R语言逻辑回归 logistic regression

R语言逻辑回归代码示例

rm(list=ls())
require(kernlab)
data(spam)
data <- spam
n <- nrow(spam)
id <- sample(1:n, floor(n*0.5))
train <- data[id,]
test <- data[-id,]

mdl <- glm(formula = type~.,
			family = binomial(link = "logit"), 
			data = train)

# 直接返回线性回归的值z
z <- predict(mdl, test, type = "link")


# 直接返回概率p
p <- predict(mdl, test, type = "response")

# 根据公式，z与p关系为
# p <- 1/(1+exp(-z))

# 0为test$type中levels=0的label，这里是nonspam
pred <- as.numeric(p>0.5)  # 这里的概率可以根据需要修改

# 转成与原标签统一的factor
pred <- factor(pred, levels=  c(0,1), labels = c("nonspam","spam"))
table(test$type,pred,dnn = c("Labels","prediction"))


###########################################
 输出的混淆矩阵为
 
         prediction
Labels    nonspam spam
  nonspam    1328   74
  spam        105  794