pacman::p_load("data.table", "dplyr", "tidyr",
"caret",
"ggplot2", "GGally")
titanic <- fread("../Titanic.csv") # 데이터 불러오기
titanic %>%
as_tibble1 Nearest Neighborhood Algorithm
Nearest Neighborhood Algorithm의 장점
- 알고리듬이 매우 간단하여 이해하기 쉽다.
- 관측된 데이터셋에 대해 분포를 가정할 필요가 없다.
- 훈련하는 동안 계산적 비용이 거의 없다.
Nearest Neighborhood Algorithm의 단점
- 범주형 예측 변수를 다룰 수 없다.
- Dummy 또는 One-hot Encoding 변환 후 사용할 수 있다.
- 데이터셋의 크기가 큰 경우, 새로운 Case와 다른 Case들과의 거리 계산에 시간이 많이 걸릴 수 있다.
- 이상치와 노이즈가 있는 경우, 예측에 큰 영향을 미친다.
- 고차원의 경우, 성능이 좋지 않다.
- 차원이 높아질수록 Case간 거리 차이가 거의 없는 것처럼 보여, 가장 가까운 거리를 찾는 것이 어렵다.
실습 자료 : 1912년 4월 15일 타이타닉호 침몰 당시 탑승객들의 정보를 기록한 데이터셋이며, 총 11개의 변수를 포함하고 있다. 이 자료에서 Target은
Survived이다.
1.1 데이터 불러오기
# A tibble: 891 × 11
Survived Pclass Name Sex Age SibSp Parch Ticket Fare Cabin Embarked
<int> <int> <chr> <chr> <dbl> <int> <int> <chr> <dbl> <chr> <chr>
1 0 3 Braund, Mr. Owen Harris male 22 1 0 A/5 21171 7.25 "" S
2 1 1 Cumings, Mrs. John Bradley (Florence Briggs Thayer) female 38 1 0 PC 17599 71.3 "C85" C
3 1 3 Heikkinen, Miss. Laina female 26 0 0 STON/O2. 3101282 7.92 "" S
4 1 1 Futrelle, Mrs. Jacques Heath (Lily May Peel) female 35 1 0 113803 53.1 "C123" S
5 0 3 Allen, Mr. William Henry male 35 0 0 373450 8.05 "" S
6 0 3 Moran, Mr. James male NA 0 0 330877 8.46 "" Q
7 0 1 McCarthy, Mr. Timothy J male 54 0 0 17463 51.9 "E46" S
8 0 3 Palsson, Master. Gosta Leonard male 2 3 1 349909 21.1 "" S
9 1 3 Johnson, Mrs. Oscar W (Elisabeth Vilhelmina Berg) female 27 0 2 347742 11.1 "" S
10 1 2 Nasser, Mrs. Nicholas (Adele Achem) female 14 1 0 237736 30.1 "" C
# ℹ 881 more rows1.2 데이터 전처리 I
titanic %<>%
data.frame() %>% # Data Frame 형태로 변환
mutate(Survived = ifelse(Survived == 1, "yes", "no")) # Target을 문자형 변수로 변환
# 1. Convert to Factor
fac.col <- c("Pclass", "Sex",
# Target
"Survived")
titanic <- titanic %>%
mutate_at(fac.col, as.factor) # 범주형으로 변환
glimpse(titanic) # 데이터 구조 확인Rows: 891
Columns: 11
$ Survived <fct> no, yes, yes, yes, no, no, no, no, yes, yes, yes, yes, no, no, no, yes, no, yes, no, yes, no, yes, yes, yes, no, yes, no, no, yes, no, no, yes, yes, no, no, no, yes, no, no, yes, no…
$ Pclass <fct> 3, 1, 3, 1, 3, 3, 1, 3, 3, 2, 3, 1, 3, 3, 3, 2, 3, 2, 3, 3, 2, 2, 3, 1, 3, 3, 3, 1, 3, 3, 1, 1, 3, 2, 1, 1, 3, 3, 3, 3, 3, 2, 3, 2, 3, 3, 3, 3, 3, 3, 3, 3, 1, 2, 1, 1, 2, 3, 2, 3, 3…
$ Name <chr> "Braund, Mr. Owen Harris", "Cumings, Mrs. John Bradley (Florence Briggs Thayer)", "Heikkinen, Miss. Laina", "Futrelle, Mrs. Jacques Heath (Lily May Peel)", "Allen, Mr. William Henry…
$ Sex <fct> male, female, female, female, male, male, male, male, female, female, female, female, male, male, female, female, male, male, female, female, male, male, female, male, female, femal…
$ Age <dbl> 22.0, 38.0, 26.0, 35.0, 35.0, NA, 54.0, 2.0, 27.0, 14.0, 4.0, 58.0, 20.0, 39.0, 14.0, 55.0, 2.0, NA, 31.0, NA, 35.0, 34.0, 15.0, 28.0, 8.0, 38.0, NA, 19.0, NA, NA, 40.0, NA, NA, 66.…
$ SibSp <int> 1, 1, 0, 1, 0, 0, 0, 3, 0, 1, 1, 0, 0, 1, 0, 0, 4, 0, 1, 0, 0, 0, 0, 0, 3, 1, 0, 3, 0, 0, 0, 1, 0, 0, 1, 1, 0, 0, 2, 1, 1, 1, 0, 1, 0, 0, 1, 0, 2, 1, 4, 0, 1, 1, 0, 0, 0, 0, 1, 5, 0…
$ Parch <int> 0, 0, 0, 0, 0, 0, 0, 1, 2, 0, 1, 0, 0, 5, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 5, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 2, 2, 0…
$ Ticket <chr> "A/5 21171", "PC 17599", "STON/O2. 3101282", "113803", "373450", "330877", "17463", "349909", "347742", "237736", "PP 9549", "113783", "A/5. 2151", "347082", "350406", "248706", "38…
$ Fare <dbl> 7.2500, 71.2833, 7.9250, 53.1000, 8.0500, 8.4583, 51.8625, 21.0750, 11.1333, 30.0708, 16.7000, 26.5500, 8.0500, 31.2750, 7.8542, 16.0000, 29.1250, 13.0000, 18.0000, 7.2250, 26.0000,…
