求教R语言做预测的方法

复制代码

现有代码是：
y<-rnorm(n = 100,mean = 2,sd = 1)
y<-sort(y)
x1<-rnorm(n = 100,mean = 2,sd = 1)
x2<-rnorm(n = 100,mean = 2,sd = 1)
x3<-rnorm(n = 100,mean = 2,sd = 1)

date1<-seq.Date(from = as.Date("1992-12-31"),by = "q",length.out = 100)

da01<-data.frame(date1,x1,x2,x3)
da01

lm1<-lm(y~x1+x2+x3,data = da01)

summary(lm1)

plot(date1,y,type = "l")



#思路是：(1) y对x1的3阶滞后项、x2的4阶滞后项、x3的3阶滞后项做回归，得出回归方程lm2
# (2) 根据x1 x2 x3 的现有数据，把y向前预测3期
# (3) 在同一张图上画出y的时间序列图，实际数值用黑实线，预测数值用黑虚线，95%置信区间用红色。

#（4）求高手把R代码补充齐全，谢谢！

1. # 数据准备
   
2. y <- rnorm(n = 100, mean = 2, sd = 1)
   
3. y <- sort(y)
   
4. x1 <- rnorm(n = 100, mean = 2, sd = 1)
   
5. x2 <- rnorm(n = 100, mean = 2, sd = 1)
   
6. x3 <- rnorm(n = 100, mean = 2, sd = 1)
   
7. 
   
8. date1 <- seq.Date(from = as.Date("1992-12-31"), by = "q", length.out = 103)
   
9. 
   
10. da02 <- data.frame(date1, y = c(y, rep(NA, 3)), x1 = c(rep(NA, 3), x1), x2 = c(rep(NA, 4), x2[1:99]), x3 = c(rep(NA, 3), x3))
    
11. train <- da02[1:100, ]
    
12. newdata <- da02[101:103, ]
    
13. 
    
14. # 建模和预测
    
15. lm2 <- lm(y ~ x1 + x2 + x3, data = train)
    
16. prediction <- predict(lm2, newdata = newdata, interval = "prediction")
    
17. predicted <- data.frame(date = date1[100:103], y = c(y[100], prediction[, "fit"]),
    
18.                         lwr =  c(y[100], prediction[, "lwr"]),
    
19.                         upr =  c(y[100], prediction[, "upr"]))
    
20. 
    
21. # 画图
    
22. plot(x = date1, y = c(y, rep(NA, 3)), type = "l", xlab = "date", ylab = "y")
    
23. lines(x = predicted$date, y = predicted$y, lty = 2)
    
24. lines(x = predicted$date, y = predicted$lwr, lty = 3, col = "red")
    
25. lines(x = predicted$date, y = predicted$upr, lty = 3, col = "red")
    
26. abline(v = predicted$date[1], col = "blue")

复制代码

cheetahfly 发表于 2018-1-15 10:37

方法可行，亲测成功，谢谢您的帮助。