coxPH

---

0.1 Cox Proportional Hazards (CoxPH) _Supervised

H2O 를 활용한 Cox Proportional Hazards (CoxPH)

[참조 1] http://docs.h2o.ai/h2o/latest-stable/h2o-docs/data-science/coxph.html

생존분석에서 쓰는 통계모형으로 준모수적 방법을 이용해 생존함수를 추정(*콕스모델)

모형의 이름인 비례위험은 시간에 상관없이 어떤 변수의 위험비는 항상 일정하다는 모형의 기본가정에서 비롯

---

0.1.1 packages

library(h2o)

## Warning: package 'h2o' was built under R version 4.0.3

# 초기 준비
h2o.init()

##  Connection successful!
## 
## R is connected to the H2O cluster: 
##     H2O cluster uptime:         4 hours 29 minutes 
##     H2O cluster timezone:       Asia/Seoul 
##     H2O data parsing timezone:  UTC 
##     H2O cluster version:        3.32.0.1 
##     H2O cluster version age:    25 days  
##     H2O cluster name:           H2O_started_from_R_user_uho906 
##     H2O cluster total nodes:    1 
##     H2O cluster total memory:   3.97 GB 
##     H2O cluster total cores:    4 
##     H2O cluster allowed cores:  4 
##     H2O cluster healthy:        TRUE 
##     H2O Connection ip:          localhost 
##     H2O Connection port:        54321 
##     H2O Connection proxy:       NA 
##     H2O Internal Security:      FALSE 
##     H2O API Extensions:         Amazon S3, Algos, AutoML, Core V3, TargetEncoder, Core V4 
##     R Version:                  R version 4.0.2 (2020-06-22)

---

0.1.2 데이터 준비

# Import the heart dataset into H2O:
heart <- h2o.importFile("http://s3.amazonaws.com/h2o-public-test-data/smalldata/coxph_test/heart.csv")

## 
  |                                                                            
  |                                                                      |   0%
  |                                                                            
  |======================================================================| 100%

heart

##   start stop event        age      year surgery transplant id
## 1     0   50     1 -17.155373 0.1232033       0          0  1
## 2     0    6     1   3.835729 0.2546201       0          0  2
## 3     0    1     0   6.297057 0.2655715       0          0  3
## 4     1   16     1   6.297057 0.2655715       0          1  3
## 5     0   36     0  -7.737166 0.4900753       0          0  4
## 6    36   39     1  -7.737166 0.4900753       0          1  4
## 
## [172 rows x 8 columns]

---

0.1.3 전처리

# Split the dataset into a train and test set:
heart_split <- h2o.splitFrame(data = heart, ratios = 0.8, seed = 1234)
train <- heart_split[[1]]
test <- heart_split[[2]]

---

0.1.4 모델링

# Build and train the model:
heart_coxph <- h2o.coxph(x = "age",
                         event_column = "event",
                         start_column = "start",
                         stop_column = "stop",
                         ties = "breslow",
                         training_frame = train)

## 
  |                                                                            
  |                                                                      |   0%
  |                                                                            
  |======================================================================| 100%

heart_coxph

## Model Details:
## ==============
## 
## H2OCoxPHModel: coxph
## Model ID:  CoxPH_model_R_1604363756625_4 
## Call:
## Surv(start, stop, event) ~ age
## 
##        coef exp(coef) se(coef)    z     p
## age 0.02257   1.02282  0.01514 1.49 0.136
## 
## Likelihood ratio test=2.42  on 1 df, p=0.1196
## n= 138, number of events= 59

# Eval performance:
perf <- h2o.performance(heart_coxph)
perf

## H2OCoxPHMetrics: coxph
## ** Reported on training data. **

---

0.1.5 예측

# Generate predictions on a test set (if necessary):
predict <- h2o.predict(heart_coxph, newdata = test)

## 
  |                                                                            
  |                                                                      |   0%
  |                                                                            
  |======================================================================| 100%

predict

##           lp
## 1 0.20941803
## 2 0.13206056
## 3 0.06687515
## 4 0.21565853
## 5 0.21565853
## 6 0.20299217
## 
## [34 rows x 1 column]

---