$ Cabin <chr> "", "C85", "", "C123", "", "", "E46", "", "", "", "G6", "C103", "", "", "", "", "", "", "", "", "", "D56", "", "A6", "", "", "", "C23 C25 C27", "", "", "", "B78", "", "", "", "", ""…
$ Embarked <chr> "S", "C", "S", "S", "S", "Q", "S", "S", "S", "C", "S", "S", "S", "S", "S", "S", "Q", "S", "S", "C", "S", "S", "Q", "S", "S", "S", "C", "S", "Q", "S", "C", "C", "Q", "S", "C", "S", "…# 2. Generate New Variable
titanic <- titanic %>%
mutate(FamSize = SibSp + Parch) # "FamSize = 형제 및 배우자 수 + 부모님 및 자녀 수"로 가족 수를 의미하는 새로운 변수
glimpse(titanic) # 데이터 구조 확인Rows: 891
Columns: 12
$ Survived <fct> no, yes, yes, yes, no, no, no, no, yes, yes, yes, yes, no, no, no, yes, no, yes, no, yes, no, yes, yes, yes, no, yes, no, no, yes, no, no, yes, yes, no, no, no, yes, no, no, yes, no…
$ Pclass <fct> 3, 1, 3, 1, 3, 3, 1, 3, 3, 2, 3, 1, 3, 3, 3, 2, 3, 2, 3, 3, 2, 2, 3, 1, 3, 3, 3, 1, 3, 3, 1, 1, 3, 2, 1, 1, 3, 3, 3, 3, 3, 2, 3, 2, 3, 3, 3, 3, 3, 3, 3, 3, 1, 2, 1, 1, 2, 3, 2, 3, 3…
$ Name <chr> "Braund, Mr. Owen Harris", "Cumings, Mrs. John Bradley (Florence Briggs Thayer)", "Heikkinen, Miss. Laina", "Futrelle, Mrs. Jacques Heath (Lily May Peel)", "Allen, Mr. William Henry…
$ Sex <fct> male, female, female, female, male, male, male, male, female, female, female, female, male, male, female, female, male, male, female, female, male, male, female, male, female, femal…
$ Age <dbl> 22.0, 38.0, 26.0, 35.0, 35.0, NA, 54.0, 2.0, 27.0, 14.0, 4.0, 58.0, 20.0, 39.0, 14.0, 55.0, 2.0, NA, 31.0, NA, 35.0, 34.0, 15.0, 28.0, 8.0, 38.0, NA, 19.0, NA, NA, 40.0, NA, NA, 66.…
$ SibSp <int> 1, 1, 0, 1, 0, 0, 0, 3, 0, 1, 1, 0, 0, 1, 0, 0, 4, 0, 1, 0, 0, 0, 0, 0, 3, 1, 0, 3, 0, 0, 0, 1, 0, 0, 1, 1, 0, 0, 2, 1, 1, 1, 0, 1, 0, 0, 1, 0, 2, 1, 4, 0, 1, 1, 0, 0, 0, 0, 1, 5, 0…
$ Parch <int> 0, 0, 0, 0, 0, 0, 0, 1, 2, 0, 1, 0, 0, 5, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 5, 0, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 2, 2, 0…
$ Ticket <chr> "A/5 21171", "PC 17599", "STON/O2. 3101282", "113803", "373450", "330877", "17463", "349909", "347742", "237736", "PP 9549", "113783", "A/5. 2151", "347082", "350406", "248706", "38…
$ Fare <dbl> 7.2500, 71.2833, 7.9250, 53.1000, 8.0500, 8.4583, 51.8625, 21.0750, 11.1333, 30.0708, 16.7000, 26.5500, 8.0500, 31.2750, 7.8542, 16.0000, 29.1250, 13.0000, 18.0000, 7.2250, 26.0000,…
$ Cabin <chr> "", "C85", "", "C123", "", "", "E46", "", "", "", "G6", "C103", "", "", "", "", "", "", "", "", "", "D56", "", "A6", "", "", "", "C23 C25 C27", "", "", "", "B78", "", "", "", "", ""…
$ Embarked <chr> "S", "C", "S", "S", "S", "Q", "S", "S", "S", "C", "S", "S", "S", "S", "S", "S", "Q", "S", "S", "C", "S", "S", "Q", "S", "S", "S", "C", "S", "Q", "S", "C", "C", "Q", "S", "C", "S", "…
$ FamSize <int> 1, 1, 0, 1, 0, 0, 0, 4, 2, 1, 2, 0, 0, 6, 0, 0, 5, 0, 1, 0, 0, 0, 0, 0, 4, 6, 0, 5, 0, 0, 0, 1, 0, 0, 1, 1, 0, 0, 2, 1, 1, 1, 0, 3, 0, 0, 1, 0, 2, 1, 5, 0, 1, 1, 1, 0, 0, 0, 3, 7, 0…# 3. Select Variables used for Analysis
titanic1 <- titanic %>%
select(Survived, Pclass, Sex, Age, Fare, FamSize) # 분석에 사용할 변수 선택
# 4. Convert One-hot Encoding for 범주형 예측 변수
dummies <- dummyVars(formula = ~ ., # formula : ~ 예측 변수 / "." : data에 포함된 모든 변수를 의미
data = titanic1[,-1], # Dataset including Only 예측 변수 -> Target 제외
fullRank = FALSE) # fullRank = TRUE : Dummy Variable, fullRank = FALSE : One-hot Encoding
titanic.Var <- predict(dummies, newdata = titanic1) %>% # 범주형 예측 변수에 대한 One-hot Encoding 변환
data.frame() # Data Frame 형태로 변환
glimpse(titanic.Var) # 데이터 구조 확인Rows: 891
Columns: 8
$ Pclass.1 <dbl> 0, 1, 0, 1, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 1, 0, 0, 0, 0,…
