tender pelican
I have an xgboost model and I am trying to find that value of threshold that maximizes the product of TPR and (1-FPR)
fpr, tpr, threshold = sklearn.metrics.roc_curve(y_test, y_pred[:,0])
fpr_tpr = np.multiply((1-fpr), tpr)
max = np.argmax(fpr_tpr)
print(fpr[max])
print(tpr[max])
I have attached my ROC curve. According to the calculation of fpr and tpr above, I get
fpr = 0.5
tpr = 0.1
Somehow this does not look right to me, having tpr value so low.
Can you please let me know what is wrong?
amber remnant