My data frame consists of 3 input columns (factors 1, 2 and 3) and output column, i.e., revenue which are time varying parameters. I am trying to predict the revenue using neural networks for the subsequent 12 instances (say months). I have trained a network to work on the 3 inputs to calculate the revenue for all the instances till now. To predict future instances of revenue, I have creared a dataframe of forecast(nnetar(col1),nnetar(col2),nnetar(col3)) i.e., forecasted each input separately and then used the same neural network to work on forecasted inputs to predict the future values of revenue.
But I don't find this very accurate. Are there any better ways to actually do this?