无论不同的解释变量(R)如何，Xgboost都会得到相同的预测结果

Question

我有以下训练和测试数据，我正在extreme gradient boosting algorithm (Xgboost)上对其执行，使用具有预测能力的数据集，尽管无论预测器值如何，模型的预测都是恒定的：

library(data.table)
require(xgboost)
library(Matrix)

sparse_matrix_train = sparse.model.matrix(clicked~.-1, data = train)
sparse_matrix_train2 = sparse.model.matrix(clicked~., data = test)
bst <- xgboost(data = sparse_matrix2, label = test2$clicked, max.depth = 2,
               eta = 0.3, nthread = 20, nround = 5,objective = "binary:logistic")
sparse_matrix_test = sparse.model.matrix(~., data = test)
test$pred_res<- predict(bst, sparse_matrix_test)

请注意，test$pred_res具有相同的值：

test$pred_res
 [1] 0.2937567 0.2937567 0.2937567 0.2937567 0.2937567 0.2937567 0.2937567 0.2937567 0.2937567
[10] 0.2937567

有人能详细解释一下这个问题吗？

数据

train <- structure(list(clicked = c(0L, 0L, 1L, 0L, 0L, 0L, 0L, 0L, 0L, 
  1L), prediction = c(0.116741800132, 0.148609212057, 0.27496222789, 
  0.0284488735744, 0.234446058524, 0.405107420156, 0.122376142849, 
  0.126600568099, 0.0636921765101, 0.385909171908), pred_res = c(0.293756693601608, 
  0.293756693601608, 0.293756693601608, 0.343792468309402, 0.293756693601608, 
  0.293756693601608, 0.293756693601608, 0.293756693601608, 0.293756693601608, 
  0.293756693601608)), .Names = c("clicked", "prediction", "pred_res"),
  row.names = c(NA, -10L), class = c("data.table", "data.frame"),
  .internal.selfref = <pointer: 0x1547c18>)

test <- structure(list(prediction = c(0.0553382017171, 0.158500277487, 
  0.155315011347, 0.118212821075, 0.0795492263212, 0.0272566752275, 
  0.159516005352, 0.218685440776, 0.0562459472969, 0.178293801444),
  pred_res = c(0.293756693601608, 0.293756693601608, 0.293756693601608, 
  0.293756693601608, 0.293756693601608, 0.293756693601608, 0.293756693601608, 
  0.293756693601608, 0.293756693601608, 0.293756693601608)),
  .Names = c("prediction","pred_res"), row.names = c(NA, -10L),
  class = c("data.table","data.frame"), .internal.selfref = <pointer: 0x1547c18>)

Answer 1

我加载了你的数据并运行了你的代码，但得到了：

“xgb.get.DMatrix(数据，标签)中出错:找不到对象'sparse_matrix2‘”

HTH，cousin_pete