In zahrakhoshmanesh/FinalProjectSTAT585: Sentiment Analysis of Reviews

importing the dataset

source_datasets=read.delim('../inst/Restaurant_Reviews.tsv',quote='',stringsAsFactors = FALSE)
str(source_datasets$Liked)
#Sys.setenv(JAVA_HOME="C:\\Program Files (x86)\\Java\\jdk1.8.0_201\\jre")

tackling imbalanced data

BalanceData<-function(x){
#install.packages("ROSE")
library(ROSE)
data(hacide)
df<-x

#over sampling
data_balanced_over <- ovun.sample(cls ~ ., data = hacide.train, method = "over",N = 1960)$data
table(data_balanced_over$cls)

data_balanced_under <- ovun.sample(cls ~ ., data = hacide.train, method = "under", N = 40, seed = 1)$data
table(data_balanced_under$cls)

data_balanced_both <- ovun.sample(cls ~ ., data = hacide.train, method = "both", p=0.5,N=1000, seed = 1)$data
table(data_balanced_both$cls)

data.rose <- ROSE(cls ~ ., data = hacide.train, seed = 1)$data
table(data.rose$cls)

#check table
table(df$Liked)
data_balanced_over <- ovun.sample(Liked ~ ., data = df, method = "over",N = 1000)$data
table(data_balanced_over$Liked)

data_balanced_under <- ovun.sample(Liked ~ ., data = df, method = "under", N = 500, seed = 1)$data
table(data_balanced_under$Liked)

data_balanced_both <- ovun.sample(Liked ~ ., data = df, method = "both", p=0.5,N=500, seed = 1)$data
table(data_balanced_both$Liked)

data.rose <- ROSE(Liked ~ ., data = df, seed = 1)$data
table(data.rose$Liked)
}

Cleaning the texts

#install.packages('tm')
#install.packages('SnowballC')
library(tm)
corpus=VCorpus(VectorSource(source_datasets$Review))
#convert all review to lower case
corpus= tm_map(corpus,content_transformer(tolower))
# remove numbers from reviews
corpus=tm_map(corpus,removeNumbers)
# remove punctuations from reviews
corpus=tm_map(corpus,removePunctuation)
# remove Stop words from reviews
corpus=tm_map(corpus,removeWords,stopwords())
# Stemming
corpus=tm_map(corpus,stemDocument)
# remove extra space that created in cleaning stage when for example number remove
corpus=tm_map(corpus,stripWhitespace)

Creating the bag of words model

#creating document term matrix of words in reviews
dtm = DocumentTermMatrix(corpus)
# reduce dimention of sparse matrix with considering 99 percent of most frequent
dtm = removeSparseTerms(dtm,0.99)
dtm
#plot(dtm, terms = findFreqTerms(tdm, lowfreq = 6)[1:25], corThreshold = 0.5)


  BigramTokenizer <-  function(x)  unlist(lapply(ngrams(words(x), 2), paste, collapse = " "), use.names = FALSE)

  tdm <- TermDocumentMatrix(corpus, control = list(tokenize = BigramTokenizer))
  tdm
  #inspect(removeSparseTerms(tdm[, 1:10], 0.99))
  # convert matrix of independent variables to data frame
dataset = as.data.frame(as.matrix(dtm))
dim(dataset)

n gram

#creating document term matrix of words in reviews
BigramTokenizer <-  function(x)  unlist(lapply(ngrams(words(x), 2), paste, collapse = " "), use.names = FALSE)

tdm <- TermDocumentMatrix(corpus, control = list(tokenize = BigramTokenizer))
tdm
library(matlib)
ntdm<-t(tdm)
  #inspect(removeSparseTerms(tdm[, 1:10], 0.99))
  # convert matrix of independent variables to data frame
dataset = as.data.frame(as.matrix(ntdm))
dim(dataset)

Building classification

3 classification are popular for text: 1- decision tree 2- naive bayes 3- random forest

preparing the dataset and divide dataset to train and test

# convert matrix of independent variables to data frame
dataset = as.data.frame(as.matrix(dtm))

#dataset$targetclass = source_datasets[-1]
str(source_datasets[[-1]])

#dataset$Liked =  source_datasets$Liked

# encode the target feature as factor
dataset$target = factor(source_datasets[[-1]],level=c(0,1))
str(dataset$target)

# split dataset to training and test set
#install.packages('caTools')
library(caTools)
set.seed(123)
split=sample.split(dataset$target, SplitRatio = 0.8)
training_set=subset(dataset,split==TRUE)
test_set=subset(dataset,split==FALSE)

visualization

#wsTDM <- Data(TermDocMatrix(dtm))
#wsHClust <- hclust(dist(wsTDM), method = "ward")

classifier : Random Forest

# fitting random forest classification to the training set
#install.packages('randomForest')
library(randomForest)
dimention<-dim(dataset)
numdim<-dimention[2]
numdim

classifier = randomForest(x=training_set[-1*numdim],
                          y=training_set$target,
                          ntree=10
                          )
#predicting the test set result
y_pred = predict(classifier,newdata=test_set[-1*numdim])

