數(shù)據(jù)分析之美：決策樹R語言實(shí)現(xiàn)

R語言實(shí)現(xiàn)決策樹

1.準(zhǔn)備數(shù)據(jù)
[plain] view plain copy
    > install.packages("tree")  
    > library(tree)  
    > library(ISLR)  
    > attach(Carseats)  
    > High=ifelse(Sales<=8,"No","Yes") //set high values by sales data to calssify  
    > Carseats=data.frame(Carseats,High) //include the high data into the data source  
    > fix(Carseats) 
2.生成決策樹
[plain] view plain copy

    > tree.carseats=tree(High~.-Sales,Carseats)  
    > summary(tree.carseats)  

[plain] view plain copy
    //output training error is 9%  
    Classification tree:  
    tree(formula = High ~ . - Sales, data = Carseats)  
    Variables actually used in tree construction:  
    [1] "ShelveLoc"   "Price"       "Income"      "CompPrice"   "Population"   
    [6] "Advertising" "Age"         "US"           
    Number of terminal nodes:  27   
    Residual mean deviance:  0.4575 = 170.7 / 373   
    Misclassification error rate: 0.09 = 36 / 400  
3. 顯示決策樹
[plain] view plain copy

    > plot(tree . carseats )  
    > text(tree .carseats ,pretty =0) 
4.Test Error

[plain] view plain copy

    //prepare train data and test data  
    //We begin by using the sample() function to split the set of observations sample() into two halves, by selecting a random subset of 200 observations out of the original 400 observations.   
    > set . seed (1)  
    > train=sample(1:nrow(Carseats),200)  
    > Carseats.test=Carseats[-train,]  
    > High.test=High[-train]  
    //get the tree model with train data  
    > tree. carseats =tree (High~.-Sales , Carseats , subset =train )  
    //get the test error with tree model, train data and predict method  
    //predict is a generic function for predictions from the results of various model fitting functions.  
    > tree.pred = predict ( tree.carseats , Carseats .test ,type =" class ")  
    > table ( tree.pred ,High. test)  
    High. test  
    tree. pred No Yes  
    No 86 27  
    Yes 30 57  
    > (86+57) /200  
    [1] 0.715 

5.決策樹剪枝
[plain] view plain copy

    /**  
    Next, we consider whether pruning the tree might lead to improved results. The function cv.tree() performs cross-validation in order to cv.tree() determine the optimal level of tree complexity; cost complexity pruning is used in order to select a sequence of trees for consideration.   
      
    For regression trees, only the default, deviance, is accepted. For classification trees, the default is deviance and the alternative is misclass (number of misclassifications or total loss).  
    We use the argument FUN=prune.misclass in order to indicate that we want the classification error rate to guide the cross-validation and pruning process, rather than the default for the cv.tree() function, which is deviance.   
      
    If the tree is regression tree,   
    > plot(cv. boston$size ,cv. boston$dev ,type=’b ’)  
    */  
    > set . seed (3)  
    > cv. carseats =cv. tree(tree .carseats ,FUN = prune . misclass ,K=10)  
    //The cv.tree() function reports the number of terminal nodes of each tree considered (size) as well as the corresponding error rate(dev) and the value of the cost-complexity parameter used (k, which corresponds to α.  
    > names (cv. carseats )  
    [1] " size" "dev " "k" " method "  
    > cv. carseats  
    $size //the number of terminal nodes of each tree considered  
    [1] 19 17 14 13 9 7 3 2 1  
    $dev  //the corresponding error rate  
    [1] 55 55 53 52 50 56 69 65 80  
    $k  // the value of the cost-complexity parameter used  
    [1] -Inf 0.0000000 0.6666667 1.0000000 1.7500000  
    2.0000000 4.2500000  
    [8] 5.0000000 23.0000000  
    $method   //miscalss for classification tree  
    [1] " misclass "  
    attr (," class ")  
    [1] " prune " "tree. sequence "  

[plain] view plain copy

    //plot the error rate with tree node size to see whcih node size is best  
    > plot(cv. carseats$size ,cv. carseats$dev ,type=’b ’)  
      
    /**  
    Note that, despite the name, dev corresponds to the cross-validation error rate in this instance. The tree with 9 terminal nodes results in the lowest cross-validation error rate, with 50 cross-validation errors. We plot the error rate as a function of both size and k.  
    */  
    > prune . carseats = prune . misclass ( tree. carseats , best =9)  
    > plot( prune . carseats )  
    > text( prune .carseats , pretty =0)  
      
    //get test error again to see whether the this pruned tree perform on the test data set  
    > tree.pred = predict ( prune . carseats , Carseats .test , type =" class ")  
    > table ( tree.pred ,High. test)  
    High. test  
    tree. pred No Yes  
    No 94 24  
    Yes 22 60  
    > (94+60) /200  
    [1] 0.77