Open MPiunti opened 1 year ago
I don't think this time management would live in Prophet, we just provide the functionality rather than handle specific requirements. I'd suggest minimising the number of CV folds by limiting the training period or the window size.
Would it be feasible a timeout management inside cross validation?
Running prophet in spark for thousands of parallel group of timeseries, i see that for certain groups cross validation is quite expensive So the idea to set a timeout for each forecast, to be handled by cross validation i.e.
if cross_validation takes more than '60 sec' the exception is handled is there any trick to mimic this ?
thanks