# Predmet: "Analyza zhlukov a klasifikacia dat"
# Studijny program: "Pravdepodobnost a matematicka statistika"
# Vyucujuci: Radoslav Harman, KAMS FMFI UK Bratislava

# Analyza zhlukov 3: hierarchicke zhlukovanie

# Data o extrasolarnych planetach (podrobnosti k datam boli minule)
planets <- read.csv("http://www.iam.fmph.uniba.sk/ospm/Harman/data/planets.csv")

planets <- planets[planets$pl_orbper < 600, ]; n <- nrow(planets)
for (i in c(4,5,8)) planets[, i] <- log(planets[, i])
planets[, 7] <- (planets[, 7])^(1/3)
planets[, 4:8] <- scale(planets[, 4:8])

# Specializovane procedury na HK nam ponuka kniznica cluster 
library(cluster)

# Aglomerativne zhlukovanie
help(agnes)

# Data pred aplikaciou HK standardizujeme a pouzijeme metodu "average" na meranie vzdialenosti zhlukov
plan.agnes1 <- agnes(planets[, 4:8], stand = TRUE, method = "average")
summary(plan.agnes1)

# Uplnu reprezentaciu dendrogramu poskytne zlozka merge
# -cislo znamena, ze zlucujeme objekt "cislo"
# (+)cislo znamena, ze spajame uz cely zhluk objektov, vytvoreny v riadku "cislo" matice 
plan.agnes1$merge

# Dendrogram si mozeme nakreslit
plot(plan.agnes1, which.plots = 2, hang = -1)
help(plot.agnes1)

# Porovnajme vysledok pouzitim inych metod merania vzdialenosti medzi zhlukmi
plan.agnes2 <- agnes(planets[, 4:8], stand = TRUE, method = "single")
plot(plan.agnes2, which.plots = 2, hang = -1)
plan.agnes3 <- agnes(planets[, 4:8], stand = TRUE, method = "complete")
plot(plan.agnes3, which.plots = 2, hang = -1)

# Divizivne zhlukovanie
help(diana)

# Data pred aplikaciou HK standardizujeme. Divizivne zhlukovanie nemeria vzdialenosti zhlukov
plan.diana <- diana(planets[, 4:8], stand = TRUE)
summary(plan.diana)
plot(plan.diana, which.plots = 2, hang = -1)