par(mar=c(4.5,4.5,1,1))
plot(faithful$eruptions,faithful$waiting,cex=1,pch=19,xlab="Durata",ylab="Attesa",cex.lab=2)
par(mfrow=c(1,2),mar=c(4.5,4.5,1,1))
hist(faithful$eruptions,xlab="Durata",ylab="",main="",cex.lab=2,col="gray",breaks = 40)
hist(faithful$waiting,xlab="Attesa",ylab="",main="",cex.lab=2,col="gray",breaks = 40)

clus <- kmeans(faithful,2)
table(clus$cluster)
par(mfrow=c(1,1),mar=c(4.5,4.5,1,1))
plot(faithful$eruptions,faithful$waiting,cex=1,pch=19,xlab="Durata",ylab="Attesa",cex.lab=2,col=clus$cluster+1)
clus <- kmeans(scale(faithful),2)
table(clus$cluster)
par(mfrow=c(1,1),mar=c(4.5,4.5,1,1))
plot(faithful$eruptions,faithful$waiting,cex=1,pch=19,xlab="Durata",ylab="Attesa",cex.lab=2,col=clus$cluster+1)

library(cluster) 
par(mar=c(5,5,1,1))
clusplot(faithful, clus$cluster, color=TRUE, shade=TRUE, labels=2, main="",lines=0)
clusplot(pam(faithful,2), color=TRUE, shade=TRUE, labels=2, main="",lines=0)

load("/Users/utente/Documents/prova r pkg/lucidi del corso/datibi16.RData")
datibi16tr <- (datibi16^(1/5)-1)*5
pairs(datibi16tr[,c(2,6,11)],cex=0.5,pch=19)
clbi16 <- clara(datibi16tr[,c(2,6,11)],k=4,samples = 50, sampsize = 500)
table(clbi16$clustering)
pairs(datibi16tr[,c(2,6,11)],cex=0.5,pch=19,col=c("red","yellow","orange","gray","blue")[clbi16$clustering])
tt <- table(clbi16$clustering,datibi16$NASCAREA)
tt / rowSums(tt)

load("~/Documents/prova r pkg/lucidi del corso/esoplan.RData")
massa <- (dati$MASS^(1/30)-1)*30
perio <- (dati$PERIOD^(1/30)-1)*30
x <- scale(cbind(massa,perio),center=T,scale=T)
x <- x[complete.cases(x),]
par(mar=c(4.5,4.5,1,1))
plot(x,cex=0.5,pch=19,xlab="Massa",ylab="Periodo",cex.lab=2)
pampl <- pam(x, 3)
plot(x,cex=0.5,pch=19,xlab="Massa",ylab="Periodo",cex.lab=2,col=c("red","yellow","orange")[pampl$clustering])

pairs(USArrests,cex=0.5,pch=19)
prusa <- prcomp(USArrests,4,scale=T)
Distanza <- dist(prusa$x, method = "euclidean")
fit <- hclust(Distanza, method="ward.D") 
par(mar=c(5,5,1,1))
plot(fit,main="",cex.lab=1.5)
groups <- cutree(fit, k=4)
rect.hclust(fit, k=4, border="red")
library(usmap)
library(ggplot2)
newdat <- as.data.frame(statepop)
newdat$clust <- rep(0,51)
newdat$clust[1:8] <- groups[1:8]
newdat$clust[10:51] <- groups[9:50]
plot_usmap(data = newdat, values = "clust", lines = "red") + 
  scale_fill_continuous(low = "yellow", high = "red", name = "Gruppi Crimine", label = scales::comma) + 
  theme(legend.position = "right")

library(factoextra)
UScale <- scale(USArrests)
par(mar=c(4.5,4.5,1,1))
fviz_nbclust(UScale, kmeans, method = "wss")+geom_vline(xintercept = 4, linetype = 2)+labs(subtitle = "Elbow method")
fviz_nbclust(UScale, kmeans, method = "silhouette")+labs(subtitle = "Silhouette method")
set.seed(12345)
fviz_nbclust(UScale, kmeans, nstart = 25,  method = "gap_stat", nboot = 50)+labs(subtitle = "Gap statistic method")
library(NbClust)
nbus <- NbClust(data = UScale, distance = "euclidean", min.nc = 2, max.nc = 15, method = "kmeans")

