Question

在插入符号中如何保存最小尺寸模型。在此示例中，gbmFit1包含gbmFit1$trainingData。保存gbmFit1会保存所有此类变量。由于我的训练数据很大，我想摆脱所有这些额外的变量，并希望以最小的尺寸保存模型。

  library(mlbench)
  library(caret)
  data(Sonar)
  x <- Sonar[, colnames(Sonar)!="Class"]
  y <- Sonar$Class
  gbmFit1 <- train(x,y, method = "gbm", verbose = FALSE)
  predict(gbmFit1, x[1:10, ]) #predict for 10 samples
  ##[1] R R R R R R R R R R
  ##Levels: M R
  dim(gbmFit1$trainingData) 
  #[1] 208  61

仅使用predict(gbmFit1$finalModel, x[1:10, ])会出错：

predict(gbmFit1$finalModel, x[1:10, ])
##Error in paste("Using", n.trees, "trees...\n") : 
##argument "n.trees" is missing, with no default

Answer 1

我认为应该这样做：

library(mlbench)
library(caret)
  data(Sonar)
  x <- Sonar[, colnames(Sonar)!="Class"]
  y <- Sonar$Class
tc1 <- trainControl(returnData = F)  # tells caret not to save training data.
  gbmFit1 <- train(x,y, method = "gbm", verbose = FALSE, trControl = tc1)
  predict(gbmFit1$finalModel, x[1:10, ], gbmFit1$finalModel$tuneValue$n.trees) # passes n.trees value to gbm.

您可能希望在此处阅读插入符号中的trainControl功能：https://topepo.github.io/caret/model-training-and-tuning.html#control

插入符号保存最小尺寸模型

1 个答案: