Split a chronologically sorted data set (S input)in a percentage of records used for training (P output),while leaving out future records (F output)for out-of-sample error metrics or concurrent visuals.
