Builtin type problem with program.

General
, , ,
1

Hello, I have a problem with this program.

library(ggplot2)
library(magrittr)
library(dplyr)
library(tidytext)
library(gutenbergr)
library(tm)
library(fields)
library(tidyr)
library(igraph)
library(reshape2)
library(wordcloud)
library(MASS)
library(e1071)

data.bbc <- as_tibble(read.table("something", stringsAsFactors = F))
for(i in seq(1, 8867, 1))
{
  if (data.bbc[i,2] == -1) data.bbc[i,2] <- 1
}
data.bbc %<>% arrange(desc(emo)) %>% slice(-c(500:(n()-500)))
data.bbc <- data.bbc[sample(1:nrow(data.bbc)),]
data.bbc$doc_id <- 1:nrow(data.bbc)
source <- DataframeSource(as.data.frame(data.bbc))
corpus <- VCorpus(source)
tdm <- DocumentTermMatrix(corpus)
bbc.svml <- svm(tdm, class, type = "C-classification", kernel = "linear")
bbc.svml.pred <- predict(bbc.svml, tdm)
table(class, bbc.svml.pred)

Every time I try to run this I have errors like this.

bbc.svml <- svm(tdm, class, type = "C-classification", kernel = "linear")
Błąd w poleceniu 'complete.cases(object)':niepoprawny 'type' (builtin) argumentu
bbc.svml.pred <- predict(bbc.svml, tdm)
Błąd w poleceniu 'predict(bbc.svml, tdm)':nie znaleziono obiektu 'bbc.svml'
table(class, bbc.svml.pred)
Błąd w poleceniu 'table(class, bbc.svml.pred)':
nie znaleziono obiektu 'bbc.svml.pred'

I 'll be thankful, if anyone 'd help me with it.

2

Hi,
It is tough to help you because we can not reproduce your error - using a reproducible example would help (see FAQ: What's a reproducible example (reprex) and how do I do one? - meta - RStudio Community).

However, my guess would be that svm() is running complete.cases() and the object is not a vector. Have a look at your data - is it in the format you expect it to be (is it completely empty?)?

Matt

3

I have edited my post. Is this what you needed?

4

That's really helpful. What are you trying at achieve with the svm() here? What is the "class" variable and where does that come from?

5

I wanted to compare objective class (emo = 0) with subiective class (emo = 1 or -1) using SVM method.

6

Okay. I think the problem is that you haven't told svm what the class is. I am not sure if svm accepts tdm's but perhaps this would work for you?

#Get your data
data.bbc <- as_tibble(read.table("https://jsienkiewicz.pl/TEXT/lab/data_bbc.csv", stringsAsFactors = F))
for(i in seq(1, 8867, 1))
{
  if (data.bbc[i,2] == -1) data.bbc[i,2] <- 1
}
data.bbc %<>% arrange(desc(emo)) %>% slice(-c(500:(n()-500)))
data.bbc <- data.bbc[sample(1:nrow(data.bbc)),]
data.bbc$doc_id <- 1:nrow(data.bbc)
# Create the document term matrix (a dtm)
library(RTextTools)
dtMatrix <- create_matrix(data.bbc["text"])
container <- create_container(dtMatrix, data.bbc$emo, trainSize=1:11, virgin=FALSE)
# train a SVM Model
model <- train_model(container, "SVM", kernel="linear", cost=1)
7

Sorry, but it didn't help. But meanwhile in somewhere else I 've found solution. This code does what I wanted without any errors.

library(ggplot2)
library(magrittr)
library(dplyr)
library(tidytext)
library(gutenbergr)
library(tm)
library(fields)
library(tidyr)
library(igraph)
library(reshape2)
library(wordcloud)
library(MASS)
library(e1071)
library(RTextTools)

data.bbc <- as_tibble(read.table("something", stringsAsFactors = F))
for(i in seq(1, 8867, 1))
{
  if (data.bbc[i,2] == -1) data.bbc[i,2] <- 1
}
data.bbc %<>% arrange(desc(emo)) %>% slice(-c(500:(n()-500)))
data.bbc <- data.bbc[sample(1:nrow(data.bbc)),]
data.bbc$doc_id <- 1:nrow(data.bbc)
source <- DataframeSource(as.data.frame(data.bbc))
corpus <- VCorpus(source)
tdm <- DocumentTermMatrix(corpus)
tdm <- as.matrix(tdm)
ind <- apply(tdm, 1, sum) > 1
tdm <- tdm[ind, ]
class <- data.bbc$emo[ind]
bbc.svml <- svm(tdm, class, type = "C-classification", kernel = "linear")
bbc.svml.pred <- predict(bbc.svml, tdm)
table(class, bbc.svml.pred)