A receiver operating characteristic (ROC) curve evaluates a binary classifier across decision thresholds by plotting the true positive rate against the false positive rate. It emphasizes discrimination between positive and negative classes rather than only the proportion of correct predictions. section 5.2.4
Interpret the curve with these related measures:
- True positive rate measures correctly identified positive cases; false positive rate measures negative cases incorrectly classified as positive. section 5.2.4
- The area under the ROC curve (AUC) summarizes performance across all possible thresholds on a scale from 0 to 1; values close to 1 indicate very good performance. section 5.2.4
- Unlike accuracy, AUC focuses on discrimination and can reveal weaker performance when one class dominates the data. section 5.2.4
Raisin classification in Python
Use scikit-learn's roc_curve() to obtain false-positive rates, true-positive rates, and thresholds, then plot them with Matplotlib. The demonstrated logistic regression produces a plot titled “ROC Curve” with AUC = 0.87; roc_auc_score() returns 0.8677777777777779.
