Is it Possible to extract Regression equation of the Regression Plots in Neural Network Toolbar

3 views (last 30 days)
Hi, I am new in using Neural Network tool bar , I have a difficulty. Eg : I have a Input Data set of 30*52 and my target data set is 30*16 . I train it and i get Regression plots. Once i get the regression plots is it possible to extract the equation from the plots. As it is specific to use Neural Network and execute the work i am badly needing to extract the data out of the Figure to a excel sheet. Kindly help me with the above Problem.

Accepted Answer

Greg Heath
Greg Heath on 20 Nov 2012
help regression
doc regression
type regression
should suffice. However, can also look at
help plotregression
doc plotregression
type plotregression
Hope this helps.
Thank you for formally accepting my answer
Greg

More Answers (4)

Vijayaraghavan
Vijayaraghavan on 20 Nov 2012
Thanks Greg . But my problem is little more complicated now as am trying to establish a Multivariate equation with nearly 20 to 25 variable and is it possible to achieve the same using Neural Network Tool bar .Since i feed in Multiple Inputs for training and i get Regression equations in the form x1 input to target = output and so.

Greg Heath
Greg Heath on 20 Nov 2012
i am surprised and confused by your last setence.
The regression equation associated with the training function is output vs target, and is only valid for 1-D targets and outputs. Inputs are not involved.
The nonlinear I/O multivariate regression equation for the default fitnet or feedforwardnet is
y = b1 + LW*tansig(b1 + IW*x)
which cannot be decomposed into a sum of fuctions of single input variables.
Greg

Vijayaraghavan
Vijayaraghavan on 20 Nov 2012
Sorry Greg that was a type error I got Output = 1.5505 target + 0.115 something of this sort. But My input set is like a1,a2,a3,a4... an and my o/p set b1,b2,b3...bn say I need a multivariate eqn using the neural network tool bar and it has to be in the form of a1(X)+a2(X)... = o/p is it possible using the neural network toolbar right away by training it by usual method or should i make any modifications to get the code to make it in my required form. Pls help me on this .
  1 Comment
Greg Heath
Greg Heath on 20 Nov 2012
I do not understand exactly what you want. The I/O equation I wrote is the default. Please write exactly what you are looking for in a similar form.
The general form is
y = activation2( b2 + LW * activation1( b1 + IW * x ) )
where
size(x) = [I N ]
size(IW) = [ H I ]
size(b1) = [ H N ] %repmat([H 1],1,N)
size(LW) = [ O H ]
size(b2) = [ O N ] %repmat([O 1],1,N)

Sign in to comment.


Imran Babar
Imran Babar on 16 May 2013
Hi I have a data of 649 patterns and with 31 inputs in 1 pattern and with 1 output only how I will draw the regression plot. Though I am using the following syntax but the graph is not a straight line
plotregression(targetdata,predicted,'Regression')
and my target data is like the following one
0 1 2 3 4 5 6 7 2 3 4 5 6 7 8 9 4 5 6 7 8 9 7 8 9 10 11 12 13 14 and so on upto 649 values for each input pattern of 31 values

Categories

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

Community Treasure Hunt

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

Start Hunting!