R: spam classification

install.packages("quanteda")
install.packages("RColorBrewer")

library(quanteda)
library(RColorBrewer)

setwd("~/Your Folder Path")
raw.data <- read.table("SMSSpamCollection", header=FALSE, sep="\t", quote="", stringsAsFactors=FALSE) 

names(raw.data) <- c("Label", "Text")
table(raw.data$Label)

set.seed(1912)
raw.data <- raw.data[sample(nrow(raw.data)),]

sms.corpus <- corpus(raw.data$Text) 
docvars(sms.corpus) <- raw.data$Label 

spam.plot <- corpus_subset(sms.corpus, docvar1 == "spam")
spam.plot <- dfm(spam.plot, tolower = TRUE, removePunct = TRUE, removeTwitter = TRUE, removeNumbers = TRUE, remove=stopwords("SMART"))

spam.col <- brewer.pal(10, "BrBG")
spam.cloud <- textplot_wordcloud(spam.plot, min.freq = 16, color = spam.col)
title("Spam Wordcloud", col.main = "grey14")

ham.plot <- corpus_subset(sms.corpus, docvar1 == "ham")
ham.plot <- dfm(ham.plot, tolower = TRUE, removePunct = TRUE, removeTwitter = TRUE, removeNumbers = TRUE, remove=c("gt", "lt",  stopwords("SMART")))
ham.col <- brewer.pal(10, "BrBG")
textplot_wordcloud(ham.plot, min.freq = 50, colors = ham.col, fixed.asp = TRUE)
title("Ham Wordcloud", col.main = "grey14")

sms.dfm <- dfm(sms.corpus, tolower = TRUE)
sms.dfm <- dfm_trim(sms.dfm, min_count = 5, min_docfreq = 3)
sms.dfm <- dfm_weight(sms.dfm, type = "tfidf")

sms.raw.train <- raw.data[1:4738,]
sms.raw.test <- raw.data[4739:nrow(raw.data),]

sms.dfm.train <- sms.dfm[1:4738,]
sms.dfm.test <- sms.dfm[4739:nrow(raw.data),]

sms.classifier <- textmodel_NB(sms.dfm.train, sms.raw.train$Label)

sms.predictions <- predict(sms.classifier, newdata = sms.dfm.test)
table(sms.predictions$nb.predicted, sms.raw.test$Label)

Referensi

• https://blog.paperspace.com/intro-to-datascience/
• https://github.com/subhash-karthik/SMS-spam-classifier/blob/master/Session1.md

Pranala Menarik

• R