Discriminant Analysis

#Discriminant Analysis example
library(rattle)
library(heplots)
data(wine)
attach(wine)
#check classes for each variable
lapply(wine, class)

$Type
[1] "factor"

$Alcohol
[1] "numeric"

$Malic
[1] "numeric"

$Ash
[1] "numeric"

$Alcalinity
[1] "numeric"

$Magnesium
[1] "integer"

$Phenols
[1] "numeric"

$Flavanoids
[1] "numeric"

$Nonflavanoids
[1] "numeric"

$Proanthocyanins
[1] "numeric"

$Color
[1] "numeric"

$Hue
[1] "numeric"

$Dilution
[1] "numeric"

$Proline
[1] "integer"

#check assumption for lda
boxM(wine[,2:14],Type)

    Box's M-test for Homogeneity of Covariance Matrices

data:  wine[, 2:14]
Chi-Sq (approx.) = 684.2, df = 182, p-value < 2.2e-16

#random partition
## 75% of the sample size
smp_size = floor(0.75 * nrow(wine))
## set the seed to make your partition reproducible
set.seed(1337)
train_ind = sample(seq_len(nrow(wine)), size = smp_size)
train = wine[train_ind, ]
test = wine[-train_ind, ]
#quadratic discriminant analysis
library(MASS)
wine.qda = qda(Type ~ ., data = train)
predicted.qda = predict(wine.qda, newdata = test)
table(test$Type, predicted.qda$class, dnn = c('Actual Group','Predicted Group'))

            Predicted Group
Actual Group  1  2  3
           1  9  0  0
           2  1 22  0
           3  0  0 13

# Model accuracy
mean(predicted.qda$class == test$Type)

[1] 0.9777778

# Partition plots
library(klaR)
partimat(wine[,2:5],Type,data=test,method="qda",main="Partition Plots")