Behavioural Analalysis

Aim

The automated extraction of writer’s attitude from the text

DataSet

We took a Hotel review dataset with three levels of emotions tagged(Positive , Negative, Neutral) for our model. 1-Negative 2-Neutral 3-Negative

Pre-processing

Pre-processing involves

• Converting Strings in to Tidy text .
• Cleaning the data like removing punctuation,removing stopwords,stemming and Sparsing the document. Process involve converting the data to Document Term Matrix

Frequency Analysis

• Frequency analysis without Cleaning the data

chennai_reviews<-read.csv("chennai_reviews.csv",header=T)
text<-data_frame(Line=1:nrow(chennai_reviews),Text=chennai_reviews$Review_Text)
text$Text<-as.character(text$Text)
text<-text%>%unnest_tokens(word,Text)
Frequeny<-text%>%count(word,sort=T)
Frequencytop20<-Frequeny[1:20,]
plot<-ggplot(Frequencytop20,aes(x=reorder(word,-n),y=n,fill=word))+geom_bar(stat="identity")

• Frequency analysis after Cleaning the data

corpus<-VCorpus(VectorSource(chennai_reviews$Review_Text))
corpus = tm_map(corpus, content_transformer(tolower))
corpus = tm_map(corpus, removePunctuation)
corpus = tm_map(corpus, removeWords, stopwords("english"))
corpus = tm_map(corpus, stemDocument)
dtm<-DocumentTermMatrix(corpus)
dtm<-removeSparseTerms(dtm,.97)
dataframe<-as.data.frame(as.matrix(dtm))
dataframe<-rbind(dataframe,colSums(dataframe))
dataframe<-data.frame(Word=colnames(dataframe),Freq=as.numeric(as.vector(dataframe[nrow(dataframe),])))
dataframe<-dataframe[order(dataframe$Freq,decreasing = T),]
dataframetop20<-dataframe[1:20,]
plot1<-ggplot(dataframetop20,aes(x=reorder(Word,-Freq),y=Freq,fill=Word))+geom_bar(stat="identity")

Sentiment Libraray Method

Mataching our data to Sentiment Liraray.

Model1

emotion_library("Its really nice place to stay especially for business and tourist purpose.")

emotion_library("Table and chair not clean. Not value for money. PC lan cable has so much dust but WiFi Internet speed is too good. Bar service is too bad only one waiter serving to all customer. Serving food like tandoori chicken is very salty.")

Model2(Improved Model)

emotion_library("Table and chair not clean. Not value for money. PC lan cable has so much dust but WiFi Internet speed is too good. Bar service is too bad only one waiter serving to all customer. Serving food like tandoori chicken is very salty.")

Naive-Bayes Model

Splitting the data taking n-grams to be 2 ,Training the data and predicting.

analysis("It was a worst day")

SVM Model1

• Taking the "1" and "2" as "not 3"
• Input

 svmanalysis("The staff of the hotel were polite.  The brick oven chefs were extremely helpful and all the meals were very good.  The room was as advertised and clean.  My stay of 2 days was very comfortable and I would recommend this hotel  to others.")

• Output

SVM_LABEL  SVM_PROB
1         3 0.9523517

• Input

svmanalysis("It was a comfortable stay and I liked the hotel srvices")

• Output

SVM_LABEL SVM_PROB
1         3 0.813309

SVM Model 2

• Input

svmanalysis("Table and chair not clean. Not value for money. PC lan cable has so much dust but WiFi Internet speed is too good. Bar service is too bad only one waiter serving to all customer. Serving food like tandoori chicken is very salty.")

• Output

 SVM_LABEL  SVM_PROB
1         2 0.8455984