[R] Help needed with Traininig a Neural Network

Bonita Willams bonitajo at gmail.com
Mon Apr 27 04:55:38 CEST 2015


I am attempting to train a dataset but am having a hard time. I am using a
dataset from UCI *archive.ics.uci.edu/ml/datasets/Congressional+Voting+Records
<http://archive.ics.uci.edu/ml/datasets/Congressional+Voting+Records>  *. I
am attempting to replicate a study which predicts the political party of
Congress members based on their voting records.

Below is some of what I have accomplished.

>dataset <- read.csv(“ucidatasethouse.vote.84.data”)

> trainset <- dataset[1:305,]

>testset <- dataset[306:435, ]



(1.) *I trained the neural net model*

> polpartynet <- neuralnet(Party ~ HandInfants + WaterProject + AdoptBudget
+ DocFeeFreeze + ElSalvadorAid + ReligiousGroupsSchools + AntiSatellTestBan
+ AidNicaraguaContras + MXMissile + Immigration + SynCorpCutback +
EducationSpending + SuperfundRighttoSue + Crime + DutyFreeExports +
ExportAdminSouthAfrica, trainset, hidden = 4, lifesign = "minimal",
linear.output = FALSE, threshold = 0.1)

hidden: 4    thresh: 0.1    rep: 1/1    steps:      16       error:
58.57427 time: 0.16 secs





(2.) *I put this together but not sure of what I was supposed to get*

> polpartynettestset.results <- compute(polpartynet, testset)







(3.) * The Training set which contains all of the columns is ‘trainset’



> colnames(trainset)

 [1] "Party"                  "HandInfants"            "WaterProject"

 [4] "AdoptBudget"            "DocFeeFreeze"           "ElSalvadorAid"

 [7] "ReligiousGroupsSchools" "AntiSatellTestBan"      "AidNicaraguaContras"

[10] "MXMissile"              "Immigration"            "SynCorpCutback"

[13] "EducationSpending"      "SuperfundRighttoSue"    "Crime"

[16] "DutyFreeExports"        "ExportAdminSouthAfrica"



(4.) *I removed “Party” column from the testset based but this may
have been a bad move?*** What should I do????



> colnames(testset)

 [1] "HandInfants"            "WaterProject"           "AdoptBudget"

 [4] "DocFeeFreeze"           "ElSalvadorAid"          "ReligiousGroupsSchools"

 [7] "AntiSatellTestBan"      "AidNicaraguaContras"    "MXMissile"

[10] "Immigration"            "SynCorpCutback"         "EducationSpending"

[13] "SuperfundRighttoSue"    "Crime"                  "DutyFreeExports"

[16] "ExportAdminSouthAfrica"



(5.) ***I would like to create a formula which will provide me neural
network results***



> results <- data.frame(actual = testset$Party, prediction = polpartynettestset.results)

Error in data.frame(actual = testset$Party, prediction =
polpartynettestset.results) :

  arguments imply differing number of rows: 0, 130



(6.) I would like to be able to round to the nearest integer to
improve readability. This is what I have tried so far…



> results$Party <- round(results$Party)

Error: object 'results' not found

> results[306:435]

Error: object 'results' not found

> polpartynettestset.results$Party <- round(polpartynettestset.results$Party)

Error in round(polpartynettestset.results$Party) :

  non-numeric argument to mathematical function





I appreciate any help that you can provide. It is possible that I am
missing something but will happily add it if you ask. I feel like a
dog chasing its tail.





 Bonita Williams

bonitajo at gmail.com

	[[alternative HTML version deleted]]



More information about the R-help mailing list