Map > Problem Definition > Data Preparation > Data Exploration > Modeling > Evaluation > Clustering
 

Evaluation - Clustering
I. Data Preparation

1- Load libraries
library(data.table)
library(formattable)
library(plotrix)
library(dplyr)
library(Rtsne)
library(MASS)
library(xgboost)
library(factoextra)
library(caTools)
library(pROC)
library(caret)
library(gains)
library(lift)
library(cluster)
 

2- Read Expressions file
df <- read.csv("GSE74763_rawlog_expr.csv")
df2 <- df[,-1]
rownames(df2) <- df[,1]
expr <- transpose(df2)
rownames(expr) <- colnames(df2)
colnames(expr) <- rownames(df2)
dim(expr)
 

3- Read Samples file
targets <- read.csv("GSE74763_rawlog_targets.csv")
colnames(targets)
dim(targets)
 

4- Merge Expressions with Samples
data <- cbind(expr, targets)
colnames(data)
dim(data)
 
II. Hierarchical Clustering (Determining Optimal Clusters)
1- Elbow Method
fviz_nbclust(expr, FUN = hcut, method = "wss")

 
2- Average Silhouette Method
fviz_nbclust(expr, FUN = hcut, method = "silhouette")

 
3- Gap Statistic Method
gap_stat <- clusGap(expr, FUN = hcut, nstart = 25, K.max = 10, B = 50)
fviz_gap_stat(gap_stat)

 
III. K-Means Clustering (Determining Optimal Clusters)
1- Elbow Method
fviz_nbclust(expr, FUN = kmeans, method = "wss")

 
2- Average Silhouette Method
fviz_nbclust(expr, FUN = kmeans, method = "silhouette")

 
3- Gap Statistic Method
gap_stat <- clusGap(expr, FUN = kmeans, nstart = 25, K.max = 10, B = 50)
fviz_gap_stat(gap_stat)