sensitivity of the classifier.
evolution
- 1955 · historyInformation Retrieval Foundations
Allen Kent and colleagues at Western Reserve University formally defined the concepts of 'precision' and 'recall' to measure the effectiveness of document retrieval systems.
- 1971 · historyCranfield Experiments
Cyril Cleverdon established these metrics as the standard for evaluating automated information retrieval systems during the influential Cranfield tests.
- 1997 · historyTREC Standardization
The Text Retrieval Conference (TREC) solidified precision and recall as the primary benchmarks for evaluating large-scale search and information extraction algorithms.
- 2006 · historyF-Measure Integration
The F-score became the standard harmonic mean for balancing precision and recall, widely adopted in the burgeoning field of machine learning classification tasks.