-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathfeature_eng.R
179 lines (137 loc) · 5.47 KB
/
feature_eng.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
#load in libraries
library(data.table)
library(caret)
library(Metrics)
library(ggplot2)
#read in data
setwd("./desktop/R/kaggle/titanic/")
train<-fread("./project/volume/data/raw/train.csv")
test<-fread("./project/volume/data/raw/test.csv")
#check out the columns and missing values
#head(train,10) #NAs in Age variable, missing values in Cabin variable
#head(test,10) #NAs in Age variable, missing values in Cabin variable
#combine dataset for time-saving
master<-rbind(train,test,fill=TRUE)
########################################################
##draw bar chart to check how important the values are##
########################################################
#0. remove 'Ticket' variable since it doesn`t look like important
master$Ticket<-NULL
#1-1. check out Sex variable.
(ggplot(master,aes(master$Survived,fill=master$Sex))
+geom_bar()
+xlab("Survived")
+ylab("number of people")
+guides(fill=guide_legend(title="Sex")))
#looks like women did more survived than men and this is plausible result
#because usually men let women and children take a boat first.
#1-2. change variable to numeric value for the modeling.
master[master$Sex=='male']$Sex='1'
master[master$Sex=='female']$Sex='0'
#2-1. check out Name variable.
#let`s sort out titles
master$Title<-0
titles<-gsub("^.*, (.*?)\\..*$", "\\1", master$Name)
master$Title<-titles
table(titles) #looks like most titles are Mr., Mrs., Miss., and Master.
main.title<-c('Mr','Mrs','Miss','Master')
master$Title[!master$Title %in% main.title]='Others'
(ggplot(master,aes(master$Survived,fill=master$Title))
+geom_bar()
+xlab("Survived")
+ylab("number of people")
+guides(fill=guide_legend(title="Title")))
#looks like women(Mrs. and Miss) did more survived than men and this is plausible result
#because usually men let women and children take a boat first.
#2-2. change variable to numeric value for the modeling.
master$Title[master$Title=="Mr"]=0
master$Title[master$Title=="Mrs"]=1
master$Title[master$Title=="Miss"]=2
master$Title[master$Title=="Master"]=3
master$Title[master$Title=="Others"]=4
master$Name<-NULL #don`t need name anymore.
#3-1. check out Name variable.
#fill out NAs from the median age of each title.
title_Age<-aggregate(master$Age,by = list(master$Title), FUN = function(x) median(x, na.rm = T))
master[is.na(master$Age), "Age"] <- apply(master[is.na(master$Age), ] , 1, function(x) title_Age[title_Age[, 1]==x["Title"], 2])
#3-2. categorize age and change to numeric value.
master[master$Age < 20]$Age -> 'Child'
master[20 <= master$Age & master$Age < 40]$Age -> 'Adult'
master[40 <= master$Age & master$Age < 60]$Age -> 'Med_Age'
master[60 <= master$Age]$Age -> 'Senior'
master$Age[master$Age %in% Child]=0
master$Age[master$Age %in% Adult]=1
master$Age[master$Age %in% Med_Age]=2
master$Age[master$Age %in% Senior]=3
#check NAs
sum(is.na(master$Age))
(ggplot(master,aes(master$Survived, fill=master$Age))
+geom_bar()
+xlab("Survived")
+ylab("number of people")
+guides(fill=guide_legend(title="Age")))
#4-1. check out Pclass variable.
(ggplot(master,aes(master$Survived,fill=master$Pclass))
+geom_bar()
+xlab("Survived")
+ylab("number of people")
+guides(fill=guide_legend(title="Passenger class")))
#5-1. check out Fare variable.
#fill out NA from the median Pclass which is 3 in this case.
master[order(master$Fare)] #NA from Pclass '3'.
Fare_NA<-aggregate(master$Fare,by = list(master$Pclass), FUN = function(x) median(x, na.rm = T))
master[is.na(master$Fare), "Fare"] <- Fare_NA[3,2]
(ggplot(master,aes(master$Fare))
+geom_density(fill = 'blue', alpha=0.4)
+geom_vline(aes(xintercept=median(Fare, na.rm=T)),colour='green', linetype='dashed', lwd=1)
+xlab("Fare")
+ylab("number of people")
+guides(fill=guide_legend(title="Fare")))
#5-2. categorize Fare values.
master[master$Fare < 15]$Fare = 0
master[15 <= master$Fare & master$Fare < 35]$Fare = 1
master[35 <= master$Fare & master$Fare < 100]$Fare = 2
master[100 <= master$Fare]$Fare = 3
#6-1. check out Embarked variable.
unique(master$Embarked) #found two missing Embarked info.
(ggplot(master,aes(master$Pclass,fill=master$Embarked))
+geom_bar()
+xlab("Class")
+ylab("number of people")
+guides(fill=guide_legend(title="Embarked")))
#most passengers are from 'S', so we replace missing value to 'S'.
master[order(master$Embarked)]
master[master$PassengerId==62]$Embarked <- 'S'
master[master$PassengerId==830]$Embarked <- 'S'
#6-2. convert to numeric values
master[master$Embarked=='S']$Embarked=0
master[master$Embarked=='Q']$Embarked=1
master[master$Embarked=='C']$Embarked=2
#7-1. check out Cabin variable
unique(master$Cabin)
#ignore this due to the sparseness.
master$Cabin<-NULL
#8-1. check out SibSp variable
(ggplot(master,aes(master$SibSp))
+geom_density(fill = 'blue', alpha=0.4)
+geom_vline(aes(xintercept=median(Fare, na.rm=T)),colour='green', linetype='dashed', lwd=1)
+xlab("FamilySize")
+ylab("number of people")
+guides(fill=guide_legend(title="FamilySize")))
table(master$SibSp)
#8-2. categorize SibSp values.
master[master$SibSp ==0]$SibSp = 0
master[1 <= master$SibSp & master$SibSp < 3]$SibSp = 1
master[3 <= master$SibSp]$SibSp = 2
master$FamilySize<-master$SibSp
master$SibSp<-NULL
#9. now we split train and test set
t.train<-master[1:891,]
a<-t.train$Survived
t.train$Survived<-NULL
t.train$Survived<-a
t.test<-master[892:1309,]
t.test$Survived<-NULL
#now save feature engineering part for the modeling script.
fwrite(t.train,"./project/volume/data/interim/train.csv")
fwrite(t.test,"./project/volume/data/interim/test.csv")