Hi. Wanted to know if we can use the past consumption values which we predicted on test data to use for training for a future time. Like: Let’s say there is test period in entire March 2015 and I build a model to predict these values, now can I use these predicted values to make a model to predict in May 2015? I think this is allowed as the predicted data is past data for predictions for May 2015.
Yes, this would be acceptable.
Thanks! One more question: If I just impute the values from the previous year at same date and time at prediction time, will this be acceptable? It doesn’t use the information: which year is this? Just what was the consumption at the same point of time last year?
Let’s say: We want to predict per quarter consumption for March 2015, and I just put March 2014 consumption values as predictions. I think this should be acceptable as this doesn’t take in to account which year is this.