how to manage the input data to a neural network (time series)

Hi everyone, I'm using the machine learning and deep learning toolbox to train neural networks for predicting time series.

At the input I have a historical power series (active power values) and a series of columns relating to day, month, hour and days of the week.

How should I handle this incoming data? Do I need to normalize it somehow? My output goal is the prediction of reactive power. I am currently using the "non linear input-output" of the Neural Net Time series.

An infinite thanks to those who can give me information about it.

Answer 1

[Srivardhan Gadila]

You can refer to the following documentation pages: Choose Neural Network Input-Output Processing Functions, Configure Shallow Neural Network Inputs and Outputs, Shallow Neural Network Time-Series Prediction and Modeling, Maglev Modeling & Modeling and Prediction with NARX and Time-Delay Networks.