As a beginner data scientist I have a few questions I would like to ask to put me on the right track. Athough this competition has ended I find this to be an educative moment for me.

I am an IoT specialist and I have done a lot of IoT projects and I would like to improve my IoT projects by adding a tinge of data science and machine learning to it.I recognize that this is a time series forecasting problem and there are a number of ways to go about that.

Firsty do I form new datasets from each soil humidity set. If that is the case then I end up with 4 new datasets containing each soil humidity profile?

Secondly, How do I predict the next 4 days or 6 days? I am going with the recurrent neural network(LSTM) approach. Do I split my data into train and test sets after which the results for the test sets then becomes what I actualy want to predict. I ask because the test sets provided each soil humidity column does not contain any other information from any column.

Any help will gladly be appreciated.