K Means Clustering.Rmd

---
title: "K Means Algorithim"
author: "Vijay S"
date: "18 May 2018"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
library(dplyr)
library(BBmisc)
```


```{r}
odi =  read.csv("odi-batting.csv")
odi$century = ifelse(odi$Runs > 99, 1, 0)
odi$ducks = ifelse(odi$Runs == 0, 1, 0)
odi$above_150 = ifelse(odi$Runs > 149, 1, 0)
odi$fifties = ifelse(odi$Runs > 49, 1, 0)
odi$missed_centuries = ifelse(odi$Runs > 90 & odi$Runs < 100, 1, 0)
View(odi)
```


```{r}
player_summary = odi %>% group_by(Player) %>% summarise(martches = n(),
                                                        total_runs = sum(Runs, na.rm = T),
                                                        avg_runs = mean(Runs, na.rm = T),
                                                        centuries = sum(century, na.rm = T),
                                                        ducks = sum(ducks, na.rm = T),
                                                        fifties = sum(fifties, na.rm = T),
                                                        above_150 = sum(above_150, na.rm = T),
                                                        missed_centuries = sum(missed_centuries, na.rm = T))
View(player_summary)
```


```{r}
top_players = player_summary %>% arrange(-total_runs) %>% head(100)
data_kmeans = top_players %>% select(-Player)
data_norm = normalize(data_kmeans, range = c(0,1), method = 'range')
model_kmeans = kmeans(data_norm, centers = 10)
top_players$cluster = model_kmeans$cluster
data_norm$cluster = model_kmeans$cluster
barplot(table(top_players$cluster))

```

# Cluster characteristics
```{r}
model_kmeans$centers
```
# Within square sums
```{r}
model_kmeans$withinss
```

## Between square sums
```{r}
model_kmeans$betweenss
```
## Plotting totalwithinss Vs Number of clusters
```{r}
model_kmeans$tot.withinss
```

```{r}
View(top_players)
dim(data_norm)
data_norm_2d = as.data.frame(cmdscale(dist(data_norm %>% select(-cluster))))
plot(data_norm_2d)
data_norm_2d$cluster = as.factor(data_norm$cluster)
library(ggplot2)
ggplot(data_norm_2d, aes(x = V1, y = V2, color = cluster)) + geom_point()

```

```{r}
hr=read.csv("HR Analytics.csv")
hr_sub = hr %>% select(Age, MonthlyIncome)
hr_sub$Age = as.numeric(hr_sub$Age)

hr_norm = normalize(hr_sub, method = 'range', range = c(0,1))
model_hr = kmeans(hr_norm, centers = 2)
hr_sub$cluster = as.factor(model_hr$cluster)
ggplot(hr_sub, aes(x = Age, y = MonthlyIncome, color = cluster)) + geom_jitter()



```

## Hierachical Clustering - It can also be used check if there is any hierachy in the rows as well
```{r}
hclust_model = hclust(dist(data_norm %>% select(-cluster)))
plot(hclust_model)

```
## Get cluster labels
```{r}
data_norm$cluster = cutree(hclust_model, k = 4)

data_norm_2d = cmdscale(dist(data_norm %>% select(-cluster)))
data_norm_2d = as.data.frame(data_norm_2d)
data_norm_2d$cluster = as.factor(data_norm$cluster)

ggplot(data_norm_2d, aes(x = V1, y = V2, color = cluster)) + geom_point()
plot(data_norm_2d)

```