$ Pclass.2 <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 1, 0,…
$ Pclass.3 <dbl> 1, 0, 1, 0, 1, 1, 0, 1, 1, 0, 1, 0, 1, 1, 1, 0, 1, 0, 1, 1, 0, 0, 1, 0, 1, 1, 1, 0, 1, 1, 0, 0, 1, 0, 0, 0, 1, 1, 1, 1, 1, 0, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 0, 1,…
$ Sex.female <dbl> 0, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 1, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, 1, 1, 0, 0, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 0,…
$ Sex.male <dbl> 1, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 0, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 1, 1, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 0, 0, 1, 1, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 1,…
$ Age <dbl> 22.0, 38.0, 26.0, 35.0, 35.0, NA, 54.0, 2.0, 27.0, 14.0, 4.0, 58.0, 20.0, 39.0, 14.0, 55.0, 2.0, NA, 31.0, NA, 35.0, 34.0, 15.0, 28.0, 8.0, 38.0, NA, 19.0, NA, NA, 40.0, NA, NA, 6…
$ Fare <dbl> 7.2500, 71.2833, 7.9250, 53.1000, 8.0500, 8.4583, 51.8625, 21.0750, 11.1333, 30.0708, 16.7000, 26.5500, 8.0500, 31.2750, 7.8542, 16.0000, 29.1250, 13.0000, 18.0000, 7.2250, 26.000…
$ FamSize <dbl> 1, 1, 0, 1, 0, 0, 0, 4, 2, 1, 2, 0, 0, 6, 0, 0, 5, 0, 1, 0, 0, 0, 0, 0, 4, 6, 0, 5, 0, 0, 0, 1, 0, 0, 1, 1, 0, 0, 2, 1, 1, 1, 0, 3, 0, 0, 1, 0, 2, 1, 5, 0, 1, 1, 1, 0, 0, 0, 3, 7,…# Combine Target with 변환된 예측 변수
titanic.df <- data.frame(Survived = titanic1$Survived,
titanic.Var)
titanic.df %>%
as_tibble# A tibble: 891 × 9
Survived Pclass.1 Pclass.2 Pclass.3 Sex.female Sex.male Age Fare FamSize
<fct> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
1 no 0 0 1 0 1 22 7.25 1
2 yes 1 0 0 1 0 38 71.3 1
3 yes 0 0 1 1 0 26 7.92 0
4 yes 1 0 0 1 0 35 53.1 1
5 no 0 0 1 0 1 35 8.05 0
6 no 0 0 1 0 1 NA 8.46 0
7 no 1 0 0 0 1 54 51.9 0
8 no 0 0 1 0 1 2 21.1 4
9 yes 0 0 1 1 0 27 11.1 2
10 yes 0 1 0 1 0 14 30.1 1
# ℹ 881 more rowsglimpse(titanic.df) # 데이터 구조 확인Rows: 891
Columns: 9
$ Survived <fct> no, yes, yes, yes, no, no, no, no, yes, yes, yes, yes, no, no, no, yes, no, yes, no, yes, no, yes, yes, yes, no, yes, no, no, yes, no, no, yes, yes, no, no, no, yes, no, no, yes, …
$ Pclass.1 <dbl> 0, 1, 0, 1, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 1, 0, 0, 0, 0,…
$ Pclass.2 <dbl> 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 1, 0,…
$ Pclass.3 <dbl> 1, 0, 1, 0, 1, 1, 0, 1, 1, 0, 1, 0, 1, 1, 1, 0, 1, 0, 1, 1, 0, 0, 1, 0, 1, 1, 1, 0, 1, 1, 0, 0, 1, 0, 0, 0, 1, 1, 1, 1, 1, 0, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 0, 1,…
$ Sex.female <dbl> 0, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 1, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, 1, 1, 0, 0, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 0,…
$ Sex.male <dbl> 1, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 0, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 1, 1, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 0, 0, 1, 1, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 1,…
$ Age <dbl> 22.0, 38.0, 26.0, 35.0, 35.0, NA, 54.0, 2.0, 27.0, 14.0, 4.0, 58.0, 20.0, 39.0, 14.0, 55.0, 2.0, NA, 31.0, NA, 35.0, 34.0, 15.0, 28.0, 8.0, 38.0, NA, 19.0, NA, NA, 40.0, NA, NA, 6…
$ Fare <dbl> 7.2500, 71.2833, 7.9250, 53.1000, 8.0500, 8.4583, 51.8625, 21.0750, 11.1333, 30.0708, 16.7000, 26.5500, 8.0500, 31.2750, 7.8542, 16.0000, 29.1250, 13.0000, 18.0000, 7.2250, 26.000…
$ FamSize <dbl> 1, 1, 0, 1, 0, 0, 0, 4, 2, 1, 2, 0, 0, 6, 0, 0, 5, 0, 1, 0, 0, 0, 0, 0, 4, 6, 0, 5, 0, 0, 0, 1, 0, 0, 1, 1, 0, 0, 2, 1, 1, 1, 0, 3, 0, 0, 1, 0, 2, 1, 5, 0, 1, 1, 1, 0, 0, 0, 3, 7,…1.3 데이터 탐색
ggpairs(titanic.df,
aes(colour = Survived)) + # Target의 범주에 따라 색깔을 다르게 표현
