สอนปลูกต้นไม้ในภาษา R (ภาค 2): วิธีสร้าง ประเมิน และปรับทูน random forest model ด้วย ranger package – ตัวอย่างการทำนายระดับการกินน้ำมันของรถใน mpg dataset

# Install ggplot2
install.packages("ggplot2")

# Load ggplot2
library(ggplot2)

# Load the dataset
data(mpg)

# Preview
head(mpg)

# A tibble: 6 × 11
  manufacturer model displ  year   cyl trans      drv     cty   hwy fl    class  
  <chr>        <chr> <dbl> <int> <int> <chr>      <chr> <int> <int> <chr> <chr>  
1 audi         a4      1.8  1999     4 auto(l5)   f        18    29 p     compact
2 audi         a4      1.8  1999     4 manual(m5) f        21    29 p     compact
3 audi         a4      2    2008     4 manual(m6) f        20    31 p     compact
4 audi         a4      2    2008     4 auto(av)   f        21    30 p     compact
5 audi         a4      2.8  1999     6 auto(l5)   f        16    26 p     compact
6 audi         a4      2.8  1999     6 manual(m5) f        18    26 p     compact

# View the structure
str(mpg)

tibble [234 × 11] (S3: tbl_df/tbl/data.frame)
 $ manufacturer: chr [1:234] "audi" "audi" "audi" "audi" ...
 $ model       : chr [1:234] "a4" "a4" "a4" "a4" ...
 $ displ       : num [1:234] 1.8 1.8 2 2 2.8 2.8 3.1 1.8 1.8 2 ...
 $ year        : int [1:234] 1999 1999 2008 2008 1999 1999 2008 1999 1999 2008 ...
 $ cyl         : int [1:234] 4 4 4 4 6 6 6 4 4 4 ...
 $ trans       : chr [1:234] "auto(l5)" "manual(m5)" "manual(m6)" "auto(av)" ...
 $ drv         : chr [1:234] "f" "f" "f" "f" ...
 $ cty         : int [1:234] 18 21 20 21 16 18 18 18 16 20 ...
 $ hwy         : int [1:234] 29 29 31 30 26 26 27 26 25 28 ...
 $ fl          : chr [1:234] "p" "p" "p" "p" ...
 $ class       : chr [1:234] "compact" "compact" "compact" "compact" ...

# Convert character columns to factor

## Get character columns
chr_cols <- c("manufacturer", "model",
              "trans", "drv",
              "fl", "class")

## For-loop through the character columns
for (col in chr_cols) {
  mpg[[col]] <- as.factor(mpg[[col]])
}

## Check the results
str(mpg)

tibble [234 × 11] (S3: tbl_df/tbl/data.frame)
 $ manufacturer: Factor w/ 15 levels "audi","chevrolet",..: 1 1 1 1 1 1 1 1 1 1 ...
 $ model       : Factor w/ 38 levels "4runner 4wd",..: 2 2 2 2 2 2 2 3 3 3 ...
 $ displ       : num [1:234] 1.8 1.8 2 2 2.8 2.8 3.1 1.8 1.8 2 ...
 $ year        : int [1:234] 1999 1999 2008 2008 1999 1999 2008 1999 1999 2008 ...
 $ cyl         : int [1:234] 4 4 4 4 6 6 6 4 4 4 ...
 $ trans       : Factor w/ 10 levels "auto(av)","auto(l3)",..: 4 9 10 1 4 9 1 9 4 10 ...
 $ drv         : Factor w/ 3 levels "4","f","r": 2 2 2 2 2 2 2 1 1 1 ...
 $ cty         : int [1:234] 18 21 20 21 16 18 18 18 16 20 ...
 $ hwy         : int [1:234] 29 29 31 30 26 26 27 26 25 28 ...
 $ fl          : Factor w/ 5 levels "c","d","e","p",..: 4 4 4 4 4 4 4 4 4 4 ...
 $ class       : Factor w/ 7 levels "2seater","compact",..: 2 2 2 2 2 2 2 2 2 2 ...

# Install
install.packages("ranger")

# Load
library(ranger)

# Split the data

## Set seed for reproducibility
set.seed(123)

## Get training rows
train_rows <- sample(nrow(mpg),
                     nrow(mpg) * 0.7)

## Create a training set
train <- mpg[train_rows, ]

## Create a test set
test <- mpg[-train_rows, ]

# Initial random forest model

## Set seed for reproducibility
set.seed(123)

## Train the model
rf_model <- ranger(hwy ~ .,
                   data = train)

# Print the model
rf_model

Ranger result

Call:
 ranger(hwy ~ ., data = train) 

Type:                             Regression 
Number of trees:                  500 
Sample size:                      163 
Number of independent variables:  10 
Mtry:                             3 
Target node size:                 5 
Variable importance mode:         none 
Splitrule:                        variance 
OOB prediction error (MSE):       1.682456 
R squared (OOB):                  0.9584596 

# Make predictions
preds <- predict(rf_model,
                 data = test)$predictions

# Get errors
errors <- test$hwy - preds

# Calculate MAE
mae <- mean(abs(errors))

# Calculate RMSE
rmse <- sqrt(mean(errors^2))

# Calculate R squared
r_sq <- 1 - (sum((errors)^2) / sum((test$hwy - mean(test$hwy))^2))

# Print the results
cat("Initial model MAE:", round(mae, 2), "\n")
cat("Initial model RMSE:", round(rmse, 2), "\n")
cat("Initial model R squared:", round(r_sq, 2), "\n")

