1.收集数据
credit <- read.csv("credit.csv")
2.探索数据(略)
3.准备数据
4.基于数据训练模型a.将数据随机化set.seed(12345)credit_rand <- credit[order(runif(1000)), ]
b.将随机数据拆分为90%的训练数据和10%的测试数据credit_train <- credit_rand[1:900,]credit_test <- credit_rand[901:1000,]
c.测试训练数据和准备数据的违约比例是否均衡prop.table(table(credit_train$default))
prop.table(table(credit_test$default))
install.packages("C50")
library(C50)
credit_model <- C5.0(credit_train[-17], credit_train$default) //建立决策分类器
5.预测模型和性能检测
credit_pred <- predict(credit_model, credit_test) //预测
测试结果
library(gmodels)
CrossTable(credit_test$default, credit_pred,
prop.chisq=FALSE,prop.c=FALSE,prop.r=FALSE,
dnn=c('actual default', 'predicted default'))
6.提高决策树的准确性
a.加入自适应增强算法credit_boost10 <- C5.0(credit_train[-21], credit_train$default, trials=10)
b.犯一些比其他错误更严重的错误
error_cost<-matrix(c(0,1,4,0), nrow=2) //建立代价矩阵,行为预测值,列为实际值
credit_cost < C5.0(credit_train[-21], credit_train$default, costs=error_cost)
Hadoop大数据云计算技术群: 342391093