library(dplyr)
## Warning: 패키지 'dplyr'는 R 버전 4.1.3에서 작성되었습니다
##
## 다음의 패키지를 부착합니다: 'dplyr'
## The following objects are masked from 'package:stats':
##
## filter, lag
## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, union
library(recipes)
## Warning: 패키지 'recipes'는 R 버전 4.1.3에서 작성되었습니다
##
## 다음의 패키지를 부착합니다: 'recipes'
## The following object is masked from 'package:stats':
##
## step
library(caret)
## Warning: 패키지 'caret'는 R 버전 4.1.3에서 작성되었습니다
## 필요한 패키지를 로딩중입니다: ggplot2
## Warning: 패키지 'ggplot2'는 R 버전 4.1.3에서 작성되었습니다
## 필요한 패키지를 로딩중입니다: lattice
# 작업형 기출2
# 보험 가입 확률을 묻는 문제
# index(id) 컬럼 포함 총 10개의 컬럼으로 되어있으며,
# train 데이터로 1491건, test 데이터로 496건의 자료를 제공
# 데이터에 index 컬럼이 숫자형으로 되어있어 모델에 포함하여 index 변수를
# 활용하지 않음
# 1 데이터 구조 파악, 데이터 분할
df<-read.csv("travel_data.csv")
# train/test: 0.75:0.25
train_list<-createDataPartition(y=df$TravelInsurance,p=0.75,list=FALSE)
df_train<-df[train_list,]
df_test<-df[-train_list,]
NROW(df_train)
## [1] 1491
NROW(df_test)
## [1] 496
df_train %>% glimpse
## Rows: 1,491
## Columns: 10
## $ INDEX <int> 1, 3, 5, 6, 7, 8, 10, 12, 14, 16, 17, 18, 19, 21, ~
## $ Age <int> 31, 28, 25, 31, 31, 28, 31, 32, 31, 28, 28, 29, 34~
## $ Employment.Type <chr> "Private Sector/Self Employed", "Private Sector/Se~
## $ GraduateOrNot <chr> "Yes", "Yes", "No", "Yes", "Yes", "Yes", "Yes", "Y~
## $ AnnualIncome <int> 1250000, 700000, 1150000, 1300000, 1350000, 145000~
## $ FamilyMembers <int> 7, 3, 4, 4, 3, 6, 9, 6, 3, 4, 7, 5, 2, 3, 4, 9, 4,~
## $ ChronicDiseases <int> 0, 1, 0, 0, 0, 1, 1, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0,~
## $ FrequentFlyer <chr> "No", "No", "No", "No", "Yes", "Yes", "No", "No", ~
## $ EverTravelledAbroad <chr> "No", "No", "No", "No", "Yes", "Yes", "No", "No", ~
## $ TravelInsurance <int> 0, 0, 0, 0, 1, 1, 0, 1, 0, 0, 1, 1, 1, 1, 0, 1, 0,~
df_test %>% glimpse
## Rows: 496
## Columns: 10
## $ INDEX <int> 0, 2, 4, 9, 11, 13, 15, 20, 24, 28, 32, 37, 40, 46~
## $ Age <int> 31, 34, 28, 33, 26, 31, 34, 28, 31, 26, 25, 34, 25~
## $ Employment.Type <chr> "Government Sector", "Private Sector/Self Employed~
## $ GraduateOrNot <chr> "Yes", "Yes", "Yes", "Yes", "Yes", "Yes", "Yes", "~
## $ AnnualIncome <int> 400000, 500000, 700000, 800000, 1400000, 1500000, ~
## $ FamilyMembers <int> 6, 4, 8, 3, 5, 6, 7, 6, 7, 4, 4, 4, 3, 4, 3, 2, 7,~
## $ ChronicDiseases <int> 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0,~
## $ FrequentFlyer <chr> "No", "No", "Yes", "Yes", "Yes", "Yes", "No", "Yes~
## $ EverTravelledAbroad <chr> "No", "No", "No", "No", "Yes", "Yes", "No", "No", ~
## $ TravelInsurance <int> 0, 1, 0, 0, 1, 1, 0, 0, 0, 0, 1, 0, 1, 1, 0, 0, 0,~
# 실제 시험에서는 INDEX 의미없는 숫자
df_train %>% mutate(index="train")->df_train
df_test %>% mutate(index='test')->df_test
bind_rows(df_train,df_test)->full
full %>% head
## INDEX Age Employment.Type GraduateOrNot AnnualIncome
## 1 1 31 Private Sector/Self Employed Yes 1250000
## 2 3 28 Private Sector/Self Employed Yes 700000
## 3 5 25 Private Sector/Self Employed No 1150000
## 4 6 31 Government Sector Yes 1300000
## 5 7 31 Private Sector/Self Employed Yes 1350000
## 6 8 28 Private Sector/Self Employed Yes 1450000
## FamilyMembers ChronicDiseases FrequentFlyer EverTravelledAbroad
## 1 7 0 No No
## 2 3 1 No No
## 3 4 0 No No
## 4 4 0 No No
## 5 3 0 Yes Yes
## 6 6 1 Yes Yes
## TravelInsurance index
## 1 0 train
## 2 0 train
## 3 0 train
## 4 0 train
## 5 1 train
## 6 1 train
# 2 목표변수, 기타변수 변환
full$TravelInsurance<-ifelse(full$TravelInsurance==0,"미가입","가입")
full$TravelInsurance<-as.factor(full$TravelInsurance)
full$GraduateOrNot<-as.factor(full$GraduateOrNot)
full$FrequentFlyer<-as.factor(full$FrequentFlyer)
full$EverTravelledAbroad<-as.factor(full$EverTravelledAbroad)
# 3 결측값 확인
colSums(is.na(full))
## INDEX Age Employment.Type GraduateOrNot
## 0 0 0 0
## AnnualIncome FamilyMembers ChronicDiseases FrequentFlyer
## 0 0 0 0
## EverTravelledAbroad TravelInsurance index
## 0 0 0
# 4 데이터 전처리
recipe(TravelInsurance~.,data=full) %>% step_YeoJohnson(Age,AnnualIncome,FamilyMembers) %>%
step_center(Age,AnnualIncome,FamilyMembers) %>%
step_scale(Age,AnnualIncome,FamilyMembers) %>% prep() %>% juice()->data
data %>%filter(index=="train") %>% select(-index)->train
data %>%filter(index=='test') %>% select(-index)->test
ctrl<-trainControl(method="cv",summaryFunction = twoClassSummary,
classProbs = TRUE)
