R Notebook

#--------------
library(randomForest)
library(e1071) #SVM and Naive Bayes
library(rpart)
library('ggplot2') # visualization
library('ggthemes') # visualization
library('scales') # visualization
library('dplyr') # data manipulation
library('mice') # imputation
library(class)

source("./titanicHelper.R")
d<-getCleanTitanicData()
train<-d[[1]]
test<-d[[2]]

#-------------SVM----------------------
svmModel<- svm(Survived~  Pclass + Sex + Age + SibSp 
                        + Parch + Fare + Embarked + Title
                        + FsizeD + Child + Mother, data = train, prob.model=TRUE)
svmPred<-predict(svmModel, newdata = data.frame(test[,c(1,3:19)]))
prepareSubmission(svmPred,"svm_submission.csv")

#-------RANDOM FOREST------------------
rf <- randomForest(train[,c('Pclass','Age','Sex','Parch','SibSp','Fare',
                            'Embarked','Title','FsizeD','Child','Mother')],
                   as.factor(train$Survived), ntree=100, importance=TRUE)
rfPred<-predict(rf,test[,c('Pclass','Age','Sex','Parch','SibSp','Fare',
                            'Embarked','Title','FsizeD','Child','Mother')])
prepareSubmission(rfPred,"rf_submission.csv")

#-------K NEAREST NEIGHBORS-------------
knnTrainX<-train[,c('Pclass','Age','Sex','Parch','SibSp','Fare','Embarked','FsizeD') ]
knnTrainY<-train[,2]
knnTestX<-test[,c('Pclass','Age','Sex','Parch','SibSp','Fare','Embarked','FsizeD')]
knnTestY<-test[,2]
#knn Clean
knnTrainX$Sex <- sapply(as.character(knnTrainX$Sex), switch, 'male' = 0, 'female' = 1)
knnTestX$Sex <- sapply(as.character(knnTestX$Sex), switch, 'male' = 0, 'female' = 1)
knnTrainX$Embarked <- sapply(as.character(knnTrainX$Embarked), switch, 'C' = 0, 'Q' = 1, 'S' = 2)
knnTestX$Embarked <- sapply(as.character(knnTestX$Embarked), switch, 'C' = 0, 'Q' = 1, 'S' = 2)
knnTrainX$FsizeD <- sapply(as.character(knnTrainX$FsizeD), switch, 'large' = 2, 'small' = 1, 'singleton' = 0)
knnTestX$FsizeD <- sapply(as.character(knnTestX$FsizeD), switch, 'large' = 2, 'small' = 1, 'singleton' = 0)

knnPred <- knn(train=knnTrainX, test=knnTestX, cl = knnTrainY, k=35)
prepareSubmission(knnPred, "knn_submission.csv")

#-------NAIVE BAYES-------------
#Naive Bayes does not function well with this dataset. It is likely because we can't
#really tell if the assumptions of conditional independence of the feature vectors work.

nb<-naiveBayes(x = train[,c('Pclass','Age','Sex')], y=train[,'Survived'])
nbtestX<-test[,c('Pclass','Age','Sex')]
nbPred<-predict(nb,newdata = nbtestX, type = "class")
prepareSubmission(nbPred, "nb_submission.csv")

#---------ENSEMBLE----------------------
knnPred<-sapply(as.character(knnPred), switch, '0'=0, '1'=1)
rfPred<-sapply(as.character(rfPred), switch, '0'=0, '1'=1)
svmPred<-sapply(as.character(svmPred), switch, '0'=0, '1'=1)
nbPred<-sapply(as.character(nbPred), switch, '0'=0, '1'=1)
ensemblePred<-round((2*knnPred+2*rfPred+2*svmPred+1*nbPred)/7)
prepareSubmission(ensemblePred, "ensemble_submission2.csv")

Code in TitanicHelper.R File

readData<-function()
{
  train <- read.csv("train.csv", stringsAsFactors=FALSE)
  test  <- read.csv("test.csv",  stringsAsFactors=FALSE)
  full  <- bind_rows(train, test)
  return (full)
}

cleanTitle<-function(full)
{
  full$Title <- gsub('(.*, )|(\\..*)', '', full$Name)
  
  uncommon <- c('Dona', 'Lady', 'the Countess','Capt', 'Col', 'Don', 
                  'Dr', 'Major', 'Rev', 'Sir', 'Jonkheer')
  
  full$Title[full$Title == 'Mlle']        <- 'Miss' 
  full$Title[full$Title == 'Ms']          <- 'Miss'
  full$Title[full$Title == 'Mme']         <- 'Mrs' 
  full$Title[full$Title %in% uncommon]  <- 'Rare Title'
  return(full)
}

calcFamilySize<-function(full)
{
  full$Surname <- sapply(full$Name,  
                         function(x) strsplit(x, split = '[,.]')[[1]][1])
  full$Fsize <- full$SibSp + full$Parch + 1
  full$Family <- paste(full$Surname, full$Fsize, sep='_')
  full$FsizeD[full$Fsize == 1] <- 'singleton'
  full$FsizeD[full$Fsize < 5 & full$Fsize > 1] <- 'small'
  full$FsizeD[full$Fsize > 4] <- 'large'
  return(full)
}

cleanFareEmbarked<-function(full)
{
  full$Embarked[c(62, 830)] <- 'C'
  full$Fare[1044] <- median(full[full$Pclass == '3' & full$Embarked == 'S', ]$Fare, na.rm = TRUE)
  return(full)
}

imputeAge<-function(full)
{
  set.seed(129)
  mice_mod <- mice(full[, !names(full) %in% c('PassengerId','Name','Ticket','Cabin','Family','Surname','Survived')], method='rf') 
  mice_output <- complete(mice_mod)
  full$Age <- mice_output$Age
  return(full)
}

createChildMother<-function(full)
{
  full$Child[full$Age < 18] <- 'Child'
  full$Child[full$Age >= 18] <- 'Adult'
  full$Mother <- 'Not Mother'
  full$Mother[full$Sex == 'female' & full$Parch > 0 & full$Age > 18 & full$Title != 'Miss'] <- 'Mother'
  full$Child  <- factor(full$Child)
  full$Mother <- factor(full$Mother)
  return(full)
}
getCleanTitanicData<-function()
{
  full<-readData()
  full<-cleanFareEmbarked(calcFamilySize(cleanTitle(full)))
  factor_vars <- c('Survived','PassengerId','Pclass','Sex','Embarked',
                   'Title','Surname','Family','FsizeD')
  full[factor_vars] <- lapply(full[factor_vars], function(x) as.factor(x))
  full<-imputeAge(full)
  full<-createChildMother(full)
  # Split the data back into a train set and a test set
  train <- data.frame(full[1:891,])
  test <- data.frame(full[892:1309,])
  return (list("train"=train,"test"=test))
}


#Prepare Submission
prepareSubmission<-function(pred,filename)
{
  submission<-data.frame()
  PassengerID<-892:1309
  submission<-data.frame("PassengerID" = PassengerID, "Survived"=pred)
  write.csv(submission,filename, row.names=FALSE)
}