CODEWITHSUNDEEP
pythonmachine-learningscikit-learnartificial-intelligence
Aadyant Khatri
Aadyant Khatri

Reputation: 1

What could be the reason for this difference?

What could be the reason for the difference in accuracies?

y_pred = model.predict(x_test)

ans = []

for i in y_pred:

    ans.append(0 if i[0]<0.5 else 1)

ans = np.array(ans)

print("Accuracy:",metrics.roc_auc_score(y_test, ans))

Accuracy: 0.7781798160210841

While

print("Accuracy:",metrics.roc_auc_score(y_test, y_pred))

Accuracy: 0.8949947592358702

Error

Upvotes: 0

Views: 51

Answers (1)

lejlot
lejlot

Reputation: 66775

ROC score is not accuracy. It is an area under the curve, that emerges when you construct your predcition by varying a threshold. In your code you removed the notion of "uncertainty", by outputing only 0s and 1s, and thus the ROC curve became flat - there is no threshold that affects your predictions.

Upvotes: 1

Related Questions