This answer suggests a way of doing leave-one-out cross-validation on time series data:
An approach that's sometimes more principled for time series is forward chaining, where your procedure would be something like this:
- fold 1 : training [1], test [2]
- fold 2 : training [1 2], test [3]
- fold 3 : training [1 2 3], test [4]
- fold 4 : training [1 2 3 4], test [5]
- fold 5 : training [1 2 3 4 5], test [6]
I realized that we can "It seems like a nice research topic" another time series from the resulting MSE of each fold. While it's not time-related, I presume the data to have some dependence structure, so time series techniques should still apply.
This derived time series seems like an interesting research topic, so my question is, has someone did any research on it? I don't even know how to pick the search keyword.
