Last active
August 29, 2015 14:22
-
-
Save oskar-j/a50e1bab232bf5a96bfa to your computer and use it in GitHub Desktop.
Make classifier for Titanic passengers and try to guess who survives the catastrophe
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
setwd("C:/big data/kaggle/titanic/") | |
library(ggplot2) | |
library(caret) | |
library(doParallel) | |
library(randomForest) | |
library(stringr) | |
registerDoParallel(detectCores()) | |
train <- read.csv("train.csv", header = T, na.strings = c(""," ","NA"), | |
sep = ",", dec = ".", stringsAsFactors = F, | |
row.names = 1) | |
test <- read.csv("test.csv", header = T, na.strings = c(""," ","NA"), | |
sep = ",", dec = ".", stringsAsFactors = F) | |
print("Data read") | |
# View(train) | |
# Partitioning data for train and test cases, p=0.75 by rule of thumb | |
# train$Survived = as.factor(train$Survived) | |
train$Pclass = as.factor(train$Pclass) | |
train$Sex = as.factor(train$Sex) | |
train$SibSp = as.factor(train$SibSp) | |
train$Parch = as.factor(train$Parch) | |
train$Embarked = as.factor(train$Embarked) | |
test$Pclass = as.factor(test$Pclass) | |
test$Sex = as.factor(test$Sex) | |
test$SibSp = as.factor(test$SibSp) | |
test$Parch = as.factor(test$Parch) | |
test$Embarked = as.factor(test$Embarked) | |
train$Nickname <- ifelse(grepl("\"[A-Za-z]+\"",train$Name), 1, 0) | |
train$Nickname = as.factor(train$Nickname) | |
test$Nickname <- ifelse(grepl("\"[A-Za-z]+\"",test$Name), 1, 0) | |
test$Nickname = as.factor(test$Nickname) | |
# outliers | |
train$Ticket <- str_replace(train$Ticket, "STON/O 2. ", "STON/O2. ") | |
train$Ticket <- str_replace(train$Ticket, "LINE", "LINE 0") | |
train$Ticket <- str_replace(train$Ticket, "SC/AH Basle ", "SC/AH/Basle ") | |
test$Ticket <- str_replace(test$Ticket, "STON/O 2. ", "STON/O2. ") | |
test$Ticket <- str_replace(test$Ticket, "LINE", "LINE 0") | |
test$Ticket <- str_replace(test$Ticket, "SC/AH Basle ", "SC/AH/Basle ") | |
train$TicketSeries <- str_extract(train$Ticket, perl('\\S+(?=\\s+)')) | |
train$TicketSeries = as.factor(train$TicketSeries) | |
test$TicketSeries <- str_extract(test$Ticket, perl('\\S+(?=\\s+)')) | |
test$TicketSeries = as.factor(test$TicketSeries) | |
train$Ticket <- ifelse(grepl("\\s", train$Ticket), | |
str_trim(sub('\\S*', '\\1', train$Ticket)), | |
train$Ticket) | |
train$Ticket = as.numeric(train$Ticket) | |
test$Ticket <- ifelse(grepl("\\s", test$Ticket), | |
str_trim(sub('\\S*', '\\1', test$Ticket)), | |
test$Ticket) | |
test$Ticket = as.numeric(test$Ticket) | |
# which(is.na(train$Ticket)) | |
sapply(train, class) | |
testIndex = createDataPartition(train$Survived, p = 3/4)[[1]] | |
testing = train[-testIndex,] | |
training = train[testIndex,] | |
testing$Survived = as.factor(testing$Survived) | |
training$Survived = as.factor(training$Survived) | |
train$Survived = as.factor(train$Survived) | |
# training <- training[ order(row.names(training)), ] | |
colNamesRD <- c("Survived", "Pclass","Sex","Age","SibSp", | |
"Parch","Ticket","Fare", | |
"Embarked", "Nickname", "TicketSeries") | |
training <- training[,colNamesRD] | |
testing <- testing[,colNamesRD] | |
training <- rfImpute(Survived ~ ., training) | |
modelFit <- train(Survived ~ ., method="rf", | |
trControl = trainControl( | |
method = "oob", | |
preProcOptions = list(thresh = 0.85) | |
), | |
data=training) | |
testing <- rfImpute(Survived ~ ., testing) | |
confusionMatrix(testing$Survived, predict(modelFit,testing)) | |
train <- train[,colNamesRD] | |
train <- rfImpute(Survived ~ ., train) | |
modelFit <- train(Survived ~ ., method="rf", | |
trControl = trainControl( | |
method = "oob", | |
preProcOptions = list(thresh = 0.85) | |
), | |
data=train) | |
# remove extra levels in ticketseries factor :( | |
idt <- which(!(test$TicketSeries %in% levels(train$TicketSeries))) | |
idp <- which(!(test$Parch %in% levels(train$Parch))) | |
test$TicketSeries[idt] <- NA | |
test$Parch[idp] <- NA | |
test <- test[,-which(names(test) %in% c("Name","Cabin"))] | |
test <- na.roughfix(test) | |
test$Survived <- predict(modelFit,test) | |
write.csv(test[,c("PassengerId", "Survived")], | |
file="results.csv", quote = F, row.names = F) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment