README.md

In topycyao/Xsurv: A package used to model survival data from advanced machine learning algorithms

Xsurv

This package implements XGBoost and LightGBM in survival data with partiallikelihood function of cox ph model and smoothed C-index as loss functions. gbm and random forests are also provided in Xsurv function to fit the model. Additional survival information can be found with XGB and LGB models,including simple survival tree,risk level and SHAP importance.

Installation

Please install from Github:

devtools::install_github("topycyao/Xsurv")

Examples

library(Xsurv)
library(survival)
#fit the data 
data('cancer')
View(cancer)

mydata<-(cancer[,-1])

datay_train<-mydata[,c(1,2)]
datax_train<-mydata[,-c(1,2)]

xs<-Xsurv(datax_train,datay_train,top_n = 5,cp=0.01) #cp is the complexity of the fitted survival tree

#other options can be used 
#xs_lgb<-Xsurv(datax_train,datay_train,option='lgb',method='C',top_n=5)
#xs_gbm<-Xsurv(datax_train,datay_train,option='gbm')
#xs_rf<-Xsurv(datax_train,datay_train,option='rf')

#model can be trained automatically by Xsurv.cv
#xs.cv<-Xsurv.cv(datax_train,datay_train,top_n = 5)

model analysis

SHAP importance

shap=xs$SHAP
shap

Suvival tree

xm<-xs$mod
xtree<-xs$tree
tr1<-xtree$tree1
plot(tr1)

Risk analysis

library(survminer)
risk=xs$risk
#plot the kaplan-meier curve for different risk levels
risk_data=risk$data
fit=survival::survfit(Surv(time,status)~risk,data=risk_data)
ggsurvplot(fit,pval = TRUE,palette = c('coral','burlywood1','cadetblue1'),size=2,
                          legend=c(0.85,0.85),legend.title='',font.x=c(15,"plain","black"),
                           font.y=c(18,"plain","black"))

Data generation

sim_dat1<-Xsurv_sim_data(size=500,dim=20,lambda=2,vu=1,
                         c_rate=0.3)
sim_dat2<-Xsurv_sim_data(size=500,dim=20,lambda=2,vu=2,
                         c_rate=0.3)    
beta=c(rep(1,5),rep(0,15))
sim_dat3<-Xsurv_sim_data(size=500,dim=20,lambda=2,vu=2,beta=beta,
                         c_rate=0.3) 

Reference

Li,K. et al. Efficient gradient boosting for prognostic biomaker discovery

topycyao/Xsurv documentation built on Aug. 6, 2022, 9:06 p.m.