multiple_linear Regression with

1、先通过代码对数据进行预处理及模型喂养

datasets = read.csv('50_Startups.csv')

#数据明确

datasets$State = factor(datasets$State,

levels = c("New York", "California", "Florida"),

labels = c(1,2,3))

#数据分割

#用

library(caTools)

#set.seed(123)里面有数据就是随机处理

set.seed(123)

split = sample.split(datasets$Profit, SplitRatio = 0.8)

training_set = subset(datasets, split == TRUE) #0.8

test_set = subset(datasets, split == FALSE) #0.2

#数据规范化（缩放）

#training_set[,2:3 ] = scale(training_set[, 2:3])

#test_set[,2:3 ] = scale(test_set[,2:3 ])

#导入formula包 为数据喂养线性函数 Fitting Linear(formula:画线的方法，lm画线的模型)

#regression = lm(formula = Profit ~ R.D.Spend + Administration + Marketing.spengd + State , data = training_set)

regression = lm(formula = Profit ~., data = training_set) #简写

#预期结果，查看预测结果

y_pred = predict(regression, newdata = test_set)

#喂养多元性回归用降梯回归

regressor = lm(formula = Profit ~ R.D.Spend + Administration + Marketing.Spend + State , data = training_set)

summary(regressor) #all in

summary(regressor) #删除state

regressor = lm(formula = Profit ~ R.D.Spend  + Marketing.Spend  , data = training_set)

summary(regressor) #删除Administration

regressor = lm(formula = Profit ~ R.D.Spend    , data = training_set)

summary(regressor) #删除Marketing.Spend

注意：可以用两种方式，很多数据的时候推荐用简写

#regression = lm(formula = Profit ~ R.D.Spend + Administration + Marketing.spengd + State , data = training_set)

regression = lm(formula = Profit ~., data = training_set) #简写

 偏差最小的数据

All in

偏差最大的数据

偏大最大的两个是State2 State3

运行代码删除偏差最大的数据

删除State

进一步删除偏差最大的数据

删除偏Administar差大的数据

最后得出R.D.Spend研发部门对毛利的影响最大

R.D.Spend