Practice_task_5_ML.Rmd

---
title: "Practice_task_5_ML"
output: html_document
---

```{r}
library(tidyverse)
library(stringi)
library(ggplot2)
library(tidyr)
library(dplyr)
library(glmnet)
library(naniar)
library(simputation)
```

```{python}
#!pip install pandas 
#!pip install matplotlib 
#!pip install seaborn 
import numpy as np 
import pandas as pd 
import matplotlib.pyplot as plt
import seaborn as sns
```

```{r}
data_r = read.csv('secondary_data.csv', sep = ";") 
data_r = data_r[, c(1,2,6,10)]
```


```{r}
data_r$class = str_replace(data_r$class, "p", "0")
data_r$class = str_replace(data_r$class, "e", "1")
data_r$does.bruise.or.bleed = str_replace(data_r$does.bruise.or.bleed, "f", "0")
data_r$does.bruise.or.bleed = str_replace(data_r$does.bruise.or.bleed, "t", "1")
```

```{r}
data_r$class = as.numeric(data_r$class)
data_r$does.bruise.or.bleed  = as.numeric(data_r$does.bruise.or.bleed )
```

```{r}
data_r %>%
  is.na() %>% 
  colSums()
```

```{r}
summary(data_r)
```


```{r}
hist(data_r$cap.diameter)
hist(data_r$stem.height)
boxplot(data_r$cap.diameter, col = "mediumturquoise", 
        ylab = 'Cap diameter',  main="Boxplot of cap diameter")
boxplot(data_r$stem.height, col = "mediumturquoise",
        ylab = 'Stem Height',  main="Boxplot of stem height")
```

```{r}
dn = quantile(data_r$cap.diameter, 0.25) - (quantile(data_r$cap.diameter, 0.75)- quantile(data_r$cap.diameter, 0.25)) * 1.5
up = quantile(data_r$cap.diameter, 0.75) + (quantile(data_r$cap.diameter, 0.75)- quantile(data_r$cap.diameter, 0.25)) * 1.5
data_r = filter(data_r, cap.diameter >= dn)
data_r = filter(data_r, cap.diameter <= up)
```

```{r}
dn1 = quantile(data_r$stem.height, 0.25) - (quantile(data_r$stem.height, 0.75)- quantile(data_r$stem.height, 0.25)) * 1.5
up1 = quantile(data_r$stem.height, 0.75) + (quantile(data_r$stem.height, 0.75)- quantile(data_r$stem.height, 0.25)) * 1.5
data_r = filter(data_r, stem.height >= dn1)
data_r = filter(data_r, stem.height <= up1)
```

```{r}
hist(data_r$cap.diameter, col = "mediumturquoise")
hist(data_r$stem.height, col = "mediumturquoise")
boxplot(data_r$cap.diameter, col = "mediumturquoise", 
        ylab = 'Cap diameter',  main="Boxplot of cap diameter")
boxplot(data_r$stem.height, col = "mediumturquoise",
        ylab = 'Stem Height',  main="Boxplot of stem height")
```

```{r}
write.csv(data_r,"/Users/olesyamba/Downloads/ML/HW_5/data.csv", row.names = FALSE)
```

```{python}
data_python = pd.read_csv("/Users/olesyamba/Downloads/ML/HW_5/data.csv")
data_python1=data_python
```

```{python}
fig, (ax1, ax2) = plt.subplots(1,2,figsize=(15,15))

sns.kdeplot(
   data=data_python, x=data_python['stem.height'], hue = 'class',
   fill=True, common_norm=False, palette="crest",
   alpha=.5, linewidth=0, ax=ax1
)
sns.kdeplot(
   data=data_python, x=data_python['stem.height'],
   fill=False, common_norm=False, color="royalblue",
   alpha=1, linewidth=1, ax=ax1
)
ax1.set_xlabel("Stem height") 
ax1.set_title(label = "Density of stem height")

sns.kdeplot(
   data=data_python, x=data_python['cap.diameter'], hue = 'class',
   fill=True, common_norm=False, palette="crest",
   alpha=.5, linewidth=0, ax=ax2
)
sns.kdeplot(
   data=data_python, x=data_python['cap.diameter'],
   fill=False, common_norm=False, color="royalblue",
   alpha=1, linewidth=1, ax=ax2
)
ax2.set_xlabel("Cap diameter")  
ax2.set_title(label = "Density of cap diameter")

plt.subplots_adjust(wspace=0.3, hspace=0.3)
plt.show()
```

```{python}
fig1, (ax3, ax4) = plt.subplots(1,2,figsize=(30,50))

sns.kdeplot(
   data=data_python, x=data_python['stem.height'], hue = 'does.bruise.or.bleed',
