# The purpose of this code is to predict and assign survival
# values to Titanic passengers with unknown survival statistics
# based on data analysis.

# Set my working directory to one where my files are stored.
setwd("//nlamsvfls07/userdata2$/LWeston/Burberry Info/Data Fellowship")
library(readr)

# Import my test and training datasets and save them in variables
train <- read.csv("~/Burberry Info/Data Fellowship/train.csv")
View(train)
test <- read.csv("~/Burberry Info/Data Fellowship/test.csv")
View(test)

# Check out what data types are included
str(train)

# Convert Pclass and Survived to factors in the training data
train$Pclass <- as.factor(train$Pclass)
train$Survived <- as.factor(train$Survived)

# Convert names to strings in the training data
train$Name <- as.character(train$Name)

# Check out what data types are included in test
str(test)

# Convert Pclass to factors in the test data
test$Pclass <- as.factor(test$Pclass)

# Convert names to strings in the test data
test$Name <- as.character(test$Name)

#Check for duplicate entries
length(unique(train$Name))

# Explore the train data a bit
table(train$Survived)

#Look at proportions
prop.table(table(train$Survived))

# Load ggplot2
library(ggplot2)
ggplot(train, aes(x = Sex, fill = factor(Survived))) +
  geom_bar(width = 0.5) +
  xlab("Sex") +
  ylab("Total count") +
  labs(fill = "Survived")

#Add new column in train data
train$ModelPrediction <- "TBD"
head(train)

#Add value to new column in train data
train$ModelPrediction <- 0
head(train)

#Check structure
str(train$ModelPrediction)

#count survival stats accuracy
sum(train$ModelPrediction == train$Survived)

#count survival stats accuracy percentage
sum(train$ModelPrediction == train$Survived)/nrow(train)

#reset ModelPrediction column to "TBD"
train$ModelPrediction <- "TBD"

#predict that males don't survive
train$ModelPrediction[train$Sex == "male"] <- 0

#predict that females do survive
train$ModelPrediction[train$Sex == "female"] <- 1

#count survival stats accuracy percentage
sum(train$ModelPrediction == train$Survived)/nrow(train)

#introducing a second variable (pclass) to our model's predictions
table(train$Pclass, train$Survived)

#compare the proportions by pclass
prop.table(table(train$Pclass, train$Survived))

#plot Pclass against survival and sex data
ggplot(train, aes(x = Sex, fill = Survived)) +
  geom_bar(width = 0.5) +
  facet_wrap(~Pclass) +
  ggtitle("Pclass") +
  xlab("Sex") +
  ylab("Total count") +
  labs(fill = "Survived")

#Reset ModelPrediction column in train dataset to 0
train$ModelPrediction <- 0
head(train)

#Set that all females survive, unless they are in pclass 3
train$ModelPrediction[train$Sex == "female"] <- 1
train$ModelPrediction[train$Pclass == "3"] <- 0
head(train)

#count survival stats accuracy percentage
sum(train$ModelPrediction == train$Survived)/nrow(train)

#plot to show relationship between Pclass, Sex, Age, and Survival
ggplot(train, aes(x = Age, fill = Survived)) +
  facet_wrap(~Sex + Pclass) +
  geom_bar(width = 10) +
  xlab("Age") +
  ylab("Total count")

#Reset ModelPrediction column in train dataset to 0
train$ModelPrediction <- 0
head(train)

#Set that all females survive, unless they are in pclass 3
train$ModelPrediction[train$Sex == "female"] <- 1
train$ModelPrediction[train$Pclass == "3"] <- 0
head(train)

#predict that all males under 18 in first and second class survive
train$ModelPrediction[(train$Pclass == "1" | train$Pclass == "2") & train$Sex == "male" & train$Age < 18] <- 1
head(train)

#count survival stats accuracy percentage
sum(train$ModelPrediction == train$Survived)/nrow(train)

#create new data frame called "submission" for passenger id data
submission <- data.frame(test$PassengerId)
head(submission)

#rename header as "Passenger Id"
names(submission)[names(submission) == "test.PassengerId"] <- "PassengerId"
head(submission)

#create a column called Survived that assumes everyone perished.
test$Survived <- 0
head(test)

#say all females survive
test$Survived[test$Sex == "female"] <- 1
head(test)

#assume all females in third class perish
test$Survived[test$Pclass == "3"] <- 0

#assume all males in first and second class under 18 survived
test$Survived[(test$Pclass == "1" | test$Pclass == "2") & test$Sex == "male" & test$Age < 18] <- 1
View(test)

#Copy values from ModelPrediction in test dataset into submission under a new column called "Survived"
submission$Survived <- test$Survived
View(test)
head(submission)

#Export submission to a csv file
write.csv(submission, file = "titanic_in_r_submission.csv")