theme_bw()
ggpairs(titanic.df,
aes(colour = Survived, alpha = 0.8)) + # Target의 범주에 따라 색깔을 다르게 표현
scale_colour_manual(values = c("purple","cyan4")) + # 특정 색깔 지정
scale_fill_manual(values = c("purple","cyan4")) + # 특정 색깔 지정
theme_bw()
1.4 데이터 분할
# Partition (Training Dataset : Test Dataset = 7:3)
y <- titanic.df$Survived # Target
set.seed(200)
ind <- createDataPartition(y, p = 0.7, list =T) # Index를 이용하여 7:3으로 분할
titanic.trd <- titanic.df[ind$Resample1,] # Training Dataset
titanic.ted <- titanic.df[-ind$Resample1,] # Test Dataset1.5 데이터 전처리 II
# 1. Imputation
titanic.trd.Imp <- titanic.trd %>%
mutate(Age = replace_na(Age, mean(Age, na.rm = TRUE))) # 평균으로 결측값 대체
titanic.ted.Imp <- titanic.ted %>%
mutate(Age = replace_na(Age, mean(titanic.trd$Age, na.rm = TRUE))) # Training Dataset을 이용하여 결측값 대체
# 2. Standardization
preProcValues <- preProcess(titanic.trd.Imp,
method = c("center", "scale")) # Standardization 정의 -> Training Dataset에 대한 평균과 표준편차 계산
titanic.trd.Imp <- predict(preProcValues, titanic.trd.Imp) # Standardization for Training Dataset
titanic.ted.Imp <- predict(preProcValues, titanic.ted.Imp) # Standardization for Test Dataset
glimpse(titanic.trd.Imp) # 데이터 구조 확인Rows: 625
Columns: 9
$ Survived <fct> no, yes, yes, no, no, no, yes, yes, yes, yes, no, no, yes, no, yes, no, yes, no, no, no, yes, no, no, yes, yes, no, no, no, no, no, yes, no, no, no, yes, no, yes, no, no, no, yes,…
$ Pclass.1 <dbl> -0.593506, -0.593506, 1.682207, -0.593506, -0.593506, -0.593506, -0.593506, -0.593506, -0.593506, 1.682207, -0.593506, -0.593506, -0.593506, -0.593506, -0.593506, -0.593506, 1.682…
$ Pclass.2 <dbl> -0.4694145, -0.4694145, -0.4694145, -0.4694145, -0.4694145, -0.4694145, -0.4694145, 2.1269048, -0.4694145, -0.4694145, -0.4694145, -0.4694145, 2.1269048, -0.4694145, -0.4694145, 2…
$ Pclass.3 <dbl> 0.888575, 0.888575, -1.123597, 0.888575, 0.888575, 0.888575, 0.888575, -1.123597, 0.888575, -1.123597, 0.888575, 0.888575, -1.123597, 0.888575, 0.888575, -1.123597, -1.123597, 0.8…
$ Sex.female <dbl> -0.7572241, 1.3184999, 1.3184999, -0.7572241, -0.7572241, -0.7572241, 1.3184999, 1.3184999, 1.3184999, 1.3184999, -0.7572241, 1.3184999, -0.7572241, 1.3184999, 1.3184999, -0.75722…
$ Sex.male <dbl> 0.7572241, -1.3184999, -1.3184999, 0.7572241, 0.7572241, 0.7572241, -1.3184999, -1.3184999, -1.3184999, -1.3184999, 0.7572241, -1.3184999, 0.7572241, -1.3184999, -1.3184999, 0.757…
$ Age <dbl> -0.61306970, -0.30411628, 0.39102893, 0.39102893, 0.00000000, -2.15783684, -0.22687792, -1.23097656, -2.00336012, 2.16751113, 0.69998236, -1.23097656, 0.00000000, 0.08207551, 0.00…
$ Fare <dbl> -0.51776394, -0.50463325, 0.37414970, -0.50220165, -0.49425904, -0.24882814, -0.44222264, -0.07383411, -0.33393441, -0.14232374, -0.05040897, -0.50601052, -0.40590999, -0.30864569…
$ FamSize <dbl> 0.04506631, -0.55421976, 0.04506631, -0.55421976, -0.55421976, 1.84292454, 0.64435239, 0.04506631, 0.64435239, -0.55421976, 3.04149669, -0.55421976, -0.55421976, 0.04506631, -0.55…glimpse(titanic.ted.Imp) # 데이터 구조 확인Rows: 266
Columns: 9
$ Survived <fct> yes, no, no, yes, no, yes, yes, yes, yes, yes, no, no, yes, yes, no, yes, no, yes, yes, no, yes, no, no, no, no, no, no, yes, yes, no, no, no, no, no, no, no, no, no, no, yes, no,…
$ Pclass.1 <dbl> 1.682207, 1.682207, -0.593506, -0.593506, -0.593506, -0.593506, -0.593506, -0.593506, -0.593506, -0.593506, -0.593506, -0.593506, -0.593506, -0.593506, -0.593506, -0.593506, 1.682…
$ Pclass.2 <dbl> -0.4694145, -0.4694145, -0.4694145, 2.1269048, -0.4694145, 2.1269048, -0.4694145, -0.4694145, -0.4694145, 2.1269048, -0.4694145, -0.4694145, 2.1269048, 2.1269048, -0.4694145, 2.12…
