You model is reaching the error threshold otherwise you would be receiving
an 'actual' error message. You model is just converging very quickly.
If
you want to see the error reducing, change lifesign="minimal" to
lifesign="full" and set the lifesign.step=1 to make it very verbose
for
this model.
Alternatively you could just look at winequal$result.matrix and you will
see how many steps were taken to reach the threshold as well as the
reached.threshold your model finished on.
Cheers,
Charles
On Tue, Jun 2, 2015 at 2:59 PM, ravishankar narayanan <ravsh98 at
gmail.com>
wrote:
> Hi,
>
> Developing a Neural Network in R to predict the quality of wine.Attached is
> the Wine Data Set.Tried twice once by selecting specific features and once
> by using all features to predict quality of wine
>
> Each time I run the Neural Network I am getting a huge error.Tried using
> feature selection and also used all features
> *With Feature Selection*
>
> *winequal<-
>
>
neuralnet(quality~volatile.acidity+citric.acid+sulphates+alcohol,winetrain,hidden=2,lifesign="minimal",linear.output=FALSE,threshold=0.1)*
>
>
>
>
>
> *hidden:2 thresh:0.1 rep:1/1 steps: 52 error:
> 8486.01111 time: 2.41secs*
>
> *No feature selection*
>
>
>
*winequal1<-neuralnet(quality~fixed.acidity+volatile.acidity+citric.acid+residual.sugar+chlorides+free.sulfur.dioxide+total.sulfur.dioxide+density+pH*
>
>
>
>
>
>
*++sulphates+alcohol,winetrain1,hidden=4,lifesign="minimal",linear.output=FALSE,threshold=0.1)*
>
> Model Produced:
>
>
>
>
>
> *hidden: 4 thresh:0.1 rep: 1/1 steps: 26 error: 8486.08992
> time: 0.1 secs*
>
> Is there any method I can use to reduce the error ? Changing the threshold
> or the number of hidden layers does not help.Any tips will be
> helpful.Thanks.
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
> http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>
[[alternative HTML version deleted]]