How to forecast a temperature using neural networks??

I am new to neural networks and learning to use it for forecasting temperature. I am using below time series as input data - Outdoor temperature - Mean radiant temperature - humidity

I want indoor temperature as output and want 120 minutes future data of it.

My first question is, should i treat it as a fitting problem or time series problem?

Secondly I have attached a code i generated through time series tool for the same. (mat file is also attached for required data.)My question is that this code generates only a training network or I can use it for forecasting? How to modify it for getting 120 minute future indoor temperature data?

Answer 1

[Greg Heath]

1.Calculate the crosscorrelation function between inputs and target

2.Determine the significant lags

3.Calculate the target autocorrelation function

4. Determine the significant lags

5. help narxnet

6. Search for design examples using greg narxnet

7. Search for more examples using greg closeloop

Come back with specific questions.

Hope this helps.

Thank you for formally accepting my answer

Greg

[Rajat]

Hello Greg,

I am also working on something similar i.e. Neural networks for flood forecasting, I have started doing my literature review, but as m new to Neural Network toolbox, could you please explain on how to begin. as i was not able to understand much from the answer provided

[sanket jhala]

As per Greg's direction I searched about use of narxnet and close loop for multi-step ahead forecasting using back propagation and found below link very very helpful.

http://www.mathworks.com/matlabcentral/answers/14970