$ Pclass.3 <dbl> -1.123597, -1.123597, 0.888575, -1.123597, 0.888575, -1.123597, 0.888575, 0.888575, 0.888575, -1.123597, 0.888575, 0.888575, -1.123597, -1.123597, 0.888575, -1.123597, -1.123597, …
$ Sex.female <dbl> 1.3184999, -0.7572241, -0.7572241, 1.3184999, -0.7572241, -0.7572241, 1.3184999, 1.3184999, -0.7572241, 1.3184999, -0.7572241, -0.7572241, 1.3184999, 1.3184999, -0.7572241, 1.3184…
$ Sex.male <dbl> -1.3184999, 0.7572241, 0.7572241, -1.3184999, 0.7572241, 0.7572241, -1.3184999, -1.3184999, 0.7572241, -1.3184999, 0.7572241, 0.7572241, -1.3184999, -1.3184999, 0.7572241, -1.3184…
$ Age <dbl> 0.62274400, 1.85855771, -0.76754642, 1.93579607, -2.15783684, 0.31379058, -1.15373820, 0.62274400, 0.00000000, -2.08059848, 0.00000000, -0.69030806, -0.07240121, -0.69030806, -0.1…
$ Fare <dbl> 0.727866891, 0.350076786, -0.502201647, -0.347551409, -0.092232621, -0.405909990, -0.502606266, -0.048220525, -0.518168555, 0.150037190, -0.502201647, -0.507064862, -0.153022808, …
$ FamSize <dbl> 0.04506631, -0.55421976, -0.55421976, -0.55421976, 2.44221062, -0.55421976, -0.55421976, 3.04149669, -0.55421976, 1.24363847, -0.55421976, -0.55421976, 0.04506631, -0.55421976, -0…1.6 모형 훈련
Nearest Neighborhood Algorithm은 다양한 Package(예를 들어, "caret", "class")를 통해 수행할 수 있다. Package "class"의 함수 knn()를 이용하면 특정 class에 대한 예측 확률만 얻을 수 있는 반면, Package "caret"의 함수 knn3()를 이용하면 각 class에 대한 예측 확률을 얻을 수 있다. 그래서 여기서는 Package "caret"을 이용하여 모형 훈련을 수행한다.
knn.model <- knn3(Survived ~ ., # Target ~ 예측 변수
data = titanic.trd.Imp, # Training Dataset
k = 4) # 이웃 개수
knn.model4-nearest neighbor model
Training set outcome distribution:
no yes
385 240 Caution! Package "caret"에서 제공하는 함수 knn3Train()를 이용하면 Training Dataset에 대한 모형 훈련과 Test Dataset에 대한 예측을 한 번에 수행할 수 있다.
# 모형 훈련 & 예측 한꺼번에
knn3Train(titanic.trd.Imp[, -1], # Training Dataset including Only 예측 변수
titanic.ted.Imp[, -1], # Test Dataset including Only 예측 변수
cl = titanic.trd.Imp[, 1], # Target of Training Dataset
k = 4) # 이웃 개수 [1] "yes" "no" "no" "yes" "no" "no" "yes" "no" "no" "yes" "no" "no" "yes" "yes" "no" "yes" "no" "no" "yes" "no" "no" "no" "no" "no" "no" "no" "no" "no" "yes" "no" "yes" "no"
[33] "no" "no" "no" "no" "no" "no" "no" "yes" "no" "no" "no" "yes" "no" "no" "no" "no" "yes" "no" "no" "no" "no" "yes" "no" "yes" "no" "no" "no" "yes" "no" "no" "yes" "no"
[65] "yes" "no" "no" "yes" "no" "no" "no" "no" "yes" "yes" "no" "no" "yes" "no" "no" "no" "yes" "no" "no" "no" "no" "no" "yes" "no" "no" "no" "no" "no" "no" "yes" "yes" "yes"
[97] "yes" "yes" "no" "no" "no" "yes" "yes" "yes" "no" "yes" "no" "yes" "no" "yes" "yes" "no" "yes" "yes" "no" "no" "yes" "no" "no" "no" "yes" "no" "no" "yes" "yes" "yes" "no" "yes"
[129] "yes" "yes" "no" "yes" "no" "no" "yes" "yes" "no" "no" "no" "no" "yes" "no" "no" "no" "yes" "no" "no" "no" "no" "no" "no" "no" "no" "no" "no" "yes" "no" "yes" "no" "no"
[161] "yes" "no" "yes" "yes" "no" "yes" "yes" "no" "no" "no" "yes" "no" "yes" "yes" "yes" "no" "yes" "no" "no" "no" "no" "no" "no" "yes" "no" "no" "no" "yes" "no" "yes" "no" "no"
[193] "yes" "no" "no" "no" "no" "yes" "no" "no" "no" "no" "no" "no" "no" "no" "no" "no" "yes" "no" "no" "yes" "yes" "no" "no" "no" "yes" "yes" "no" "no" "yes" "no" "no" "yes"
[225] "yes" "yes" "no" "no" "no" "no" "yes" "no" "no" "no" "yes" "no" "no" "yes" "yes" "no" "no" "yes" "no" "no" "no" "yes" "no" "yes" "no" "no" "yes" "no" "no" "no" "no" "yes"
[257] "yes" "yes" "no" "yes" "no" "yes" "yes" "no" "yes" "yes"
attr(,"prob")
no yes
[1,] 0.0000000 1.00000000
[2,] 0.7500000 0.25000000
[3,] 0.7500000 0.25000000
[4,] 0.5000000 0.50000000
[5,] 0.7500000 0.25000000
[6,] 1.0000000 0.00000000
[7,] 0.2500000 0.75000000
