Klasifikasi Dataset Palmer Penguin Menggunakan Support Vector Machine (SVM)
Muhamad Ramdhani
1. Dataset yang Digunakan
Dataset yang digunakan adalah Palmer Penguins, berisi informasi morfologi dan spesies penguin dari kepulauan Palmer.
library(palmerpenguins)## Warning: package 'palmerpenguins' was built under R version 4.4.3library(dplyr)##
## Attaching package: 'dplyr'## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, unionlibrary(ggplot2)
data("penguins")
head(penguins)2. Eksplorasi Data dan Visualisasi Awal
Eksplorasi dilakukan dengan visualisasi scatter plot antara panjang paruh dan panjang sirip untuk memisahkan spesies.
summary(penguins)## species island bill_length_mm bill_depth_mm
## Adelie :152 Biscoe :168 Min. :32.10 Min. :13.10
## Chinstrap: 68 Dream :124 1st Qu.:39.23 1st Qu.:15.60
## Gentoo :124 Torgersen: 52 Median :44.45 Median :17.30
## Mean :43.92 Mean :17.15
## 3rd Qu.:48.50 3rd Qu.:18.70
## Max. :59.60 Max. :21.50
## NA's :2 NA's :2
## flipper_length_mm body_mass_g sex year
## Min. :172.0 Min. :2700 female:165 Min. :2007
## 1st Qu.:190.0 1st Qu.:3550 male :168 1st Qu.:2007
## Median :197.0 Median :4050 NA's : 11 Median :2008
## Mean :200.9 Mean :4202 Mean :2008
## 3rd Qu.:213.0 3rd Qu.:4750 3rd Qu.:2009
## Max. :231.0 Max. :6300 Max. :2009
## NA's :2 NA's :2ggplot(penguins, aes(x = flipper_length_mm, y = bill_length_mm, color = species)) +
geom_point(size = 2) +
labs(title = "Distribusi Penguin berdasarkan Fitur Morfologi") +
theme_minimal()## Warning: Removed 2 rows containing missing values or values outside the scale range
## (`geom_point()`).
3. Model SVM Linear dan RBF dengan K-Fold CV
Model dievaluasi menggunakan K-Fold Cross Validation (k=5). Berikut implementasi dan ringkasan performa model.
library(e1071)## Warning: package 'e1071' was built under R version 4.4.3library(caret)## Warning: package 'caret' was built under R version 4.4.3## Loading required package: latticepenguins_clean <- penguins %>%
na.omit() %>%
select(species, bill_length_mm, flipper_length_mm)
train_control <- trainControl(method = "cv", number = 5)
svm_linear <- train(species ~ ., data = penguins_clean, method = "svmLinear", trControl = train_control)
svm_rbf <- train(species ~ ., data = penguins_clean, method = "svmRadial", trControl = train_control)
svm_linear## Support Vector Machines with Linear Kernel
##
## 333 samples
## 2 predictor
## 3 classes: 'Adelie', 'Chinstrap', 'Gentoo'
##
## No pre-processing
## Resampling: Cross-Validated (5 fold)
## Summary of sample sizes: 267, 267, 266, 266, 266
## Resampling results:
##
## Accuracy Kappa
## 0.951877 0.9241863
##
## Tuning parameter 'C' was held constant at a value of 1svm_rbf## Support Vector Machines with Radial Basis Function Kernel
##
## 333 samples
## 2 predictor
## 3 classes: 'Adelie', 'Chinstrap', 'Gentoo'
##
## No pre-processing
## Resampling: Cross-Validated (5 fold)
## Summary of sample sizes: 265, 266, 267, 268, 266
## Resampling results across tuning parameters:
##
## C Accuracy Kappa
## 0.25 0.9521416 0.9246351
## 0.50 0.9551719 0.9295420
## 1.00 0.9551719 0.9295158
##
## Tuning parameter 'sigma' was held constant at a value of 2.332133
## Accuracy was used to select the optimal model using the largest value.
## The final values used for the model were sigma = 2.332133 and C = 0.5.4. Visualisasi Decision Boundary (2D)
Visualisasi decision boundary dibuat untuk membandingkan model linear dan RBF.
model_linear <- svm(species ~ ., data = penguins_clean, kernel = "linear")
model_rbf <- svm(species ~ ., data = penguins_clean, kernel = "radial")
x_range <- seq(min(penguins_clean$bill_length_mm), max(penguins_clean$bill_length_mm), length = 200)
y_range <- seq(min(penguins_clean$flipper_length_mm), max(penguins_clean$flipper_length_mm), length = 200)
grid <- expand.grid(bill_length_mm = x_range, flipper_length_mm = y_range)
grid$pred_linear <- predict(model_linear, grid)
grid$pred_rbf <- predict(model_rbf, grid)
# SVM Linear
ggplot() +
geom_tile(data = grid, aes(x = bill_length_mm, y = flipper_length_mm, fill = pred_linear), alpha = 0.3) +
geom_point(data = penguins_clean, aes(x = bill_length_mm, y = flipper_length_mm, color = species), size = 2) +
labs(title = "Decision Boundary - SVM Linear") +
theme_minimal()
# SVM RBF
ggplot() +
geom_tile(data = grid, aes(x = bill_length_mm, y = flipper_length_mm, fill = pred_rbf), alpha = 0.3) +
geom_point(data = penguins_clean, aes(x = bill_length_mm, y = flipper_length_mm, color = species), size = 2) +
labs(title = "Decision Boundary - SVM RBF") +
theme_minimal()
5. Interpretasi Parameter C dan Gamma
Parameter C mengontrol margin dan regularisasi. Gamma mengontrol sensitivitas terhadap titik data.
set.seed(123)
tune.out <- tune(svm, species ~ ., data = penguins_clean,
kernel = "radial",
ranges = list(cost = c(0.1, 1, 10), gamma = c(0.01, 0.1, 1)))
summary(tune.out)##
## Parameter tuning of 'svm':
##
## - sampling method: 10-fold cross validation
##
## - best parameters:
## cost gamma
## 10 1
##
## - best performance: 0.04180036
##
## - Detailed performance results:
## cost gamma error dispersion
## 1 0.1 0.01 0.24367201 0.07043843
## 2 1.0 0.01 0.05989305 0.03432061
## 3 10.0 0.01 0.04483066 0.02052299
## 4 0.1 0.10 0.05989305 0.03432061
## 5 1.0 0.10 0.04483066 0.02052299
## 6 10.0 0.10 0.04777184 0.02813163
## 7 0.1 1.00 0.04786096 0.02037625
## 8 1.0 1.00 0.04786096 0.02037625
## 9 10.0 1.00 0.04180036 0.034971886. Kesimpulan dan Refleksi
SVM efektif dalam klasifikasi spesies penguin. Model RBF menunjukkan akurasi yang lebih baik dibanding linear. Tuning parameter sangat mempengaruhi akurasi dan generalisasi model.