[R] R 정리 18 - svm, knn

# SVM 분류 모델 : 데이터 분류 결정 마진에 영향을 주는 관측치(Support Vectors)를 이용
# Kernel trick을 이용해 비선형 데이터에 대한 분류도 가능

library(e1071)

# train/test 생략
head(iris, 3)
# Sepal.Length Sepal.Width Petal.Length Petal.Width Species
# 1          5.1         3.5          1.4         0.2  setosa
# 2          4.9         3.0          1.4         0.2  setosa
# 3          4.7         3.2          1.3         0.2  setosa
x <- subset(iris, select = -Species)
y <- iris$Species
x
y

svm_model <- svm(Species ~ ., data=iris)
svm_model
# Parameters:
#   SVM-Type:  C-classification 
# SVM-Kernel:  radial 
# cost:  1 
# 
# Number of Support Vectors:  51
summary(svm_model)

svm_model1 <- svm(x, y)
summary(svm_model1)

dist(iris[, -5])
cmdscale(dist(iris[, -5])) # 거리 값을 공간적인 배치형식으로 출력
# [,1]         [,2]
# [1,] -2.684125626  0.319397247
# [2,] -2.714141687 -0.177001225
# [3,] -2.888990569 -0.144949426
# [4,] -2.745342856 -0.318298979
# [5,] -2.728716537  0.326754513
# [6,] -2.280859633  0.741330449
# [7,] -2.820537751 -0.089461385
# [8,] -2.626144973  0.163384960

# support vector : + , 나머지 데이터는 o
plot(cmdscale(dist(iris[, -5])), col=as.integer(iris[,5]), pch=c("o","+")[1:150 %in% svm_model$index + 1])

 

pred <- predict(svm_model, x)
head(pred)
t <- table(pred, y)
t
# y
# pred         setosa versicolor virginica
# setosa         50          0         0
# versicolor      0         48         2
# virginica       0          2        48
sum(diag(t)) / nrow(x) # 분류 정확도 : 0.9733333

# 최상의 모델을 만들기 위해 파라미터 조절 - svm 튜닝

svm_tune <- tune(svm, train.x = x, train.y = y, kernel='radial',
                 ranges = list(cost=10^(-1:2)), gamma=2^(-2:2))
                # 10-fold cross validation : 교차 검정
svm_tune
# - sampling method: 10-fold cross validation 
# 
# - best parameters:
#   cost
# 1
# 
# - best performance: 0.03333333 

svm_model_after_tune <- svm(Species ~ ., data=iris, kernel='radial', cost=1, gamma=0.5)
summary(svm_model_after_tune)

pred <- predict(svm_model_after_tune, x)
t <- table(pred, y)
t
# y
# pred         setosa versicolor virginica
# setosa         50          0         0
# versicolor      0         48         2
# virginica       0          2        48
sum(diag(t)) / nrow(x) # 0.9733333

---

# KNN : 최근접 이웃 알고리즘을 사용. k개의 원 안에 데이터 수 
# 또는 거리(유클디안 거리 계산법)에 의해 새로운 데이터를 분류.
# 비모수적 검정 분류 모형

install.packages("ggvis")
library(ggvis)

head(iris, 2)
iris %>% ggvis(~Petal.Length, ~Petal.Width, fill=~factor(Species))

---