#####################################################################

#                Importare i dati

#####################################################################


originale<-read.table("Dataset_whiskey.txt", header=TRUE,sep="\t")

str(originale)
View(originale)


# Esplorazione preliminare dei dati

plot(originale)

boxplot(originale[,-c(1:2)]) # solo colonne con valori numerici


# Boxplot per categorie (esempio)

boxplot(originale$B~originale$type)


#####################################################################

#                CORRELAZIONE TRA VARIABILI

#####################################################################


# Installare pacchetto corrplot

install.packages("corrplot")

# Caricare il pacchetto

library(corrplot)


# Generare la matrice di correlazione:

Cr<-cor(originale[,-c(1:2)]) # solo colonne con valori numerici


# Visualizzare la matrice di correlazione

windows() #nuova finestra per il grafico
corrplot(Cr)



#####################################################################

#                CLUSTERING GERARCHICO

#####################################################################

dataset<-as.matrix(originale[,-c(1:2)]) # Seleziono solo la parte di dataset con valori numerici

Type<-originale$type
Sample<-originale$sample


#-- Normalizzazione per variabile (E' NECESSARIA!!!)

dataNorm<-scale(dataset, center = TRUE, scale = TRUE)

#-- Clustering gerarchico

Dist<-dist(dataNorm, method="euclidean")

HcC<- hclust(Dist, method="complete")
HcS<- hclust(Dist, method="single")
HcA<- hclust(Dist, method="average")

DendC<-as.dendrogram(HcC)
DendS<-as.dendrogram(HcS)
DendA<-as.dendrogram(HcA)

plot(DendC,main="Complete")
windows()
plot(DendS,main="Single")
windows()
plot(DendA,main="Average")

#--- Selezione numero cluster

Cut<-cutree(HcC, k= 2)

Newdata<-data.frame(originale, K2= Cut)
View(Newdata)

#--- Esplorazione risultati

Box<-boxplot(Newdata$A~Newdata$K2,main="Variabile A")

Cluster1<-Newdata[which(Newdata$K2==1),]


###################################################################

#                       PCA

###################################################################

dataset<-as.matrix(originale[,-c(1:2)])

Type<-originale$type
Sample<-originale$sample


PCA<-prcomp(dataset, center= TRUE, scale=TRUE)


varianze<-PCA$sdev^2

varianzecum<-cumsum(varianze/sum(varianze)*100) 

#-- Scree plot

plot(varianze,pch=16,type="o")

abline(h=1,col="gray")

#-- Varianza cumulata %

plot(varianzecum,pch=16,type="o")
abline(h=c(50,75),col="gray")


#-- Plot loadings

plot(PCA$rotation[,1], PCA$rotation[,2],pch=16)
text(PCA$rotation[,1], PCA$rotation[,2],labels=colnames(dataset),cex=0.8,pos=3)
abline(h=0,col="gray")
abline(v=0,col="gray")

#-- Plot scores

plot(PCA$x[,1], PCA$x[,2],pch=16)
text(PCA$x[,1], PCA$x[,2],labels=Sample,cex=0.8,pos=3)
abline(h=0,col="gray")
abline(v=0,col="gray")


plot(PCA$x[,1], PCA$x[,2],type="n")
text(PCA$x[,1], PCA$x[,2],labels=Type,cex=0.8)
abline(h=0,col="gray")
abline(v=0,col="gray")

#-- a colori

Color<-as.factor(Type)
levels(Color)<-c("red","blue")
Color<-as.character(Color)

plot(PCA$x[,1], PCA$x[,2],type="n")
points(PCA$x[,1], PCA$x[,2],col=Color,cex=1.2,pch=16)
abline(h=0,col="gray")
abline(v=0,col="gray")


#-- Biplot

biplot(PCA, choices =c(1,2))


#-- Proiezione di dati

newdata<-dataset[c(1,20),]

New<-predict(PCA,newdata)


plot(PCA$x[,1], PCA$x[,2],type="n")
points(PCA$x[,1], PCA$x[,2],col=Color,cex=1.2,pch=16)
abline(h=0,col="gray")
abline(v=0,col="gray")

points(New[,1],New[,2],col="green",pch=18,cex=2)