AI 용어집
인공지능 완전 사전
True Positive (True Positive)
Case where the model correctly predicts the positive class, corresponding to observations correctly identified as belonging to the target class.
True Negative (True Negative)
Case where the model correctly predicts the negative class, representing observations correctly excluded from the target class.
False Positive (False Positive)
Type I error where the model incorrectly predicts the positive class, corresponding to false alarms or observations incorrectly classified as positive.
False Negative (False Negative)
Type II error where the model incorrectly predicts the negative class, representing missed detections of actually positive observations.
Recall (Recall/Sensitivity)
Model's ability to detect all positive observations, measured by the ratio TP/(TP+FN) indicating the detection rate of the target class.
Error Rate (Error Rate)
Proportion of incorrectly classified observations by the model, calculated as (FP+FN)/(Total) representing the overall error performance.
Accuracy
Overall proportion of correct predictions, calculated as (TP+TN)/(Total), measuring the general performance of the classifier.
Balanced Accuracy
Average of recall for each class, providing a metric adapted to imbalanced datasets by giving equal weight to all classes.
MCC (Matthews Correlation Coefficient)
Correlation coefficient between binary observations and predictions, considered a robust measure even for imbalanced classes.
Precision-Recall Curve
Graph plotting precision against recall for different thresholds, particularly useful for evaluating performance on imbalanced datasets.