Confusion Matrix

A Confusion Matrix is a square table that provides a detailed breakdown of a classification model's predictions against ground truthGround TruthGround truth refers to data collected at the Earth's surface to validate and calibrate information derived from remot... labels. Each row represents the instances of an actual class, while each column represents the instances of a predicted class. The diagonal elements show correct predictions (true positives for each class), while off-diagonal elements reveal specific types of misclassification. For a binary classifier, the four cells represent true positives, true negatives, false positives, and false negatives. Multi-class confusion matrices extend this to any number of categories. Confusion Matrices in Geospatial ClassificationConfusion matrices are the standard evaluation tool for land cover classificationLand Cover ClassificationLand cover classification is the process of categorizing Earth's surface into distinct classes such as forest, cropla... accuracy assessment, as recommended by mapping agencies and remote sensingRemote SensingRemote sensing is the science of collecting data about Earth's surface without direct physical contact, primarily usi... standards. They reveal not just overall accuracy but precisely which land cover classes are confused with one another. For example, a confusion matrix might show that deciduous and mixed forest classes are frequently confused, while water and urban classes are reliably distinguished. This class-specific insight guides targeted improvements, such as adding training samples for confused classes or incorporating additional features to distinguish them. Producer's accuracy (recall) and user's accuracy (precision) for each class are derived directly from the confusion matrix, along with the overall accuracy and the Kappa statistic. Derived Metrics and Practical UsageFrom a confusion matrix, numerous performance metrics are computed. Overall accuracy is the sum of diagonal elements divided by the total. Per-class precision, recall, and F1 scores provide class-specific performance assessment. The Kappa coefficient accounts for chance agreement, providing a more rigorous accuracy measure. For geospatial applications, error matrices are typically accompanied by area-weighted estimates that account for the different proportions of land cover classes in the study area, preventing dominant classes from inflating overall accuracy.