R语言-决策树

C5.0

> ###########################决策树
> ########C5.0
> setwd("/Users/yaozhilin/Downloads/R_edu/data")
> orgdata<-read.csv("Allelectronics.csv")
> summary(orgdata)
      age       income            student          credit_rating      buys_computer     
 Min.   :1   Length:14          Length:14          Length:14          Length:14         
 1st Qu.:1   Class :character   Class :character   Class :character   Class :character  
 Median :2   Mode  :character   Mode  :character   Mode  :character   Mode  :character  
 Mean   :2                                                                              
 3rd Qu.:3                                                                              
 Max.   :3                                                                              
> #C5.0只能处理分类变量
> orgdata<-as.data.frame(lapply(orgdata[,1:5],as.factor))
> library(C50)
> #编辑决策树的控制数据
> #minCases控制类的样本量，CF置信因子越大，模型越大，winnow是否筛选变量（做xy的相关检验）
> tr<-C5.0Control(minCases = 1,CF=0.95,winnow = FALSE,noGlobalPruning = TRUE)
> model<-C5.0(buys_computer~.,data = orgdata,trials=1,rules=FALSE,control=tr)
> plot(model)

> #生成规则
> rules<-C5.0(buys_computer~.,data = orgdata,trials=1,rules=TRUE,control=tr)
> summary(rules)

Call:
C5.0.formula(formula = buys_computer ~ ., data = orgdata, trials = 1, rules = TRUE,
 control = tr)


C5.0 [Release 2.07 GPL Edition]      Wed Nov  4 15:12:38 2020
-------------------------------

Class specified by attribute `outcome'

Read 14 cases (5 attributes) from undefined.data

Rules:

Rule 1: (3, lift 2.2)
    age = 1
    student = no
    ->  class no  [0.800]

Rule 2: (2, lift 2.1)
    age = 3
    credit_rating = excellent
    ->  class no  [0.750]

Rule 3: (4, lift 1.3)
    age = 2
    ->  class yes  [0.833]

Rule 4: (3, lift 1.2)
    age = 3
    credit_rating = fair
    ->  class yes  [0.800]

Rule 5: (7/1, lift 1.2)
    student = yes
    ->  class yes  [0.778]

Default class: yes


Evaluation on training data (14 cases):

            Rules     
      ----------------
        No      Errors

         5    1( 7.1%)   <<


       (a)   (b)    <-classified as
      ----  ----
         4     1    (a): class no
               9    (b): class yes


    Attribute usage:

     85.71%    age
     71.43%    student
     35.71%    credit_rating


Time: 0.0 secs

cart方法

######cart实现决策树
library(rpart)
library(rpart.plot)
#minsplit:每个节点中最小样本量，minbucket:节点中所含样本最小数，cp:复杂参数，通常是阈值。
tc<-rpart.control(minsplit = 1,minbucket = 1,cp=0.001,maxdepth = 6,xval = 10)
cmodel<-rpart(buys_computer~.,orgdata,parms = list(split="gini"),
              method = "class",control = tc)
rpart.plot(cmodel,branch=1,extra=106,under=TRUE,faclen=0,cex=0.8)

#进行剪枝
cmodel_p<-prune(cmodel,cp=0.3)
rpart.plot(cmodel_p)