I am really new to using LSTMs for time series forecasts. I am having a little problem defining the problem was hoping someone could help me.
Say I have a bunch of claim payments (see image). I would like to forecast sum of all claim payments for Apr-17 using previous data. Should my training dataset have:
A) individual claims and monthly payments (Jan-16 to Mar-17 in the image) ?
The examples I've seen usually have a single array of values, similar to a time series forecast of aggregated numbers. I've looked at Multivariate LSTMS and I am not sure they apply for this scenario because all values represent the same type of payment.
Any guidance would be appreciated