train(TravelInsurance~.,data=train,
method='rpart',metric="ROC",
trControl=ctrl)->rpfit
rpfit
## CART
##
## 1491 samples
## 9 predictor
## 2 classes: '가입', '미가입'
##
## No pre-processing
## Resampling: Cross-Validated (10 fold)
## Summary of sample sizes: 1342, 1341, 1341, 1343, 1342, 1342, ...
## Resampling results across tuning parameters:
##
## cp ROC Sens Spec
## 0.00929368 0.7751709 0.5762055 0.9653509
## 0.06226766 0.7342430 0.4981481 0.9663925
## 0.39591078 0.5516366 0.1137666 0.9895066
##
## ROC was used to select the optimal model using the largest value.
## The final value used for the model was cp = 0.00929368.
# rpart cp complexity parameter 이 포인트에서 분할가지수를 선택한다.
confusionMatrix(rpfit)
## Cross-Validated (10 fold) Confusion Matrix
##
## (entries are percentual average cell counts across resamples)
##
## Reference
## Prediction 가입 미가입
## 가입 20.8 2.2
## 미가입 15.3 61.7
##
## Accuracy (average) : 0.8249
test %>% glimpse
## Rows: 496
## Columns: 10
## $ INDEX <int> 0, 2, 4, 9, 11, 13, 15, 20, 24, 28, 32, 37, 40, 46~
## $ Age <dbl> 0.5290744, 1.4182229, -0.5150177, 1.1367593, -1.31~
## $ Employment.Type <fct> Government Sector, Private Sector/Self Employed, P~
## $ GraduateOrNot <fct> Yes, Yes, Yes, Yes, Yes, Yes, Yes, Yes, Yes, Yes, ~
## $ AnnualIncome <dbl> -1.4862039, -1.1612874, -0.5655285, -0.2877223, 1.~
## $ FamilyMembers <dbl> 0.8341005, -0.3760749, 1.7698222, -1.1525825, 0.27~
## $ ChronicDiseases <int> 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0,~
## $ FrequentFlyer <fct> No, No, Yes, Yes, Yes, Yes, No, Yes, Yes, Yes, No,~
## $ EverTravelledAbroad <fct> No, No, No, No, Yes, Yes, No, No, Yes, No, No, No,~
## $ TravelInsurance <fct> 미가입, 가입, 미가입, 미가입, 가입, 가입, 미가입, ~
predict(rpfit,test,type='prob')->rffit1
predict(rpfit,test,type="raw")->rffit2
confusionMatrix(rffit2,test$TravelInsurance)
## Confusion Matrix and Statistics
##
## Reference
## Prediction 가입 미가입
## 가입 106 11
## 미가입 66 313
##
## Accuracy : 0.8448
## 95% CI : (0.8098, 0.8755)
## No Information Rate : 0.6532
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.6295
##
## Mcnemar's Test P-Value : 7.561e-10
##
## Sensitivity : 0.6163
## Specificity : 0.9660
## Pos Pred Value : 0.9060
## Neg Pred Value : 0.8259
## Prevalence : 0.3468
## Detection Rate : 0.2137
## Detection Prevalence : 0.2359
## Balanced Accuracy : 0.7912
##
## 'Positive' Class : 가입
##
importance<-varImp(rpfit,scale=FALSE)
print(importance)
## rpart variable importance
##
## Overall
## AnnualIncome 198.329
## EverTravelledAbroadYes 124.682
## FamilyMembers 68.946
## Age 38.981
## FrequentFlyerYes 37.468
## Employment.TypePrivate Sector/Self Employed 18.536
## INDEX 5.253
## GraduateOrNotYes 0.000
## ChronicDiseases 0.000
## `Employment.TypePrivate Sector/Self Employed` 0.000
library(pROC)
## Type 'citation("pROC")' for a citation.
##
## 다음의 패키지를 부착합니다: 'pROC'
## The following objects are masked from 'package:stats':
##
## cov, smooth, var
rffit2_num<-as.numeric(rffit2)
result<-roc(test$TravelInsurance,rffit2_num)
## Setting levels: control = 가입, case = 미가입
## Setting direction: controls < cases
result
##
## Call:
## roc.default(response = test$TravelInsurance, predictor = rffit2_num)
##
## Data: rffit2_num in 172 controls (test$TravelInsurance 가입) < 324 cases (test$TravelInsurance 미가입).
## Area under the curve: 0.7912
result$auc
## Area under the curve: 0.7912