[R] neural net

arun smartpink111 at yahoo.com
Fri Dec 14 02:39:58 CET 2012


Hi,
I tried your dataset.  I couldn't reproduce the Error: message.  Instead,

mydata<-read.table(text="
drug param1 param2 param3 param4 param5 class
A 111 15 125 40 0.5 1
B 347 13 280 55 3 2
C 335 9 119 89 -40 1
D 477 37 75 2 0 1
E 863 24 180 10 5 2
F 737 28 150 15 6 2
G 390 63 167 12 0 3
H 209 93 200 48 45 3
I 376 72 201 45 -60 3
J 262 16 205 49 25 3
K 273 39 267 53 11 1
L 192 33 164 19 15 2
M 282 2 213 86 30 1
N 111 11 198 68 -21 1
O 387 20 143 12 16 2
P 674 15 78 -20 -17 2
R 734 54 140 24 7 2
S 272 46 159 57 28 2
T 245 37 90 6 31 2
",sep="",header=TRUE,stringsAsFactors=TRUE)

library(neuralnet)
nn <- neuralnet(
  class~param1+param2+param3+param4+param5+param5,  #param5 is duplicated(typo?)
  data=mydata, hidden=2, err.fct="ce",
  linear.output=FALSE)
#Warning message:
#'err.fct' was automatically set to sum of squared error (sse), because the response is not binary 


nn
#Call: neuralnet(formula = class ~ param1 + param2 + param3 + param4 +     param5 + param5, data = mydata, hidden = 2, err.fct = "ce",     linear.output = FALSE)
#
#1 repetition was calculated.
#
  #      Error Reached Threshold Steps
#1 12.50980687    0.009804371657    30
plot(nn)

sessionInfo()
R version 2.15.0 (2012-03-30)
Platform: x86_64-pc-linux-gnu (64-bit)

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=C                 LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       

attached base packages:
[1] grid      stats     graphics  grDevices utils     datasets  methods  
[8] base     

other attached packages:
[1] neuralnet_1.31 MASS_7.3-16    stringr_0.6    reshape_0.8.4  plyr_1.7.1    

loaded via a namespace (and not attached):
[1] tools_2.15.0
A.K.



----- Original Message -----
From: Katarzyna Nurzynska <paxkn at nottingham.ac.uk>
To: PIKAL Petr <petr.pikal at precheza.cz>; "r-help at r-project.org" <r-help at r-project.org>
Cc: 
Sent: Thursday, December 13, 2012 10:56 AM
Subject: Re: [R] neural net

Hi
Thanks for your reply. I have compared my data with some other which works and I cannot see the difference... 

The structure of my data is shown below: 

> str(data) 
'data.frame':   19 obs. of  7 variables: 
$ drug    : Factor w/ 19 levels "A","B","C","D",..: 1 2 3 4 5 6 7 8 9 10 ... 
$ param1  : int  111 347 335 477 863 737 390 209 376 262 ... 
$ param2 : int  15 13 9 37 24 28 63 93 72 16 ... 
$ param3     : int  125 280 119 75 180 150 167 200 201 205 ... 
$ param4     : int  40 55 89 2 10 15 12 48 45 49 ... 
$ param5     : num  0.5 3 -40 0 5 6 0 45 -60 25 ... 
$ Class   : int  1 2 1 1 2 2 3 3 3 3 ... 

> summary(data) 
      drug        param1         param2         param3             param4              param5             Class      
A      : 1   Min.   :111.0   Min.   : 2.0   Min.   : 75.0   Min.   :-20.00   Min.   :-60.000   Min.   :1.000  
B      : 1   1st Qu.:253.5   1st Qu.:15.0   1st Qu.:132.5   1st Qu.: 12.00   1st Qu.:  0.000   1st Qu.:1.000  
C      : 1   Median :335.0   Median :28.0   Median :164.0   Median : 40.00   Median :  6.000   Median :2.000  
D      : 1   Mean   :383.0   Mean   :33.0   Mean   :166.0   Mean   : 35.26   Mean   :  4.447   Mean   :1.895  
E      : 1   3rd Qu.:433.5   3rd Qu.:42.5   3rd Qu.:200.5   3rd Qu.: 54.00   3rd Qu.: 20.500   3rd Qu.:2.000  
F      : 1   Max.   :863.0   Max.   :93.0   Max.   :280.0   Max.   : 89.00   Max.   : 45.000   Max.   :3.000  
(Other):13                                                            

The structure of the example data which worked is shown below: 

> str(infert) 
'data.frame':   248 obs. of  8 variables: 
$ education     : Factor w/ 3 levels "0-5yrs","6-11yrs",..: 1 1 1 1 2 2 2 2 2 2 ... 
$ age           : num  26 42 39 34 35 36 23 32 21 28 ... 
$ parity        : num  6 1 6 4 3 4 1 2 1 2 ... 
$ induced       : num  1 1 2 2 1 2 0 0 0 0 ... 
$ case          : num  1 1 1 1 1 1 1 1 1 1 ... 
$ spontaneous   : num  2 0 0 0 1 1 0 0 1 0 ... 
$ stratum       : int  1 2 3 4 5 6 7 8 9 10 ... 
$ pooled.stratum: num  3 1 4 2 32 36 6 22 5 19 ... 