[8,] 1.0000000 0.00000000
[9,] 1.0000000 0.00000000
[10,] 0.0000000 1.00000000
[11,] 1.0000000 0.00000000
[12,] 0.8000000 0.20000000
[13,] 0.2500000 0.75000000
[14,] 0.2500000 0.75000000
[15,] 0.7500000 0.25000000
[16,] 0.0000000 1.00000000
[17,] 0.5000000 0.50000000
[18,] 0.5000000 0.50000000
[19,] 0.0000000 1.00000000
[20,] 1.0000000 0.00000000
[21,] 1.0000000 0.00000000
[22,] 0.8571429 0.14285714
[23,] 1.0000000 0.00000000
[24,] 1.0000000 0.00000000
[25,] 1.0000000 0.00000000
[26,] 1.0000000 0.00000000
[27,] 0.7500000 0.25000000
[28,] 0.5000000 0.50000000
[29,] 0.2500000 0.75000000
[30,] 0.7500000 0.25000000
[31,] 0.0000000 1.00000000
[32,] 1.0000000 0.00000000
[33,] 1.0000000 0.00000000
[34,] 0.8000000 0.20000000
[35,] 0.7500000 0.25000000
[36,] 0.7500000 0.25000000
[37,] 1.0000000 0.00000000
[38,] 1.0000000 0.00000000
[39,] 0.7500000 0.25000000
[40,] 0.0000000 1.00000000
[41,] 1.0000000 0.00000000
[42,] 0.7500000 0.25000000
[43,] 1.0000000 0.00000000
[44,] 0.0000000 1.00000000
[45,] 1.0000000 0.00000000
[46,] 0.5000000 0.50000000
[47,] 1.0000000 0.00000000
[48,] 0.7500000 0.25000000
[49,] 0.0000000 1.00000000
[50,] 1.0000000 0.00000000
[51,] 1.0000000 0.00000000
[52,] 0.7500000 0.25000000
[53,] 0.7500000 0.25000000
[54,] 0.2500000 0.75000000
[55,] 1.0000000 0.00000000
[56,] 0.2500000 0.75000000
[57,] 1.0000000 0.00000000
[58,] 0.7500000 0.25000000
[59,] 0.7500000 0.25000000
[60,] 0.0000000 1.00000000
[61,] 0.7500000 0.25000000
[62,] 1.0000000 0.00000000
[63,] 0.0000000 1.00000000
[64,] 0.7500000 0.25000000
[65,] 0.0000000 1.00000000
[66,] 0.7500000 0.25000000
[67,] 0.7500000 0.25000000
[68,] 0.2500000 0.75000000
[69,] 1.0000000 0.00000000
[70,] 1.0000000 0.00000000
[71,] 1.0000000 0.00000000
[72,] 1.0000000 0.00000000
[73,] 0.5000000 0.50000000
[74,] 0.0000000 1.00000000
[75,] 1.0000000 0.00000000
[76,] 0.7500000 0.25000000
[77,] 0.2000000 0.80000000
[78,] 1.0000000 0.00000000
[79,] 1.0000000 0.00000000
[80,] 0.7500000 0.25000000
[81,] 0.2500000 0.75000000
[82,] 1.0000000 0.00000000
[83,] 0.7500000 0.25000000
[84,] 1.0000000 0.00000000
[85,] 0.5000000 0.50000000
[86,] 1.0000000 0.00000000
[87,] 0.5000000 0.50000000
[88,] 0.7500000 0.25000000
[89,] 1.0000000 0.00000000
[90,] 0.8000000 0.20000000
[91,] 1.0000000 0.00000000
[92,] 1.0000000 0.00000000
[93,] 0.7500000 0.25000000
[94,] 0.2000000 0.80000000
[95,] 0.0000000 1.00000000
[96,] 0.0000000 1.00000000
[97,] 0.0000000 1.00000000
[98,] 0.0000000 1.00000000
[99,] 0.8000000 0.20000000
[100,] 1.0000000 0.00000000
[101,] 1.0000000 0.00000000
[102,] 0.0000000 1.00000000
[103,] 0.0000000 1.00000000
[104,] 0.5000000 0.50000000
[105,] 1.0000000 0.00000000
[106,] 0.0000000 1.00000000
[107,] 1.0000000 0.00000000
[108,] 0.0000000 1.00000000
[109,] 1.0000000 0.00000000
[110,] 0.0000000 1.00000000
[111,] 0.0000000 1.00000000
[112,] 1.0000000 0.00000000
[113,] 0.2000000 0.80000000
[114,] 0.4000000 0.60000000
[115,] 1.0000000 0.00000000
[116,] 1.0000000 0.00000000
[117,] 0.0000000 1.00000000
[118,] 0.5000000 0.50000000
[119,] 0.8000000 0.20000000
[120,] 1.0000000 0.00000000
[121,] 0.0000000 1.00000000
[122,] 0.7500000 0.25000000
[123,] 0.5000000 0.50000000
[124,] 0.0000000 1.00000000
[125,] 0.5000000 0.50000000
[126,] 0.5000000 0.50000000
[127,] 0.7500000 0.25000000
[128,] 0.2500000 0.75000000
[129,] 0.5000000 0.50000000
[130,] 0.5000000 0.50000000
[131,] 0.7500000 0.25000000
[132,] 0.2500000 0.75000000
[133,] 1.0000000 0.00000000
[134,] 1.0000000 0.00000000
[135,] 0.0000000 1.00000000
[136,] 0.0000000 1.00000000
[137,] 0.5000000 0.50000000
[138,] 1.0000000 0.00000000
[139,] 1.0000000 0.00000000
[140,] 1.0000000 0.00000000
