r语言实现岭回归_78-预测分析－Ｒ语言实现-岭回归与LASSO回归

> library(pacman)

> p_load(dplyr, readr, caret)

以上一节中未去除离群值的MSE为3619.029，修正R2为0.8603和去除离群值后的MSE为2690.545,修正R2为0.8706为基准，以及两个模型在测试集上的MSE分别为2914.014和1672.859，对模型进行改进。

> results

+ "original", 3619.029, 0.8603, 2914.014,

+ "remove_out", 2690.545, 0.8706, 1672.859)

> results

## # A tibble: 2 x 4

## model mse r_square test_mse

##

## 1 original 3619. 0.860 2914.

## 2 remove_out 2691. 0.871 1673.

1、数据预处理

> machine

> names(machine)

+ "cach", "chmin", "chmax", "prp", "erp")

> machine

>

> set.seed(123)

> ind

>

> dtrain

> dtest

2、缩减特征集

> ct

> set.seed(123)

> fit.step

+ trControl = ct, preProcess = c("corr"), trace = F)

>

> summary(fit.step$finalModel)

##

## Call:

## lm(formula = .outcome ~ myct + mmin + mmax + cach + chmax, data = dat)

##

## Residuals:

## Min 1Q Median 3Q Max

## -163.94 -29.68 3.25 28.52 355.05

##

## Coefficients:

## Estimate Std. Error t value Pr(>|t|)

## (Intercept) -6.024e+01 8.909e+00