剪枝决策树

R 中的信用风险建模

Lore Dirick

Manager of Data Science Curriculum at Flatiron School

大型决策树的问题

过于复杂：不再清晰
在测试集上过拟合
解决方案：用 printcp()、plotcp() 进行剪枝

Printcp 与 tree_undersample

printcp(tree_undersample)

Classification tree:
rpart(formula = loan_status ~ ., data = undersampled_training_set, method = "class",
 control = rpart.control(cp = 0.001))
Variables actually used in tree construction:
age    annual_inc     emp_cat     grade    home_ownership   ir_cat     loan_amnt     
Root node error: 2190/6570 = 0.33333
n= 6570 
        CP    nsplit  rel error   xerror      xstd
1  0.0059361      0    1.00000   1.00000   0.017447
2  0.0044140      4    0.97443   0.99909   0.017443
3  0.0036530      7    0.96119   0.98174   0.017366
4  0.0031963      8    0.95753   0.98904   0.017399
               ...  
16 0.0010654     76    0.84247   1.02511   0.017554
17 0.0010000     79    0.83927   1.02511   0.017554

Plotcp 与 tree_undersample

屏幕截图 2020-06-22 5.57.10 PM.png

Plotcp 与 tree_undersample

屏幕截图 2020-06-22 5.56.53 PM.png

$CP = 0.003653$

绘制剪枝后的树

屏幕截图 2020-06-23 6.16.04 PM.png

ptree_undersample=prune(tree_undersample,
                        cp = 0.003653)

plot(ptree_undersample,
     uniform=TRUE)

text(ptree_undersample)

绘制剪枝后的树

屏幕截图 2020-06-23 6.15.42 PM.png

ptree_undersample=prune(tree_undersample,
                        cp = 0.003653)

plot(ptree_undersample,
     uniform=TRUE)

text(ptree_undersample,
     use.n=TRUE)

rpart.plot 包中的 prp()

屏幕截图 2020-06-22 6.05.09 PM.png

library(rpart.plot)
prp(ptree_undersample)

rpart.plot 包中的 prp()

屏幕截图 2020-06-22 6.04.33 PM.png

library(rpart.plot)
prp(ptree_undersample, extra = 1)

开始练习！

R 中的信用风险建模