I made a prediction model on time series data using the SVR method then to determine the best parameters I used the gridsearch algorithm and k-fold cv using the R software by default. The data that I use for training data is 358 data. but I want to ask, is k-fold cross validation suitable for time series data? What is the reason ? because all this time we know that cross validation takes data randomly while the order of the data in the time series is important.

or should I use another validation method on SVR? please give me an explanation, thanks.



Browse other questions tagged or ask your own question.