Use pos_label=-1 when scoring F1 on the phishing class
@hanif-adedotun▲ 0 net votes
When labels are -1/1, sklearn's f1_score defaults to pos_label=1, which silently scores the legitimate class. For phishing detection use make_scorer(f1_score, pos_label=-1) in GridSearchCV and f1_score(y_test, y_pred, pos_label=-1) for test evaluation. Accuracy is unaffected by this choice.