load("/Users/utente/Documents/prova r pkg/lucidi del corso/fertvsasvita.RData")
plot(wdd$`value.Fertility rate, total (births per woman)`,wdd$`value.Life expectancy at birth, total (years)`,cex=1,pch=19,col=wdd$region)
library(mclust)
risgap <- Mclust(na.omit(wdd[,3:4]),G=1:9)
plot(risgap)

load("/Users/utente/Documents/prova r pkg/lucidi del corso/RemSens.RData")
library(raster)
plotRGB(ls8)
nr <- getValues(ls8)
kmls8 <- kmeans(na.omit(nr), centers = 10, iter.max = 500, nstart = 5, algorithm="Lloyd")
kmls8
table(kmls8$cluster)
sclu <- ls8[[1]]
knr <- setValues(sclu, kmls8$cluster)
plot(knr, main = 'Unsupervised classification', col = terrain.colors(10))
plotRGB(sent3)
nr <- getValues(sent3)
kmsent3 <- kmeans(na.omit(nr), centers = 4, iter.max = 500, nstart = 5, algorithm="Lloyd")
kmsent3
table(kmsent3$cluster)
sclu <- sent3[[1]]
knr <- setValues(sclu, kmsent3$cluster)
plot(knr, main = 'Unsupervised classification', col = c("darkgreen","gray","cyan","gold"))

load("/Users/utente/Documents/prova r pkg/lucidi del corso/dna.RData")
set.seed(12345)
fviz_nbclust(dna[,2:13], kmeans, nstart = 25,  method = "gap_stat", nboot = 50)+labs(subtitle = "Gap statistic method")
library(NbClust)
nbus <- NbClust(data = dna[,2:13], distance = "euclidean", min.nc = 2, max.nc = 15, method = "kmeans")
cldna <- kmeans(dna[,2:13],centers = 3)
kk <- prcomp(dna[,2:13],2)$x[,1:2]
dna$Region <- as.character(dna$Region)
par(mfrow=c(1,1),mar=c(4.5,4.5,1,1))
plot(kk,cex=1,pch=19,col=c("red","blue","green")[cldna$cluster])
text(kk,dna$Region,pos=1,col=c("red","blue","green")[cldna$cluster],cex=1.2)

load("/Users/utente/Documents/prova r pkg/lucidi del corso/fifa19.RData")
cargioc <- fifa19[rowSums(is.na(fifa19[,55:88])) == 0,]
fifa19cl <- kmeans(cargioc[,55:88],centers = 10)
clusplot(cargioc[,55:88], fifa19cl$cluster)
fifa19pr <- prcomp(cargioc[,55:88])$x[,1:2]
par(mar=c(4.5,4.5,1,1))
plot(fifa19pr,cex=0.5,pch=19,col=fifa19cl$cluster)
table(cargioc$Position,fifa19cl$cluster)
aggregate(cargioc[,90:92],by=list(clus=fifa19cl$cluster),mean,na.rm=T)

load("/Users/utente/Documents/prova r pkg/lucidi del corso/musica.RData")
pairs(evol[,28:41],cex=0.5,pch=19,col=evol$cluster)
kk <- table(evol$year,evol$cluster)
vv <- kk/rowSums(kk)
tt <- t(vv)
barplot(tt,xlab="Year",ylab="Perc. cluster",main="Evoluzione dei generi")
kk <- table(evol$artist_name,evol$cluster)
prinart <- matrix("",ncol(kk),10)
freqart <- matrix(0,ncol(kk),10)
for (i in 1:ncol(kk))
{
  prinart[i,] <- names(sort(kk[,i],decreasing = T)[1:10])
  freqart[i,] <- sort(kk[,i],decreasing = T)[1:10]
}