library(tidyverse)

# Aufgabe 1
combine_df = read_csv('https://github.com/NikoStein/dma_data/raw/main/combine.csv')


#a) Wählen Sie geeignete Variablen aus und clustern Sie die Daten mittels kMeans um 
# Positionen mit ähnlichen Anforderungen an die Spieler zu identifizieren. 
combine_df %>%
  ggplot(aes(x=weight_kg, y=bench, color = position, alpha = 0.1)) + 
  geom_point() + 
  theme_bw()

combine_df %>%
  select(-bench) %>%
  na.omit() -> combine_df

combine_df %>%
  select(height_in, weight_kg, vertical_cm, broad_jump_cm, x3cone, shuttle, x40yd) %>%
  scale() %>%
  kmeans(centers = 3) -> clustering

combine_df$cluster = clustering$cluster

combine_df %>%
  group_by(cluster, position) %>%
  summarise(count = n()) %>%
    ggplot(aes(x=position, y=as.factor(cluster), size=count)) + 
  geom_point() + 
  xlab('Position') + ylab("Cluster") +
  theme_bw()


#b)
combine_df %>%
  select(height_in, weight_kg, vertical_cm, broad_jump_cm, x3cone, shuttle, x40yd) %>%
  scale() %>%
  dist() -> dist_mat

hclust_avg = hclust(dist_mat, method = 'complete')
plot(hclust_avg)
cut_avg = cutree(hclust_avg, k=4)

combine_df$cluster = cut_avg

combine_df %>%
  group_by(cluster, position) %>%
  summarise(count = n()) %>%
  ggplot(aes(x=position, y=as.factor(cluster), size=count)) + 
  geom_point() + 
  xlab('Position') + ylab("Cluster") +
  theme_bw()


# Aufgabe 2
mall_df = read_csv('https://github.com/NikoStein/dma_data/raw/main/Mall_Customers.csv')

# a) Clustern Sie die Kunden mittels des kMeans Algorithmus. 
mall_df %>%
  select(-CustomerID) %>%
  mutate(Gender = if_else(Gender == 'Male', 0, 1)) %>%
  scale() %>%
  kmeans(centers = 3) -> clusters

clusters


# b) Testen Sie verschiedene Anzahlen an Clustern. Welche Anzahl erscheint Ihnen sinnvoll?
cluster_data = NULL

for (i in 2:20){
  mall_df %>%
    select(-CustomerID) %>%
    mutate(Gender = if_else(Gender == 'Male', 0, 1)) %>%
    scale() %>%
    kmeans(centers = i) -> clusters
  
  cluster_data %>%
    rbind(data.frame(centers=i, between=clusters$betweenss, within=clusters$tot.withinss)) -> cluster_data
}


cluster_data %>%
  ggplot(aes(x=centers)) + 
  geom_line(aes(y=between), color='red') +
  geom_line(aes(y=within), color='blue') +
  theme_bw()