[141,] 0.0000000 1.00000000
[142,] 1.0000000 0.00000000
[143,] 1.0000000 0.00000000
[144,] 1.0000000 0.00000000
[145,] 0.0000000 1.00000000
[146,] 1.0000000 0.00000000
[147,] 0.7500000 0.25000000
[148,] 1.0000000 0.00000000
[149,] 1.0000000 0.00000000
[150,] 1.0000000 0.00000000
[151,] 0.7500000 0.25000000
[152,] 0.5000000 0.50000000
[153,] 0.8000000 0.20000000
[154,] 1.0000000 0.00000000
[155,] 0.7500000 0.25000000
[156,] 0.2500000 0.75000000
[157,] 0.7500000 0.25000000
[158,] 0.5000000 0.50000000
[159,] 1.0000000 0.00000000
[160,] 1.0000000 0.00000000
[161,] 0.5000000 0.50000000
[162,] 0.7500000 0.25000000
[163,] 0.0000000 1.00000000
[164,] 0.2500000 0.75000000
[165,] 0.7500000 0.25000000
[166,] 0.0000000 1.00000000
[167,] 0.0000000 1.00000000
[168,] 0.5000000 0.50000000
[169,] 1.0000000 0.00000000
[170,] 0.7500000 0.25000000
[171,] 0.5000000 0.50000000
[172,] 1.0000000 0.00000000
[173,] 0.0000000 1.00000000
[174,] 0.2000000 0.80000000
[175,] 0.0000000 1.00000000
[176,] 1.0000000 0.00000000
[177,] 0.0000000 1.00000000
[178,] 1.0000000 0.00000000
[179,] 1.0000000 0.00000000
[180,] 1.0000000 0.00000000
[181,] 0.7500000 0.25000000
[182,] 1.0000000 0.00000000
[183,] 1.0000000 0.00000000
[184,] 0.0000000 1.00000000
[185,] 1.0000000 0.00000000
[186,] 1.0000000 0.00000000
[187,] 1.0000000 0.00000000
[188,] 0.0000000 1.00000000
[189,] 1.0000000 0.00000000
[190,] 0.0000000 1.00000000
[191,] 1.0000000 0.00000000
[192,] 1.0000000 0.00000000
[193,] 0.5000000 0.50000000
[194,] 1.0000000 0.00000000
[195,] 1.0000000 0.00000000
[196,] 1.0000000 0.00000000
[197,] 0.7500000 0.25000000
[198,] 0.5000000 0.50000000
[199,] 1.0000000 0.00000000
[200,] 1.0000000 0.00000000
[201,] 0.8571429 0.14285714
[202,] 0.5000000 0.50000000
[203,] 1.0000000 0.00000000
[204,] 1.0000000 0.00000000
[205,] 1.0000000 0.00000000
[206,] 1.0000000 0.00000000
[207,] 1.0000000 0.00000000
[208,] 0.7500000 0.25000000
[209,] 0.1666667 0.83333333
[210,] 1.0000000 0.00000000
[211,] 0.8000000 0.20000000
[212,] 0.2500000 0.75000000
[213,] 0.0000000 1.00000000
[214,] 1.0000000 0.00000000
[215,] 0.7500000 0.25000000
[216,] 1.0000000 0.00000000
[217,] 0.0000000 1.00000000
[218,] 0.0000000 1.00000000
[219,] 1.0000000 0.00000000
[220,] 1.0000000 0.00000000
[221,] 0.5000000 0.50000000
[222,] 0.7500000 0.25000000
[223,] 1.0000000 0.00000000
[224,] 0.0000000 1.00000000
[225,] 0.0000000 1.00000000
[226,] 0.0000000 1.00000000
[227,] 0.7500000 0.25000000
[228,] 0.7500000 0.25000000
[229,] 1.0000000 0.00000000
[230,] 1.0000000 0.00000000
[231,] 0.2500000 0.75000000
[232,] 0.7500000 0.25000000
[233,] 1.0000000 0.00000000
[234,] 1.0000000 0.00000000
[235,] 0.5000000 0.50000000
[236,] 1.0000000 0.00000000
[237,] 1.0000000 0.00000000
[238,] 0.0000000 1.00000000
[239,] 0.5000000 0.50000000
[240,] 0.6000000 0.40000000
[241,] 0.7500000 0.25000000
[242,] 0.0000000 1.00000000
[243,] 1.0000000 0.00000000
[244,] 0.7500000 0.25000000
[245,] 1.0000000 0.00000000
[246,] 0.0000000 1.00000000
[247,] 1.0000000 0.00000000
[248,] 0.0000000 1.00000000
[249,] 1.0000000 0.00000000
[250,] 1.0000000 0.00000000
[251,] 0.2500000 0.75000000
[252,] 0.7500000 0.25000000
[253,] 0.7500000 0.25000000
[254,] 1.0000000 0.00000000
[255,] 1.0000000 0.00000000
[256,] 0.0000000 1.00000000
[257,] 0.0000000 1.00000000
[258,] 0.0000000 1.00000000
[259,] 0.9166667 0.08333333
[260,] 0.0000000 1.00000000
[261,] 0.7500000 0.25000000
[262,] 0.2500000 0.75000000
[263,] 0.2500000 0.75000000
[264,] 1.0000000 0.00000000
[265,] 0.2500000 0.75000000
[266,] 0.5000000 0.500000001.7 모형 평가
Caution! 모형 평가를 위해 Test Dataset에 대한 예측 class/확률 이 필요하며, 함수 predict()를 이용하여 생성한다.
