Error when plotting

[basaldella]

Hello,

I have a strange error when trying to use plotnet. Let's assume I have two models, called m1 and m2, which are trained using the exact same formula on two different datasets.

If I do

> plotnet(m1,alpha=0.6)

Everything works fine.

But if run

> plotnet(m2,alpha=0.6)

What i get is

Error in as.character(forms) : 
  cannot coerce type 'closure' to vector of type 'character'

Do you have any idea of what could cause this problem?

[fawda123]

Can you explain the models and input data a bit more? I can't diagnose anything from the example.

[basaldella]

Yeah, I'm sorry, you're right. I created a gist with the example code and data, I hope this helps.

In the code, as you can see, the formulae are the equal, the datasets (generated with the same tool) are very similar, but somehow plotnet(akecModel) fails, while plotnet(semevalModel) works. The same happens when I try to run garson and lekprofile as well.

[fawda123]

You need to rename the formula object to something else. formula is already a function in R and that's causing problems in the NeuralNetTools functions. This should work:

library(nnet)
library(NeuralNetTools)

formulaA = AKEC ~ 
  tf +
  Freq_Sentence + 
  Height + 
  LifeSpan + 
  Depth +
  tf.idf + 
  Maximality

formulaS = SEMEVAL2010 ~ 
  tf +
  Freq_Sentence + 
  Height + 
  LifeSpan + 
  Depth +
  tf.idf + 
  Maximality

AKEC_train <- read.csv("https://gist.githubusercontent.com/basaldella/ebb672efbb12c565dec585a6ed0916b4/raw/b9dc49dffa6e8caa758cacd0a5acf8148b73c89e/akec.top.csv",stringsAsFactors = TRUE)
SEMEVAL_train <- read.csv("https://gist.githubusercontent.com/basaldella/ebb672efbb12c565dec585a6ed0916b4/raw/b9dc49dffa6e8caa758cacd0a5acf8148b73c89e/semeval.top.csv",stringsAsFactors = TRUE)

set.seed(1203) 
akecModel = nnet(formulaA,
                 size = length(attr(terms(formulaA),"variables")) * 2,
                 entropy=T,data = AKEC_train)

set.seed(1203) 
semevalModel = nnet(formulaS,
                    size = length(attr(terms(formulaS),"variables")) * 2,
                    entropy=T,data = SEMEVAL_train)

plotnet(akecModel)
plotnet(semevalModel)

[basaldella]

Now it works like a charm. Sorry for my mistake, but I'm far from being an R expert... Thanks a lot for your help!

[fawda123]

No problem, I'm happy to help.