ensemble

################################--SER 599 Thesis--###############################
#-------------------------------------------------------------------------------#
#--Version 2.0/2.1 : Japa Swadia--#
#--Classification of requirements statements using ensemble methods and RTextTools--#

#Load data set for training
#Read csv file containing Requirement statement and Requirement Category columns
load_train <- read.csv("C:/Users/japas_000/Documents/Thesis/model/train_data_big.csv", header = TRUE, nrows = 1495)

#Prepare data frame for training
train_df <- as.data.frame(load_train[ , -1])

#Install and load tm package for text mining
library(tm)

#Load required packages
library(class) # KNN model
library(SnowballC) # Stemming words
library(e1071) # SVM and ensemble classification
library(ggplot2) # Plotting data

#Install and load RTextTools package
library(RTextTools)

# Create matrix of text terms and preprocess the statements
doc_matrix <- create_matrix(train_df$Requirement.Statement, language="english", removeNumbers=TRUE,
                            stemWords=TRUE)

# Create container to hold the matrix, Category column and specify training and test data size
container <- create_container(doc_matrix, as.numeric(factor(load_train$Requirement.Type)), trainSize=2:754, testSize=755:1495, virgin=FALSE)

# Apply ensemble learning for training model
svm.pred <- train_model(container,"SVM") #apply Support Vector Machine algorithm
#glmnet.pred <- train_model(container, "GLMNET") # apply Generalized Linear Model Network algorithm
maxent.pred <- train_model(container, "MAXENT") # apply Maximum Entropy algorithm
boosting.pred <- train_model(container, "BOOSTING") # apply Boosting algorithm
#bagging.pred <- train_model(container, "BAGGING") # apply Bagging algorithm
rf.pred <- train_model(container, "RF") # apply Random Forests algorithm
#nnet.pred <- train_model(container, "NNET") # apply Neural Network algorithm
#tree.pred <- train_model(container, "TREE") # apply Decision Tree algorithm

# Now classify test data using trained ensemble models
svm.classify <- classify_model(container, svm.pred) # using Support Vector Machine algorithm
glmnet.classify <- classify_model(container, glmnet.pred) # using Generalized Linear Model Network algorithm
maxent.classify <- classify_model(container, maxent.pred) # using Maximum Entropy algorithm
boosting.classify <- classify_model(container, boosting.pred) # using Boosting algorithm
bagging.classify <- classify_model(container, bagging.pred) # using Bagging algorithm
rf.classify <- classify_model(container, rf.pred) # using Random Forests algorithm
nnet.classify <- classify_model(container, nnet.pred) # using Neural Network algorithm
tree.classify <- classify_model(container, tree.pred) # using Decision Tree algorithm

# Cross-validation
SVM <- cross_validate(container, 5, "SVM") 
GLMNET <- cross_validate(container, 5, "GLMNET") 
MAXENT <- cross_validate(container, 5, "MAXENT") 
BAGGING <- cross_validate(container, 5, "BAGGING") 
BOOSTING <- cross_validate(container, 5, "BOOSTING")
RF <- cross_validate(container, 5, "RF") 
NNET <- cross_validate(container, 5, "NNET") 
TREE <- cross_validate(container, 5, "TREE")

# Create analytics from ensemble methods
analytics <- create_analytics(container, cbind(svm.classify, glmnet.classify, maxent.classify, boosting.classify, 
                                               bagging.classify, rf.classify, nnet.classify, tree.classify))

# Get summary of analytics
summary(analytics)

# Get summaries and comparison of different algorithms
topic_summary <- analytics@label_summary
alg_summary <- analytics@algorithm_summary
ens_summary <-analytics@ensemble_summary
doc_summary <- analytics@document_summary

# Produce csv files of predicted output
write.csv(topic_summary, file="C:/Users/japas_000/Documents/Thesis/model/label_summary.csv")
write.csv(alg_summary, file="C:/Users/japas_000/Documents/Thesis/model/algo_summary.csv")
write.csv(ens_summary, file="C:/Users/japas_000/Documents/Thesis/model/ensemble_summary.csv")
write.csv(doc_summary, file="C:/Users/japas_000/Documents/Thesis/model/doc_summary3.csv")