NBAClustering.Rmd

---
title: "NBAClustering"
author: "Eric Drew"
date: "2022-10-12"
output: pdf_document
---

```{r setup, include=FALSE}
library(tidyverse)
library(datasets)
library(arules)
library(arulesViz)
library(ggplot2)
library(dplyr)
library(rpart)
library(rpart.plot)
library(TH.data)
library(ISLR2)
library(lattice)
library(stats)
library(rattle)
library(RColorBrewer)
#library(caret)
library(ROCR)
library(cluster)  
library(factoextra) 
library(gridExtra)
library(NbClust)
library(dendextend)
library(class)
library(ClustOfVar)
library(MASS)
library(kableExtra)
library(partykit)
library(dbscan)
```

```{r data}
#https://www.basketball-reference.com/leagues/NBA_2022_per_game.html
#per game Stats Data
pergame <- read.csv("C:\\Users\\ericd\\OneDrive - North Carolina State University\\Desktop\\PersonalProjects\\NBA-Clustering\\perGameStats.csv")

#Remove the "total" observation(for now)
pergame = pergame[order(pergame[,'Player'],-pergame[,'G']),]
pergame = pergame[!duplicated(pergame$Player),]


pergame <- pergame[,-c(2,3,4)]

```


```{r clean}
#missing values check
sum(is.na(pergame))

#remove rows with missing values
pergame <- na.omit(pergame)

#set player name to the row names and drop column
row.names(pergame) <- pergame$Player
pergame <- pergame[,-1]

pergame <- as.data.frame(pergame)


#filter to players who play more than 15mpg
pergame <- pergame %>% filter(pergame$PTS >= 13)


labelvec <- c('LeBron James','Joel Embiid','Kevin Durant','Luka Dončić','Rudy Gobert','Bam Adebayo',
                 'Anthony Davis', 'Montrezl Harrell','Aaron Gordon','Josh Hart','Reggie Jackson',
              'LaMelo Ball','Stephen Curry','Carmelo Anthony','Kyle Kuzma', 'Deandre Ayton', 'James Harden',
              'Jimmy Butler','Bradley Beal','Jalen Green')
```

```{r cluster}
scaleData <- scale(pergame)

clus2=kmeans(scaleData,centers=3,nstart = 10000)
clus2

fviz_nbclust(scaleData, kmeans, method = "wss",k.max = 9)
fviz_nbclust(scaleData, kmeans, method = "silhouette",k.max = 9)

fviz_cluster(clus2,data = scaleData,geom=c('text','point'), pointsize=.8,labelsize = 12,label.select=labelvec,
             main='Cluster Plot of NBA Players(min. 13PPG)')


#bind the clusters onto each players observation
profile.kmeans=cbind(pergame,clus2$cluster)

#create the summary statistics for the important per game stats
all.k=profile.kmeans %>% group_by(clus2$cluster) %>%
   summarise('Points/Game'=round(mean(PTS),2),'Rebs/Game'=round(mean(TRB),2),'AST/Game'=round(mean(AST),2),
             'Blocks/Game'=round(mean(BLK),2))


#rename clusters to what they are categorized as
all.k$`clus2$cluster` <- as.character(all.k$`clus2$cluster`)
all.k[1,1] = 'Role Players'
all.k[2,1] = 'Stars'
all.k[3,1] = 'Big Men'

names(all.k)[names(all.k) == 'clus2$cluster'] <- 'Player Category'

#print table with ordering the stars first
all.k[c(2,3,1),]

#league averages for the stats to be highlighted(stars:points. big men:blocks/rebounds)
mean(pergame$PTS) #18.8
mean(pergame$BLK) #0.6
mean(pergame$TRB) #5.9

NbClust(scaleData,method="kmeans",min.nc=2,max.nc = 4)
```