library(caret)library(earth) #利用包内的数据data(etitanic)#使用baseR中的model.matrixhead(model.matrix(survived ~ ., data = etitanic)[,-1])

#数据集etitantic中pclass变量有1st,2nd, 3rd三个水平，需要转换为虚拟变量#model.matrix()函数将pclass转换为三个变量并把第一个水平作为参考，同时生成一个常数列#使用caret包的dummyVars（）函数dummies <- dummyVars(survived ~ ., data = etitanic)head(predict(dummies, newdata = etitanic))

#preprocess函数#大海哥先介绍一下它的用法preProcess(x,  需要处理的数据集，数据集为数值型的矩阵或数据框method = c("center","scale"), 数据处理方#  thresh = 0.95, PCA的阈值pcaComp = NULL, PCA主成分个数na.remove = TRUE, 是否去除缺失值#   k = 5,   knn算法的k值knnSummary = mean, knn插值的方法outcome = NULL,  结局变量fudge = 0.2, 公差值numUnique = 3, Box-Cox变换需要多少个唯一值 verbose = FALSE, 是否显示处理过程 freqCut = 95/5, 最常见值与第二常见值的比值 uniqueCut = 10, 不同值占总样本数的百分比 cutoff = 0.9,   相关性的阈值 rangeBounds = c(0, 1), 指定范围内变换数据预处理方法method还有："BoxCox","YeoJohnson", "expoTrans", "center","scale", "range", "knnImpute","bagImpute", "medianImpute", "pca","ica", "spatialSign", "corr", "zv","nzv", and "conditionalX"
#然后我们找个示例演示一下吧library(AppliedPredictiveModeling)data(schedulingData)#引入示例数据str(schedulingData)

pp <- preProcess(schedulingData[, -8],          method = c("center", "scale", "YeoJohnson", "nzv")

#处理完成后，看看能不能拿来进行预测predict(pp, newdata =schedulingData[1:6, -8])

还可以进行数据拆分data(iris)iris <- na.omit(iris)set.seed(123)  # 设置随机种子，以确保结果可重复train_indices <- createDataPartition(iris $Species, p = 0.7, list = FALSE)train_data <-iris[train_indices, ]test_data <- iris[-train_indices, ]

#基本参数设定fitControl <- trainControl(## 10-fold CV                          method = "repeatedcv",                          number = 5,                          ## repeated ten times                          repeats = 3)# method：重采样方法，包括"boot", "boot632", "optimism_boot","boot_all", "cv", "repeatedcv","LOOCV", "LGOCV" (for repeated training/test splits),"none" (only fits one model to the entire training set),"oob" (only for random forest, bagged trees, bagged earth, bagged flexible discriminant analysis, or conditional tree forest models), timeslice,"adaptive_cv", "adaptive_boot" or"adaptive_LGOCV"#number用于设定交叉验证和重复交叉验证的折数、bootstrap抽样次数 #repeats 用于设定交叉验证的次数
#模型拟合set.seed(1)gbmFit1 <- train(factor(Species) ~ .,                  data = train_data,                  method = "gbm",                  trControl = fitControl,                 ## This last option is actually one                 ## for gbm() that passes through                 verbose = FALSE)

迭代次数，即树（在函数中调用）n.trees树的复杂性，称为interaction.depth学习速率：算法适应的速度，称为shrinkage节点中开始拆分的训练集样本的最小数量（n.minobsinnode)#定义调参过程gbmGrid <-  expand.grid(interaction.depth = c(1, 5, 9),                        n.trees = (1:20)*100,                        shrinkage =0.1,                       n.minobsinnode = 20)                       nrow(gbmGrid)

#开始调参set.seed(2)gbmFit2 <- train(factor(Species) ~ .,                  data = train_data,                  method = "gbm",                  trControl = fitControl,                 verbose = FALSE,                  tuneGrid = gbmGrid)gbmFit2$bestTune  #调优结果#这一步需要消耗一点点时间哦！

#可以看到，给出了调优结果gbmFit2$finalModel #最终模型

#我们还可以根据结果，绘制重采样曲线ggplot(gbmFit2) #以Accuracy为基准

#也可以选择ggplot(gbmFit2,metric = "Kappa") #以Kappa为基准

#这一步也是我们最重要的一步了predict(gbmFit2, newdata = head(test_data))#默认预测分类

#预测成功啦！#还可以预测概率哦！predict(gbmFit2,         newdata = head(test_data), type = "prob")#预测概率