#making the confusion matrix
cm_randomforest=table(test_set[,numdim],y_pred)
#cm_randomforest
#str(cm_randomforest)
TP=cm_randomforest[1,1]
TP
FP=cm_randomforest[1,2]
FP
FN=cm_randomforest[2,1]
FN
TN=cm_randomforest[2,2]
TN

Accuracy = (TP + TN) / (TP + TN + FP + FN)
Accuracy

Precision = TP / (TP + FP)
Precision

Recall = TP / (TP + FN)
Recall

F1_Score = 2 * Precision * Recall / (Precision + Recall)
F1_Score

Naive Bayes classifier

# Naive Bayes

#install.packages('e1071')
library(e1071)
classifier = naiveBayes(x = training_set[-692],
                        y = training_set$target)

# Predicting the Test set results
y_pred = predict(classifier, newdata = test_set[-692])

# Making the Confusion Matrix
cm_naivebayes = table(test_set[, 692], y_pred)
cm_naivebayes
TP=cm_naivebayes[1,1]
TP
FP=cm_naivebayes[1,2]
FP
FN=cm_naivebayes[2,1]
FN
TN=cm_naivebayes[2,2]
TN

Accuracy = (TP + TN) / (TP + TN + FP + FN)
Accuracy

Precision = TP / (TP + FP)
Precision

Recall = TP / (TP + FN)
Recall

F1_Score = 2 * Precision * Recall / (Precision + Recall)
F1_Score

Decision Tree Classifier

# Decision Tree Classification


# Splitting the dataset into the Training set and Test set
#install.packages('caTools')
library(caTools)
set.seed(123)

# Fitting Decision Tree Classification to the Training set
#install.packages('rpart')
library(rpart)
classifier = rpart(formula = Liked ~ .,
                   data = training_set)

# Predicting the Test set results
y_pred = predict(classifier, newdata = test_set[-692], type = 'class')

# Making the Confusion Matrix
cm_decisiontree = table(test_set[, 692], y_pred)

cm_decisiontree

TP=cm_decisiontree[1,1]
TP
FP=cm_decisiontree[1,2]
FP
FN=cm_decisiontree[2,1]
FN
TN=cm_decisiontree[2,2]
TN

Accuracy = (TP + TN) / (TP + TN + FP + FN)
Accuracy

Precision = TP / (TP + FP)
Precision

Recall = TP / (TP + FN)
Recall

F1_Score = 2 * Precision * Recall / (Precision + Recall)
F1_Score

Visualization

#' Visualize dataset and get some insight of data.
#'
#' @param x dataset.
#' @return some diagram and insight of data
#' @author Zahra Khoshmanesh
#' @export
#' @import tidyverse
#' @examples
#' VisualizeData('./inst/Restaurant_Reviews.tsv')


VisualizeData<-function(dataset){

  library(tidyverse)
  library(tm)
  library(qdap)

  #source_datasets=read.delim(dataset,quote='',stringsAsFactors = FALSE)
  source_datasets=read.delim('./inst/Restaurant_Reviews.tsv',quote='',stringsAsFactors = FALSE)

  #check the number of two class label
  check_class <-table(source_datasets[[2]])
  class_label<-names(source_datasets)[2]
  class_text<-names(source_datasets)[1]
  if (check_class[1]!=check_class[2]){

    print("dataset is imbalance, balance it with calling BalanceData ")

  }  else {
    print("dataset is balanced dataset and no need to balance it")
  }

  corpus=VCorpus(VectorSource(source_datasets[[1]]))
  #convert all review to lower case
  corpus= tm_map(corpus,content_transformer(tolower))
  # remove numbers from reviews
  corpus=tm_map(corpus,removeNumbers)
  # remove punctuations from reviews
  corpus=tm_map(corpus,removePunctuation)
  # remove Stop words from reviews
  corpus=tm_map(corpus,removeWords,stopwords())
  # Stemming
  corpus=tm_map(corpus,stemDocument)
  # remove extra space that created in cleaning stage when for example number remove
  corpus=tm_map(corpus,stripWhitespace)
  myTDM <- TermDocumentMatrix(corpus)
  findFreqTerms(myTDM,lowfreq = 20, highfreq = Inf)



  a <- tolower(source_datasets[[1]])
  a <- removePunctuation(a)
  a <- removeNumbers(a)
  a <-rm_stopwords(a, tm::stopwords("english"))


  freq_term<-freq_terms(a)
  ## S3 method for class 'freq_terms'
  plot(freq_term, plot = TRUE)

  library(tidytext)
  library(dplyr)
  text_df <- tibble(text = source_datasets[[1]])

   tidy_text <- text_df %>%
    unnest_tokens(word, text)

  data(stop_words)
  tidy_text <- tidy_text %>%
    anti_join(stop_words)

  tidy_text  %>%
    count(word, sort = TRUE) 

  library(ggplot2)
  tidy_text %>%
    count(word, sort = TRUE) %>%
    filter(n > 10) %>%
    mutate(word = reorder(word, n)) %>%
    ggplot(aes(word, n)) +
    geom_col() +
    xlab(NULL) +
    coord_flip()
}

zahrakhoshmanesh/FinalProjectSTAT585 documentation built on June 4, 2019, 1:57 p.m.