#Cluster Analysis assessment in proposing a surgical technique for benign prostatic enlargement_ICIMTH 2022

install.packages("ggpubr")
install.packages("factoextra")
library(ggpubr)
library(factoextra)
citation()
citation("ggpubr")
citation("factoextra")
citation("dplyr")

#import csv from location
data <- read.csv(file.choose())
#preview file
summary(data)
# identify working directory
getwd()
# Rename the first column
colnames(data)[1]<-'SurgMethod'
summary(data)
#convert to factor
data$SurgMethod<- factor(data$SurgMethod)

data1<-data[ ,c('SurgMethod','Age','PSA','ProstateVolume','Hbbefore','Nabefore', 'PVRbefore', 'Qmaxbefore', 'IPSSbefore')]
summary(data1)
data.noclass = subset(data1, select = -c(1))
summary(data.noclass)

#Scale the variables
library(dplyr)

data.noclass_Scale<-data.noclass %>%
  mutate_if(is.numeric, scale)
summary(data.noclass_Scale)

#Clustering with K-means in 3 clusters
set.seed(100) 
data.noclass_Scale.cl <- kmeans(data.noclass_Scale, centers=3)


#Create a column with the clustering results in our original dataframe
data$cluster   <- fitted(data.noclass_Scale.cl, method = "classes") 

#Comparison results
table(data$SurgMethod, data$cluster)


#visualize k-means clusters
fviz_cluster(data.noclass_Scale.cl, data = data.noclass_Scale,
             palette = c("#2E9FDF", "#00AFBB", "#E7B800"), 
             geom = "point",
             ellipse.type = "convex", 
             ggtheme = theme_bw()
)

#Compute principal component analysis (PCA) to reduce the data into small dimensions for visualization

res.pca <- prcomp(data.noclass_Scale,  scale = TRUE)
# Coordinates of individuals
ind.coord <- as.data.frame(get_pca_ind(res.pca)$coord)
# Add clusters obtained using the K-means algorithm
ind.coord$cluster <- factor(data$cluster)

ind.coord$SurgMethod <- data$SurgMethod
# Data inspection
head(ind.coord$SurgMethod)

# Percentage of variance explained by dimensions
eigenvalue <- round(get_eigenvalue(res.pca), 1)
variance.percent <- eigenvalue$variance.percent
head(eigenvalue)


#Visualize k-means clusters
#Color individuals according to the cluster groups
ggscatter(
  ind.coord, x = "Dim.1", y = "Dim.2", 
  color = "cluster", palette = "npg", ellipse = TRUE, ellipse.type = "convex",
  shape = "SurgMethod", size = 1.5,  legend = "right", ggtheme = theme_bw(),
  xlab = paste0("Dim 1 (", variance.percent[1], "% )" ),
  ylab = paste0("Dim 2 (", variance.percent[2], "% )" )
) +
  stat_mean(aes(color = cluster), size = 4)