Precision and Recall

Precision and Recall are fundamental evaluation metrics for classification models that provide complementary perspectives on prediction quality. Precision (positive predictive value) measures the proportion of positive predictions that are actually correct: of all instances the model labeled as a particular class, how many truly belong to that class. Recall (sensitivity or true positive rate) measures the proportion of actual positive instances that were correctly identified: of all instances that truly belong to a class, how many did the model find. A model with high precision but low recall is conservative, making few mistakes but missing many true instances, while a model with high recall but low precision finds most true instances but also produces many false alarms. Precision-Recall Tradeoffs in Geospatial ApplicationsGeospatial applications often face important tradeoffs between precision and recall depending on the consequences of different error types. Building detection for damage assessment prioritizes recall because missing a damaged building (false negative) has serious humanitarian consequences, even if some false detections require additional verification. Deforestation alerting may also emphasize recall to ensure no illegal clearing goes undetected. Conversely, high-confidence mapping products prioritize precision to ensure reported features actually exist, accepting that some real features will be missed. Property valuationProperty ValuationProperty Valuation is the process of estimating the monetary value of real estate based on location, physical charact... models need balanced precision and recall across price ranges to avoid systematic over- or under-estimation. Metrics and Practical EvaluationThe precision-recall curve plots precision against recall at different classification thresholds, providing a complete picture of the tradeoff. Average Precision (AP) summarizes the curve as a single number, commonly used to evaluate object detectionObject DetectionObject Detection is a computer vision technique that identifies and localizes specific objects within images or video... models. In multi-class geospatial classification, per-class precision and recall identify which land cover types are reliably mapped versus which suffer from confusion. Macro-averaged precision and recall treat all classes equally, while micro-averaged metrics weight by class frequency.