0
对于我的每个模型,我想要为总数MSE
创建boxplots
以显示模型误差的变化性。为了生成显示错误分布的boxplot
,多次抽样和拟合模型的最佳方法是什么?在R中的多个样本上绘制多个模型的MSE
对于每个模型,我将生成所有数据(列车和测试)的预测。然后我将根据预测计算出列车MSE
和测试MSE
的subsetting
。以下函数将同时计算MSE
两个值。
是否有更好的方法来重新采样和plot
每个模型的MSE每个重新采样?任何帮助,将不胜感激。
calcMSE = function(model,modelLabel,dataSet,trainIdx,newX=NULL)
{
# The predict method for glmnet will need to be called differently from the
# other predict methods.
if ("glmnet" %in% class(model)) {
predVals = predict(model,newX,type="response")
} else {
predVals = predict(model,data)
}
MSE = list(
name = modelLabel,
train = mean((predVals[trainIdx] - data$y[trainIdx])^2),
test = mean((predVals[-trainIdx] - data$y[-trainIdx])^2)
)
return(MSE)
}
modelMSEs = data.frame(Model = rep(NA,8),Train.MSE = rep(NA,8),Test.MSE = rep(NA,8))
modelMSEs[1,] = calcMSE(modelA1,"A1",Data,trainIdx)
modelMSEs[2,] = calcMSE(modelA2,"A2",Data,trainIdx)
modelMSEs[3,] = calcMSE(modelB1,"B1",Data,trainIdx)
modelMSEs[4,] = calcMSE(modelB2,"B2",Data,trainIdx)
modelMSEs[5,] = calcMSE(modelC1,"C1",Data,trainIdx)
modelMSEs[6,] = calcMSE(modelC2,"C2",Data,trainIdx)
print(modelMSEs)