Crossvalidation Question

[haifengl]

Crossvalidation Question

[haifengl]

The first parameter of CrossValidation is the number of samples in the data. The second is the number of rounds of cross validation. In your case, it should be 2. BTW, it is better to use a larger number for cross validation (e.g. 5 or 10). Otherwise, the estimated error rate may be very biased.

Besides, Bootstrap may be a better chance for more accurate estimation (with a large number of rounds such as 100). But it takes longer time. Worth to try if your data is not too large.

[Xyclade]

Thank you for the quick feedback. The main reason I want to do a 50% training-testing split is because this is exactly how it is done in an example in R, and my goal is to work through that example using Smile instead. I will experiment a bit more and use Bootstrap too to see how that works. Thanks again!