Predicting if a time-series nonlinear signal will reach end positions( 0 or max)

2 views (last 30 days)
Hello All, I am trying to build a Matlab/Simulink model to predict if a signal would reach end values. However the incoming data is realtime/dynamic , I have less control over the past data(only a few samples to say) in order to avoid more delay in the control system.
Can this done using Neural Nets considering less delay/few past samples? I went through some answers related to market trends here but seemed complex. Could a simple NN model be built with limited(Sorry,I am beginner to NN) or any simple method is possible ?
Please any idea or direction would be helpful. Thank you in advance
Note:In realtime data is NON LINEAR/EXPONENTIAL.
  2 Comments
Greg Heath
Greg Heath on 28 Jul 2017
Why don't you pick a dataset from MATLAB's NN example library?
help nndatasets
doc nndatasets
Greg
Pranav
Pranav on 31 Jul 2017
Hi Greg, Thank you for the reply.
I am using simplenar_dataset model given one only time-series to predict the next value of exponential curve. Is there a time delay introduced using this model.
What could be least number of previous samples used without/less delay to predict? In the example , it takes 100 samples which is too high for my application.

Sign in to comment.

Accepted Answer

Greg Heath
Greg Heath on 31 Jul 2017
The answer to questions you have about input and feedback delays can be answered if you have relevant design data by using
1. SIGNIFICANT DELAYS OF THE INPUT-TARGET CROSS-CORRELATION
FUNCTION
2. SIGNIFICANT DELAYS OF THE TARGET AUTO-CORRELATION FUNCTION
3. If you use the NN Toolbox function NNCORR
help NNCORR
doc NNCORR
It will be worthwhile to see some of my NEWSGROUP and ANSWERS
posts. Search with
greg nncorr
Hope this helps.
Thank you for formally accepting my answer
Greg

More Answers (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!