TN = np.sum((labels == 0) & (predicted_labels == 0))
FP = np.sum((labels == 0) & (predicted_labels == 1))
TPR.append(float(TP) / (TP + FN))
FPR.append(float(FP) / (TN + FP))
idx = np.argsort(probabilities)
np.argsort(probs)
return np.array(TPR)[idx], np.array(FPR)[idx], probabilities[idx]
logit = LogisticRegression()
logit.fit(X_train_rank, y_train_rank)
probs = logit.predict_proba(X_test)[:, 1:]
probs = np.array([x[0] for x in probs])
tpr, fpr, thresholds = roc_curve(probs, np.array(y_test))
plt.plot(fpr, tpr)
plt.plot([0, 1], [0, 1], ls="--", c=".3", alpha=0.5)
plt.xlabel("False Positive Rate (1 - Specificity)")
plt.ylabel("True Positive Rate (Sensitivity, Recall)")
plt.axvline(0.4)
plt.axhline(0.6)
plt.title("ROC plot of fake data")
plt.show()
logit = LogisticRegression()
logit.fit(X_rank, y_rank)
