caret包进行机器学习

caret 是一个用于创建机器学习工作流的一般包,方法使用如下：

library(caret)
library(kernlab)
data(iris)
head(iris)

使用createDataPartition创建一系列测试/训练分区。输入为预测结果的一列，和用于训练的数据集比例，一般用四分之三来训练，四分之一来测试，即p = 0.75。

intrain<-createDataPartition(iris$Species,p = 0.75,list = F)
head(intrain)
training<-iris[intrain,]
testing<-iris[-intrain,]

选用特定训练模型

modelFit<-train(Species~.,data=training,method = "pls",preProc = c("center", "scale"))
modelFit$finalModel

使用模型对测试数据集进行预测

predictions <- predict(modelFit,newdata=testing)
predictions 
confusionMatrix(predictions,testing$Species)

# Confusion Matrix and Statistics
# 
#             Reference
# Prediction   setosa versicolor virginica
#   setosa         12          0         0
#   versicolor      0          6         2
#   virginica       0          6        10
# 
# Overall Statistics
#                                           
#                Accuracy : 0.7778          
#                  95% CI : (0.6085, 0.8988)
#     No Information Rate : 0.3333          
#     P-Value [Acc > NIR] : 5.965e-08       
#                                           
#                   Kappa : 0.6667          
#                                           
#  Mcnemar's Test P-Value : NA              
# 
# Statistics by Class:
# 
#                      Class: setosa Class: versicolor
# Sensitivity                 1.0000            0.5000
# Specificity                 1.0000            0.9167
# Pos Pred Value              1.0000            0.7500
# Neg Pred Value              1.0000            0.7857
# Prevalence                  0.3333            0.3333
# Detection Rate              0.3333            0.1667
# Detection Prevalence        0.3333            0.2222
# Balanced Accuracy           1.0000            0.7083
#                      Class: virginica
# Sensitivity                    0.8333
# Specificity                    0.7500
# Pos Pred Value                 0.6250
# Neg Pred Value                 0.9000
# Prevalence                     0.3333
# Detection Rate                 0.2778
# Detection Prevalence           0.4444
# Balanced Accuracy              0.7917