# 예측 class 생성
knn.pred <- predict(knn.model,
newdata = titanic.ted.Imp[,-1], # Test Dataset including Only 예측 변수
type = "class") # 예측 class 생성
knn.pred %>%
as_tibble# A tibble: 266 × 1
value
<fct>
1 yes
2 no
3 no
4 no
5 no
6 no
7 yes
8 no
9 no
10 yes
# ℹ 256 more rows1.7.1 ConfusionMatrix
CM <- caret::confusionMatrix(knn.pred, titanic.ted.Imp$Survived,
positive = "yes") # confusionMatrix(예측 class, 실제 class, positive = "관심 class")
CMConfusion Matrix and Statistics
Reference
Prediction no yes
no 148 28
yes 16 74
Accuracy : 0.8346
95% CI : (0.7844, 0.8772)
No Information Rate : 0.6165
P-Value [Acc > NIR] : 7.118e-15
Kappa : 0.6422
Mcnemar's Test P-Value : 0.09725
Sensitivity : 0.7255
Specificity : 0.9024
Pos Pred Value : 0.8222
Neg Pred Value : 0.8409
Prevalence : 0.3835
Detection Rate : 0.2782
Detection Prevalence : 0.3383
Balanced Accuracy : 0.8140
'Positive' Class : yes
1.7.2 ROC 곡선
# 예측 확률 생성
test.knn.prob <- predict(knn.model,
newdata = titanic.ted.Imp[,-1], # Test Dataset including Only 예측 변수
type = "prob") # 예측 확률 생성
test.knn.prob %>%
as_tibble# A tibble: 266 × 2
no yes
<dbl> <dbl>
1 0 1
2 0.75 0.25
3 0.75 0.25
4 0.5 0.5
5 0.75 0.25
6 1 0
7 0.25 0.75
8 1 0
9 1 0
10 0 1
# ℹ 256 more rowstest.knn.prob <- test.knn.prob[,2] # "Survived = yes"에 대한 예측 확률
ac <- titanic.ted.Imp$Survived # Test Dataset의 실제 class
pp <- as.numeric(test.knn.prob) # 예측 확률을 수치형으로 변환1.7.2.1 Package “pROC”
pacman::p_load("pROC")
knn.roc <- roc(ac, pp, plot=T, col="gray") # roc(실제 class, 예측 확률)
auc <- round(auc(knn.roc),3)
legend("bottomright", legend = auc, bty = "n")
Caution! Package "pROC"를 통해 출력한 ROC 곡선은 다양한 함수를 이용해서 그래프를 수정할 수 있다.
# 함수 plot.roc() 이용
plot.roc(knn.roc,
col="gray", # Line Color
print.auc = TRUE, # AUC 출력 여부
print.auc.col = "red", # AUC 글씨 색깔
print.thres = TRUE, # Cutoff Value 출력 여부
print.thres.pch = 19, # Cutoff Value를 표시하는 도형 모양
print.thres.col = "red", # Cutoff Value를 표시하는 도형의 색깔
auc.polygon = TRUE, # 곡선 아래 면적에 대한 여부
auc.polygon.col = "gray90") # 곡선 아래 면적의 색깔
# 함수 ggroc() 이용
ggroc(knn.roc) +
annotate(geom = "text", x = 0.9, y = 1.0,
label = paste("AUC = ", auc),
size = 5,
color="red") +
theme_bw()
1.7.2.2 Package “Epi”
pacman::p_load("Epi")
# install_version("etm", version = "1.1", repos = "http://cran.us.r-project.org")
ROC(pp, ac, plot="ROC") # ROC(예측 확률, 실제 class) 
1.7.2.3 Package “ROCR”
pacman::p_load("ROCR")
knn.pred <- prediction(pp, ac) # prediction(예측 확률, 실제 class)
knn.perf <- performance(knn.pred, "tpr", "fpr") # performance(, "민감도", "1-특이도")
plot(knn.perf, col = "gray") # ROC Curve
perf.auc <- performance(knn.pred, "auc") # AUC
auc <- attributes(perf.auc)$y.values
legend("bottomright", legend = auc, bty = "n")
1.7.3 향상 차트
1.7.3.1 Package “ROCR”
knn.perf <- performance(knn.pred, "lift", "rpp") # Lift Chart
plot(knn.perf, main = "lift curve",
colorize = T, # Coloring according to cutoff
lwd = 2) 