I came across this excellent question and answer by Adam Sypniewski. Here method of using RNNs (e.g. LSTMs and GRUs) for predicting time series where the dependent variable depends on several independent variables. Just wondering, is anyone aware of some Keras code for this?
Please go through this tutorial:
The author focuses on LSTM-RNNs with explicit instructions on how to re-shape your dataset into the format that Keras requires, which can be VERY tricky if not properly understood.
From the same question you mentioned, you will find an interesting link to Keras documentation. If you navigate to Multi-input and multi-output models you'll find an example code that does exactly this.