I am working on a LSTM model whose purpose is to estimate (in real time) the arrival date (eta) of a boat, based on the remaining kilometers to destination and several other variables. My data frame looks like this.
However, the eta is not known until the boat is arrived and therefore, I don't have access to this variable in real time.
So far I managed to build a multivariate time series model that looks back several time steps in the past (based on this tutorial) and predict the eta based on it. However it also takes into account the eta at the prior time steps, which does not work for my use case.
What I would like is a LSTM model that is suited for a response that is unknown as the time series evolve.
What would you advise so that I can fix this issue?
P.S.: I had an idea of using the remaining kilometers (bird_distance) as the response. However, I have to predict when it will reach 0, which I can't solve either.
Any help is appreciated