R 회귀분석

###회귀분석 - 추정(결과변수가 연속변수일 경우)에 활용하는 대표적인 분석방법 ##단순회귀분석과 다중회귀분석,로지스틱회지분석  ##전처리 auto=read.csv('autoparts.csv') dim(auto) is.na(auto) complete.cases(auto) auto[!complete.cases(auto),] head(auto) auto1=auto[auto$prod_no=='90784-76001',c(2:11)] head(auto1) summary(auto1) boxplot(auto1) boxplot(auto1$separation) auto2=auto1[auto$c_thickness<1000,] boxplot(auto2) hist(auto2$c_thickness,breaks = 30) ##교차검증 ###모델링_단순선형회귀모형 head(women) summary(women) #이상치확인 m=lm(weight~height,data=women) ;m summary(m) plot(women$height,women$weight) #산점도 abline(m,col='green') ##예측 predict new.data=data.frame(height=75) predict(m,newdata=new.data) new.data=data.frame(height=c(75,76)) predict(m,newdata=new.data) ##신뢰구간 예측 interval='confidence' / 95% predict(m,newdata=new.data,interval = 'confidence') head(auto2) m=lm(c_thickness~fix_time,data=auto2) new.data=data.frame(fix_time=86.1) predict(m,newdata=new.data) ###다중선형회귀분석 m=lm(c_thickness~fix_time+a_speed,data=auto2) n=lm(c_thickness~.,data=auto2) summary(m) summary(n) ##복수 데이터 예측 new.data=data.frame(fix_time=86.1,a_speed=0.610,b_speed=1.718,separation=241.9,s_separation=657.3,                             rate_term=95,mpa=78.2,load_time=18.1,highpressure_time=74) predict(m,newdata=new.data) new.data=data.frame(fix_time=c(86.1,86.1),a_speed=c(0.610,0.603),b_speed=c(1.718,1.704),separation=c(241.9,242.5),                             rate_term=c(95,95),mpa=c(78.2,77.9),load_time=c(18.1,18.2),highpressure_time=c(74,56)) predict(m,newdata=new.data,interval = 'confidence') test=auto2[32:50,] new.data=data.frame(fix_time=test$fix_time,a_speed=test$a_speed,b_speed=test$b_speed,separation=test$separation,                           s_separation=test$s_separation,rate_terms=test$rate_terms,mpa=test$mpa,load_time=test$load_time,                           highpressure_time=test$highpressure_time) predict(m,newdata=new.data) ###최적변수찾기 step head(swiss) m=lm(Fertility~.,data=swiss) summary(m) step(m,direction = 'forward') #전진선택법 step(m,direction = 'backward') #후진제거법 step(m,direction = 'both') #단계별선택법 m=lm(c_thickness~.,data=auto2) step(m,direction = 'forward') step(m,direction = 'backward') step(m,direction = 'both') ##교차분석_sample, split #교차분석(K-fold CV)을 위한 데이터셋 나누기 t_index=sample(1:nrow(auto2),size=nrow(auto2)) #행의 갯수만큼 변수를 만든다 split_index=split(t_index,1:10) #10개그룹으로 나눈다 class(split_index) length(split_index) split_index$'1' split_index[[1]]#1의 밸류를 다 가져오기 auto=read.csv('autoparts.csv') auto1=auto[auto$prod_no=='90784-76001',c(2:11)] auto2=auto1[auto1$c_thickness<1000,] mse=c() for(i in 1:10){   test=auto2[split_index[[i]],]   train=auto2[-split_index[[i]],]   m=lm(c_thickness~.,data=train)   m_pred=predict(m,test)   mse[i]=mean((test$c_thickness-m_pred)^2) }   mse mean(mse) #mse평균제곱오차 split_index=split(t_index,1:5) mse=c() for(i in 1:5){   test=auto2[split_index[[i]],]   train=auto2[-split_index[[i]],]   m=lm(c_thickness~.,data=train)   m_pred=predict(m,test)   mse[i]=mean((test$c_thickness-m_pred)^2) }   mse mean(mse) ##간단한 성능평가 방법 t_index=sample(1:nrow(auto2),size=nrow(auto2)*0.7) train=auto2[t_index,] test=auto2[-t_index,] nrow(train);nrow(test) m=lm(c_thickness~.,data=train) m_pred=predict(m,test) mean((test$c_thickness-m_pred)^2) ##LASSO 자동 변수 선택 기법 - 독립변수 선택할 때  xmat=as.matrix(auto2[1:9]) #종속변수 제외하고는 매트릭스 형태로 바꿔줌/ 데이터프레임[]는 열을 의미 yvec=auto2$c_thickness library(glmnet) install.packages('glmnet') fit.lasso=glmnet(x=xmat,y=yvec,alpha=1,nlambda = 100) #알파 : 비용(1이 제일 큼) fit.lasso.cv=cv.glmnet(x=xmat,y=yvec,nfolds=10,alpha=1,lambda = fit.lasso$lambda) #nfold수행횟수 plot(fit.lasso.cv) fit.lasso.param=fit.lasso.cv$lambda.min #람다 가장 작은 값 fit.lasso.param=fit.lasso.cv$lambda.1se #처음 커지는 람다 값 fit.lasso.tune=glmnet(x=xmat,y=yvec,alpha = 1,lambda=fit.lasso.param) coef(fit.lasso.tune) #계수 #연습문제 train=read.csv('occupancy_train.csv') head(train) xmat=as.matrix(train[2:6]) occ=train$Occupancy fit.lasso=glmnet(x=xmat,y=occ,alpha = 1,nlambda = 50) fit.lasso.cv=cv.glmnet(x=xmat,y=occ,nfolds=10,alpha=1,lambda=fit.lasso$lambda) plot(fit.lasso.cv) fit.lasso.param=fit.lasso.cv$lambda.min fit.lasso.tune=glmnet(x=xmat,y=occ,alpha = 1,lambda = fit.lasso.param) coef(fit.lasso.tune) fit.lasso.param1=fit.lasso.cv$lambda.1se fit.lasso.tune1=glmnet(x=xmat,y=occ,alpha = 1,lambda = fit.lasso.param1) coef(fit.lasso.tune1)