R- 没有分类变量的多元线性回归的机器学习

如何解决R- 没有分类变量的多元线性回归的机器学习

您好，我想为我的数据集开发一个多元线性回归方程。我有 2 个独立的连续变量和 24 个数据点（具有因变量结果）。我实际上已经使用 lm( y~ x1 + x2,data=df) 函数创建了一个回归方程，但我想使用机器学习并将数据拆分为训练集和测试集。然后我想使用我获得的最佳模型来预测 PM 浓度。当我查看网站时，我总是会找到包含分类变量的示例，但我没有看到任何使用连续变量的示例。

我怎样才能找到一个很好的例子，我应该遵循什么路径？

谢谢

编辑：这是我尝试过的事情之一。

> library(caret)
> dim(df)
[1] 24  3
> head(df)
# A tibble: 6 x 3
     PM   SO2 Wind_speed
  <dbl> <dbl>      <dbl>
1 108.   7.57        0.9
2 115.   9.45        0.9
3  74.5 13.4         1.2
4  77.2  7.73        1.3
5  57.0  5.08        1.3
6  52.6  8.59        1.2
> summary(df)
       PM              SO2           Wind_speed   
 Min.   : 52.64   Min.   : 4.090   Min.   :0.600  
 1st Qu.: 76.84   1st Qu.: 7.397   1st Qu.:0.800  
 Median :105.69   Median : 9.265   Median :1.000  
 Mean   :118.62   Mean   :17.089   Mean   :1.004  
 3rd Qu.:158.02   3rd Qu.:15.070   3rd Qu.:1.200  
 Max.   :261.84   Max.   :75.270   Max.   :1.500  
> control <- trainControl(method="cv",number=10)
> summary(control)
                  Length Class 
method            1      -none-
number            1      -none-
repeats           1      -none-
search            1      -none-
p                 1      -none-
initialWindow     0      -none-
horizon           1      -none-
fixedWindow       1      -none-
skip              1      -none-
verboseIter       1      -none-
returnData        1      -none-
returnResamp      1      -none-
savePredictions   1      -none-
classprobs        1      -none-
summaryFunction   1      -none-
selectionFunction 1      -none-
preProcoptions    6      -none-
sampling          0      -none-
index             0      -none-
indexOut          0      -none-
indexFinal        0      -none-
timingSamps       1      -none-
predictionBounds  2      -none-
seeds             1      -none-
adaptive          4      -none-
trim              1      -none-
allowParallel     1      -none-
                  Mode     
method            character
number            numeric  
repeats           logical  
search            character
p                 numeric  
initialWindow     NULL     
horizon           numeric  
fixedWindow       logical  
skip              numeric  
verboseIter       logical  
returnData        logical  
returnResamp      character
savePredictions   logical  
classprobs        logical  
summaryFunction   function 
selectionFunction character
preProcoptions    list     
sampling          NULL     
index             NULL     
indexOut          NULL     
indexFinal        NULL     
timingSamps       numeric  
predictionBounds  logical  
seeds             logical  
adaptive          list     
trim              logical  
allowParallel     logical  
> metric <- "Accuracy"
> set.seed(7)
> fit.lda <- train(PM~.,data=df,method="lda",metric=metric,trControl=control)
Hata: wrong model type for regression
>set.seed(7)
> fit.cart <- train(PM~.,data= df,method="rpart",trControl=control)
Hata: Metric Accuracy not applicable for regression models