I have a target variable, 'y,' observed at quarterly intervals, with 10 exogenous variables, each with their respective future forecast values. My aim is to forecast 'y' specifically 5 years ahead (equivalent to 20 quarters). Additionally, I have access to future forecasts for the 10 exogenous variables. So, I'm training the model using data up to 2017 and fine-tuning it to forecast the period from 2018 to 2022. Are there alternative data splitting methods I should consider to meet this objective of forecasting 5 years into the future?