after training networ not have a good results

1 view (last 30 days)
muhammed kareem
muhammed kareem on 1 Nov 2014
Commented: muhammed kareem on 6 Nov 2014
Dear all I make a network with 5 input and 1 target everything is ok and my r2 is 0.92 but when I make sim to new data to predict the result is too bad and r2 is 0.3 please help

Sign in to answer this question.

Accepted Answer

Greg Heath
Greg Heath on 2 Nov 2014
The new and original data should be consistent with samples from the same probability distribution.
1. How much data do you have?
2. Do the two data sets have the same source?
3. Are the 5 new target vs input data plots close to the 5 old data graphs?
4. Do the the two data sets have the same mean and covariance matrix?
5. Are you using FITNET or the obsolete NEWFIT?
6. How many hidden nodes?
7. code?
Hope this helps.
Thank you for formally accepting my answer
Greg
  1 Comment
muhammed kareem
muhammed kareem on 6 Nov 2014
Thanks for replay The new data is same source with training data I use both fit and newff I changed hidden node many time but the calculation is 20 My data is 257 point with 5 element also I make repmat to increase the data it look good in training but bad in simlution
Dear friend I will try cluster analysis for my data and if this not help I will ask you again

Sign in to comment.

More Answers (0)

Categories

Find more on Sequence and Numeric Feature Data Workflows in Help Center and File Exchange

Tags

Products

Community Treasure Hunt

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

Start Hunting!