I would like to divide my data of only 100 instances into training and testing an use the training data to fit a curve(interpolate) and use the testing data to calculate the error at the interpolated points. I have been using 80:20 ratio but i have been reading on the internet that this ratio is good for large dataset. So, could someone please help me how to decide this with a reference?

  • $\begingroup$ If you randomly choose the training set, I advise repeating the split and regression several times to ensure similar parameters and results each time. If the results are not approximately repeatable for different random data splits, this would be a consideration that there was a problem. The more often that different random splits yield essentially the same results, the more confident you can be in those results. $\endgroup$ – James Phillips May 28 at 10:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.