In the above snap from the week 3 optional lab on logistic regression, where i have added more points to the training data. I see that the logistic regression still mis-classifies certain points. does this not make the logistic regression fail when classifying this kind of data just like linear regression which fails when a point too farther is added?
Which additional points did you add?
the crosses and circles you may see in the attached figure.
Hello @AzharAli, yes, logistic regression model can make wrong prediction.
While linear regression model is a line in the features space so that outliners will get pretty bad predictions, logistic regression model is a hyperplane in the features space so that False (benign) samples that cross the hyperplane and locate inside the cluster of True (malignant) samples would be easily mispredicted.
Cheers,
Raymond
In general all models output an answer with a probability below 100%, so there are always cases when the model fails!