Hi all,
I have considered neural network to classify the health status of the cow.
I found a very neatly written R codes for classification method in
here<http://home.strw.leidenuniv.nl/~jarle/IAC/RRoutines/classification-example.R>
.
It would be very helpful if you can answer some of the questions, that I am
struggling with,
I have set of time series data from different animals, I use data on 10
animals to train the model and 1 animal to test the model at a time.
Below is the simple Rcode , I used before complicating myself. But the
prediction resulted in classifying absolute 0 and 1, even though I have
asked results in integer value(#round()).
- In section 6.2 of neural network and related method for
classification, by B.D.
Ripley<http://www.jstor.org/stable/2346118?origin=JSTOR-pdf>it is
specified as below.
[image: Inline image 9]
I tried to transfer my data in the range [0,1] before feeding them to
‘nnet’. But it does not seemed to improve my result.
- I would like to know, on specifying size of the hidden layer
and decay value.
- How important is it specifying ‘linout=T’ in the function? I
want output in terms probability, so I have not specified it!!!!!!!!
- In Rcode mensioned above, I see ‘r$data’, in all the places, I
just wonder whether training and test data are considered as same!!!
#----------------------------------------------------------------------------------
## training NNET
Nnet <- nnet(x=data, y=class.ind(dat$trt_window),
size=2,rang=0.05,skip=T, decay=0.000001, maxit=500)#
## validation for each lactation
pred <- predict(Nnet, newdata=y, type = "raw")
#-----------------------------------------------------------------------------------
Your help is very much appreciated,
Thanks for your time advance.
[[alternative HTML version deleted]]