Initial model MAE: 0.79
Initial model RMSE: 1.07
Initial model R squared: 0.95

# Define hyperparametres
ntree_vals <- c(300, 500, 700)
mtry_vals <- 2:5
min_node_vals <- c(1, 5, 10)

# Create a hyperparametre grid
grid <- expand.grid(num.trees = ntree_vals,
                    mtry = mtry_vals,
                    min.node.size = min_node_vals)

# Instantiate an empty data frame
hpt_results <- data.frame()

# For-loop through the hyperparametre grid
for (i in 1:nrow(grid)) {
  
  ## Get the combination
  params <- grid[i, ]
  
  ## Set seed for reproducibility
  set.seed(123)
  
  ## Fit the model
  model <- ranger(hwy ~ .,
                  data = train,
                  num.trees = params$num.trees,
                  mtry = params$mtry,
                  min.node.size = params$min.node.size)
  
  ## Make predictions
  preds <- predict(model,
                   data = test)$predictions
  
  ## Get errors
  errors <- test$hwy - preds
  
  ## Calculate MAE
  mae <- mean(abs(errors))
  
  ## Calculate RMSE
  rmse <- sqrt(mean(errors^2))
  
  ## Store the results
  hpt_results <- rbind(hpt_results,
                       cbind(params,
                             MAE = mae,
                             RMSE = rmse))
}

# View the results
hpt_results

   num.trees mtry min.node.size       MAE      RMSE
1        300    2             1 0.8101026 1.0971836
2        500    2             1 0.8012484 1.0973957
3        700    2             1 0.8039271 1.1001252
4        300    3             1 0.7434543 1.0051344
5        500    3             1 0.7417985 1.0069989
6        700    3             1 0.7421666 1.0028184
7        300    4             1 0.6989314 0.9074216
8        500    4             1 0.7130704 0.9314843
9        700    4             1 0.7141147 0.9292718
10       300    5             1 0.7157657 0.9370918
11       500    5             1 0.7131899 0.9266787
12       700    5             1 0.7091556 0.9238312
13       300    2             5 0.8570125 1.1673637
14       500    2             5 0.8515116 1.1736009
15       700    2             5 0.8522571 1.1756648
16       300    3             5 0.7885005 1.0654548
17       500    3             5 0.7872713 1.0664734
18       700    3             5 0.7859149 1.0581331
19       300    4             5 0.7561500 0.9790160
20       500    4             5 0.7623437 0.9869463
21       700    4             5 0.7611660 0.9813048
22       300    5             5 0.7615190 0.9777769
23       500    5             5 0.7615861 0.9804616
24       700    5             5 0.7613151 0.9788333
25       300    2            10 0.9257704 1.2391377
26       500    2            10 0.9292344 1.2611164
27       700    2            10 0.9258555 1.2635794
28       300    3            10 0.8790601 1.1635695
29       500    3            10 0.8704461 1.1594165
30       700    3            10 0.8704562 1.1507016
31       300    4            10 0.8609516 1.0887466
32       500    4            10 0.8672105 1.0962367
33       700    4            10 0.8624934 1.0875710
34       300    5            10 0.8558867 1.0811168
35       500    5            10 0.8567463 1.0783473
36       700    5            10 0.8536824 1.0751511

# Visualise the results
ggplot(hpt_results,
       aes(x = mtry,
           y = RMSE,
           color = factor(num.trees))) +
  
  ## Use scatter plot
  geom_point(aes(size = min.node.size)) +
  
  ## Set theme to minimal
  theme_minimal() +
  
  ## Add title, labels, and legends
  labs(title = "Hyperparametre Tuning Results",
       x = "mtry",
       y = "RMSE",
       color = "num.trees",
       size = "min.node.size")

# Define the best hyperparametres
best_num.tree <- 300
best_mtry <- 4
best_min.node.size <- 2.5

# Fit the model
rf_model_new <- ranger(hwy ~ .,
                       data = train,
                       num.tree = best_num.tree,
                       mtry = best_mtry,
                       min.node.size = best_min.node.size)

# Evaluate the model

## Make predictions
preds_new <- predict(rf_model_new,
                     data = test)$predictions

## Get errors
errors_new <- test$hwy - preds_new

## Calculate MAE
mae_new <- mean(abs(errors_new))

## Calculate RMSE
rmse_new <- sqrt(mean(errors_new^2))

## Calculate R squared
r_sq_new <- 1 - (sum((errors_new)^2) / sum((test$hwy - mean(test$hwy))^2))

## Print the results
cat("Final model MAE:", round(mae_new, 2), "\n")
cat("Final model RMSE:", round(rmse_new, 2), "\n")
cat("Final model R squared:", round(r_sq_new, 2), "\n")

Final model MAE: 0.71
Final model RMSE: 0.93
Final model R squared: 0.96

# Compare the two models
model_comp <- data.frame(Model = c("Initial", "Final"),
                         MAE = c(round(mae, 2), round(mae_new, 2)),
                         RMSE = c(round(rmse, 2), round(rmse_new, 2)),
                         R_Squared = c(round(r_sq, 2), round(r_sq_new, 2)))

# Print
model_comp

    Model  MAE RMSE R_Squared
1 Initial 0.85 1.08      0.95
2   Final 0.71 0.93      0.96

# Fit the model with importance
rf_model_new <- ranger(hwy ~ .,
                       data = train,
                       num.tree = best_num.tree,
                       mtry = best_mtry,
                       min.node.size = best_min.node.size,
                       importance = "permutation") # Add importance

# Install vip package
install.packages("vip")

# Load vip package
library(vip)

# Get variabe importance
vip(rf_model_new)  +
  
  ## Add title and labels
  labs(title = "Variable Importance - Final Random Forest Model",
       x = "Variables",
       y = "Importance") +
  
  ## Set theme to minimal
  theme_minimal()