0
$\begingroup$

Hey I am building logistic model. My train data has 4000 observations and my test set has 1000 observations. What suprised me is fact that for train set I get AUC 0.9 but on my test set I get AUC 0.92. What does is mean? There must be sth wrong with my model or on the contrary the model is very good?

Improve this question
$\endgroup$
3
  • $\begingroup$ Was the split into training and test sets done randomly? $\endgroup$
    – rolando2
    Commented May 31, 2021 at 12:15
  • 1
    $\begingroup$ "There must be sth wrong with my model" - no, for instance, data in your test set might be "easier" to predict. Like studying past exams that are difficult and then the actual exam is comparatively easy $\endgroup$
    – Laksan Nathan
    Commented May 31, 2021 at 12:16
  • $\begingroup$ @rolando2 I use createDataPartition function from caret package. Distribution of my depedent variable is almost the same in test and train sample $\endgroup$
    – Math122
    Commented May 31, 2021 at 12:23

1 Answer 1

Reset to default
1
$\begingroup$

Well the model should be better on the training data generally since it uses that data to train the parameters. However due to randomness in the sample the accuracy can go a little bit up and down so it is not that suprising that the test data accuracy was higher in this case, the difference in accuracy is probably neglible. Overall you can probably say that the model is good since the model obviously can explain unobserved data.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.