As a short example suppose we have a kind of pollution sensor and a jet fan in a tunnel.

Jet fan turn on/off according to the automatic scenario based on pollution sensor value.

Pollution itself is based on some parameters like number of cars in the tunnel, direction of wind,speed of wind etc.

And number of cars can be predicted according to date and time (month,day,hour,...) and some other parameters.


I want to generate an online simulation of this system, so that for the current state of the system can “predict” value of pollution sensor in the short-term furture (ex 1 hour).

What is the best neural network architecture for this and what will be good input features?

What I thought about input features is

Sensed value in the previous time period

X1:Pollution value

X2:wind speed

Actuator status :

X3:jetfan status (on/off)

date and time :

X4:season X5:month X6:day in week

  • 1
    $\begingroup$ You have shadowed specification, the most of important data is encoded in the "other parameters". I do not have the pollution sensor, so I do not know what data it generates, how often samples, what do you mean by predict? How much data you have already? What have you tried? What do you expect in the answers? I think this is a very broad question, a bit unclear, so maybe you could fill in the details? Why do you need neural networks? $\endgroup$ – Evil Jun 4 '16 at 20:17
  • 1
    $\begingroup$ This seems to be, basically, "Neural nets: how do I use one?" which is a very broad question. Also, as @EvilJS says, you've not given enough details to answer the question for the specific example you give. $\endgroup$ – David Richerby Jun 4 '16 at 21:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.