I have implemented a 5 fold Cross-Validation method for classifiers in MATLAB. To choose the best model i evaluate the best model as the one that obtained a better average accuracy.

Now i would like to know if accuracy is an acceptable standard measure or it is advisable to use other metrics or even statistics. For example, there is any difference between magnifying the average accuracy or minimizing the MSE?

  • $\begingroup$ First of all, MSE is a regression measure, and accuracy is a classification measure. Which kind is your problem? $\endgroup$ Nov 5, 2019 at 12:29
  • $\begingroup$ @ItamarMushkin Actually I had the same doubt too, but since i am using the 'trainbr' (Bayesian Regularization) learning function the choice of MSE is forced. Think at the same also with crossentropy. My problem is to classify different phases of a movement. At each validation should I consider the average accuracy or the average performance? $\endgroup$
    – Mirko Job
    Nov 5, 2019 at 12:58
  • $\begingroup$ What's average performance? $\endgroup$ Nov 5, 2019 at 13:05
  • $\begingroup$ ... If your learning function forces the MSE loss, it is simply not appropriate for a classification task. $\endgroup$ Nov 5, 2019 at 13:06


Your Answer

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

Browse other questions tagged or ask your own question.