How to understand which variables are redundant in a neural network classification problem

Hi all,

I'm working on a LSTM network and I'm trying to predict if a pump system will fail.

I have a dataset with 50 sensors, that are my input variables, and I would like to reduce the number, so I can streamline the neural net a bit.

I would like to know if there is a function that identifies similarities between the variables, so as to eliminate the redundant ones.

I don't know if a variable represents the humidity or another one represents the temperature, they are indicated only as sensors. For this reason I did normalization on the dataset.

Thanks in advance.

Answer 1

[Srivardhan Gadila]

One way is to find correlation between the variables. Refer to corr, Correlation and correlation functions in MATLAB.