Time series cross-validation for lasso
6 views (last 30 days)
I am trying to estimate a lasso model using time series data and I would like to use cross-validation to select the lamda parameter. However, K-flod validation randomly shuffles the data and therefore it is not optimal in a time-series context; something like forward chaining would be more appropriate. I have no clue how to do this in matlab as the lasso or the cvpartition functions do not offer such option.
Swetha Polemoni on 31 Mar 2021
Edited: Swetha Polemoni on 31 Mar 2021
Hi Nikolaos Vasilas
K-fold validation without random shuffling of data/ time series cross-validation is not supported by Matlab This requirement has been brought to the notice of developers. It might be considered in future release.