> summary(infert) 
   education        age            parity         induced            case         spontaneous        stratum      pooled.stratum 
0-5yrs : 12   Min.   :21.00   Min.   :1.000   Min.   :0.0000   Min.   :0.0000   Min.   :0.0000   Min.   : 1.00   Min.   : 1.00  
6-11yrs:120   1st Qu.:28.00   1st Qu.:1.000   1st Qu.:0.0000   1st Qu.:0.0000   1st Qu.:0.0000   1st Qu.:21.00   1st Qu.:19.00  
12+ yrs:116   Median :31.00   Median :2.000   Median :0.0000   Median :0.0000   Median :0.0000   Median :42.00   Median :36.00  
               Mean   :31.50   Mean   :2.093   Mean   :0.5726   Mean   :0.3347   Mean   :0.5766   Mean   :41.87   Mean   :33.58  
               3rd Qu.:35.25   3rd Qu.:3.000   3rd Qu.:1.0000   3rd Qu.:1.0000   3rd Qu.:1.0000   3rd Qu.:62.25   3rd Qu.:48.25  
               Max.   :44.00   Max.   :6.000   Max.   :2.0000   Max.   :1.0000   Max.   :2.0000   Max.   :83.00   Max.   :63.00  

So still not sure how to solve the problem .... 






_________
From: PIKAL Petr [petr.pikal at precheza.cz]
Sent: 13 December 2012 07:16
To: dada; r-help at r-project.org
Subject: RE: [R] neural net

Hi

> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-
> project.org] On Behalf Of dada
> Sent: Thursday, December 13, 2012 12:41 AM
> To: r-help at r-project.org
> Subject: [R] neural net
>
> Hi
> I would like to do neural netowrk analysis on my data. It look like
> this:
>
> drug  param1  param2  param3  param4  param5  class
> A     111     15      125     40      0.5     1
> B     347     13      280     55      3       2
> C     335     9       119     89      -40     1
> D     477     37      75      2       0       1
> E     863     24      180     10      5       2
> F     737     28      150     15      6       2
> G     390     63      167     12      0       3
> H     209     93      200     48      45      3
> I     376     72      201     45      -60     3
> J     262     16      205     49      25      3
> K     273     39      267     53      11      1
> L     192     33      164     19      15      2
> M     282     2       213     86      30      1
> N     111     11      198     68      -21     1
> O     387     20      143     12      16      2
> P     674     15      78      -20     -17     2
> R     734     54      140     24      7       2
> S     272     46      159     57      28      2
> T     245     37      90      6       31      2
>
> I have entered the code below:
> > nn <- neuralnet(
> + class~param1+param2+param3+param4+param5+param5,
> + data=mydata, hidden=2, err.fct="ce",
> + linear.output=FALSE)
>
> However the error appeared:
> Error in model.frame.default(formula.reverse, data) :
>   object is not a matrix
>
> I changed the data frame to matrix:
> mydata.mat=as.matrix(mydata)

This is not very wise. It changes all numeric values to character. From documentation and your data frame there is nothing obviously wrong. However you did not provide info about structure of your data something like

summary(mydata) or str(mydata).

Documentation does not say much about how neuralnet reacts e.g. on NA values. The best way how to proceed seems to try recommended data and compare them to your data to see where are the differences.

Or you can read sources to see where the error message originates.

Regards
Petr

>
> but still the same error appears...
>
> How can I fix this ? I would be very grateful for your help  !
>
>
>
> --
> View this message in context: http://r.789695.n4.nabble.com/neural-net-
> tp4652927.html
> Sent from the R help mailing list archive at Nabble.com.
>
> ______________________________________________
> R-help at r-project.org mailing list
> 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.This message and any attachment are intended solely for the addressee and may contain confidential information. If you have received this message in error, please send it back to me, and immediately delete it.   Please do not use, copy or disclose the information contained in this message or in any attachment.  Any views or opinions expressed by the author of this email do not necessarily reflect the views of the University of Nottingham.

This message has been checked for viruses but the contents of an attachment
may still contain software viruses which could damage your computer system:
you are advised to perform your own checks. Email communications with the
University of Nottingham may be monitored as permitted by UK legislation.

______________________________________________
R-help at r-project.org mailing list
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.





More information about the R-help mailing list