利用机器学习预测乳腺癌

rm(list=ls())
x <- read.csv("breast.csv", row.names=1)  
head(x)
x <- x[-1]
class(x$class)
x$class <- as.factor(x$class)
levels(x$class)
levels(x$class) <- c('benign', 'nalignant')
head(x)
pheatmap(x[-10],cluster_rows = F,cluster_cols = F)


nrow(x)
#进行抽样，df分成df.train与dr.validate  
set.seed(1234)  
train <- sample(nrow(x), 0.7*nrow(x))  
x.train <- x[train,]  
x.validate <- x[-train,]  
table(x.train$class)  
table(x.validate$class)  


# 逻辑回归- Logistic regression with glm()  
fit <- glm(class~., data=x.train, family=binomial())  
summary(fit)  
prob <- predict(fit, x.validate, type="response")  
result <- factor(prob > .5, levels=c(FALSE, TRUE), labels=c("benign", "malignant"))  
logit.perf <- table(x.validate$class, result,dnn=c("Actual", "Predicted"))  
logit.perf

http://vassarstats.net/clin1.html

library(dplyr)
newdata <- dplyr::sample_n(x,5)
newdata <- newdata[-10]
prob <-predict(fit,newdata = newdata,type = 'response')
result <- factor(prob > .5, levels=c(FALSE, TRUE), labels=c("benign", "malignant"))  
result

sx.voiceclouds.cn