在R中使用xgboost

假设X为训练数据，y为label，为0或者1.用xgboost建立分类模型代码如下
调用caret包中的createFolds方法，进行10倍交叉验证
最后画出AUC曲线

library(xgboost)
library(caret)
library(caTools)
library(pROC)

set.seed(123)
folds <- createFolds(y, k = 10)
cv = lapply(folds, function(x) {
  training_fold = X[-x, ]
  training_fold_y = y[-x]
  test_fold = X[x, ]
  test_fold_y = y[x]
  # we stick our XGBoost classifier in here
  classifier = xgboost(
  	data = as.matrix(training_fold), 
  	label = training_fold_y, 
  	nrounds = 100, 
  	verbose = 0,
  	params = list(objective = "binary:logistic", 
  				  eval_metric = "auc", 
  				  max_depth = 5,
  				  eta = 0.4, 
  				  min_child_weight = 7, 
  				  subsample = 0.8, 
  				  colsample_bytree = 0.9))
  y_predict = predict(classifier, newdata = as.matrix(test_fold)) # again need a matrix
  y_pred = (y_predict >= 0.5) # here we are setting up the binary outcome of 0 or 1
  roc <- roc(test_fold_y, y_predict)
  return(roc)
})

# plot
plot.roc(cv$Fold01, col = "#8DD3C7")
lines.roc(cv$Fold02, col = "#FFFFB3")
lines.roc(cv$Fold03, col = "#BEBADA")
lines.roc(cv$Fold04, col = "#FB8072")
lines.roc(cv$Fold05, col = "#80B1D3")
lines.roc(cv$Fold06, col = "#FDB462")
lines.roc(cv$Fold07, col = "#B3DE69")
lines.roc(cv$Fold08, col = "#FCCDE5")
lines.roc(cv$Fold09, col = "#D9D9D9")
lines.roc(cv$Fold10, col = "#BC80BD")
legend("bottomright", legend = paste0('Fold',1:10,': auc ',as.vector(unlist(lapply(cv, function(x) round(as.numeric(x$auc),2))))), 
       col = RColorBrewer::brewer.pal('Set3', n=10),
       lty = 1,
       box.lty=0, bg="transparent")