   fill=True, common_norm=False, palette="crest",
   alpha=.5, linewidth=0, ax=ax3
)
sns.kdeplot(
   data=data_python, x=data_python['stem.height'],
   fill=False, common_norm=False, color="royalblue",
   alpha=1, linewidth=1, ax=ax3
)
ax3.set_xlabel("stem.height")  
ax3.set_title(label = "Density of stem height")

sns.kdeplot(
   data=data_python, x=data_python['cap.diameter'], hue = 'does.bruise.or.bleed',
   fill=True, common_norm=False, palette="crest",
   alpha=.5, linewidth=0, ax=ax4
)
sns.kdeplot(
   data=data_python, x=data_python['cap.diameter'],
   fill=False, common_norm=False, color="royalblue",
   alpha=1, linewidth=1, ax=ax4
)
ax4.set_xlabel("cap.diameter")  
ax4.set_title(label = "Density of cap diameter")
sns.move_legend(ax3, "upper right")
plt.subplots_adjust(wspace=0.3, hspace=0.3)
plt.show()
```

```{r}
ks.test(data_r$cap.diameter, 'pnorm') 
```

```{r}
ks.test(data_r$stem.height, 'pnorm') 
```

```{r}
library(RColorBrewer)
library(ggpubr)
ggqqplot(data_r$cap.diameter, palette=RColorBrewer::brewer.pal(7, "YlGnBu"), col = "mediumturquoise", title = "QQPlot of cap diameter")
```

```{r}
library(ggpubr)
ggqqplot(data_r$stem.height, title = "QQPlot of stem height", col = "mediumturquoise")
```

```{python}
sns.jointplot(data = data_python, x=data_python['class'], y=data_python['cap.diameter'], kind="kde", fill = True, color="mediumturquoise")
plt.suptitle("Simultaneous distribution of class and cap diameter")
plt.show()

sns.jointplot(data = data_python, x=data_python['class'], y=data_python['does.bruise.or.bleed'], kind="kde", fill = True, color="olivedrab")
plt.suptitle("Simultaneous distribution of class  and does bruise or bleed")
plt.show()

sns.jointplot(data = data_python, x=data_python['class'], y=data_python['stem.height'], kind="kde", fill = True, color="skyblue")
plt.suptitle("Simultaneous distribution of class and stem height")
plt.show()

sns.jointplot(data = data_python, x=data_python['cap.diameter'], y=data_python['does.bruise.or.bleed'], kind="kde", fill = True, color="magenta")
plt.suptitle("Simultaneous distribution of cap diameter and does bruise or bleed")
plt.show()

sns.jointplot(data = data_python, x=data_python['cap.diameter'], y=data_python['stem.height'], kind="kde", fill = True, color="mediumslateblue")
plt.suptitle("Simultaneous distribution of cap diameter and stem height")
plt.show()

sns.jointplot(data = data_python, x=data_python['does.bruise.or.bleed'], y=data_python['stem.height'], kind="kde", fill = True, color="gold")
plt.suptitle("Simultaneous distribution of does bruise or bleed and stem height")
plt.show()
```


```{python}
class1 = data_python[data_python['class']==1]
class0 = data_python[data_python['class']==0]

doesbruise1 = data_python[data_python['does.bruise.or.bleed']==1]
doesbruise0 = data_python[data_python['does.bruise.or.bleed']==0]
```

```{python}
#!pip install scipy 
from scipy.stats import kstest

stat, p_value = kstest(class1['cap.diameter'], class0['cap.diameter'])
print(f" Kolmogorov-Smirnov Test for different values of class on variable cap.diameter: statistic={stat:.4f}, p-value={p_value:.4f} \n \n")

stat, p_value = kstest(class1['stem.height'], class0['stem.height'])
print(f" Kolmogorov-Smirnov Test for different values of class on variable stem.height: statistic={stat:.4f}, p-value={p_value:.4f} \n \n")

stat, p_value = kstest(doesbruise1['cap.diameter'], doesbruise0['cap.diameter'])
print(f" Kolmogorov-Smirnov Test for different values of does.bruise.or.bleed on variable cap.diameter: statistic={stat:.4f}, p-value={p_value:.4f} \n \n")

stat, p_value = kstest(doesbruise1['stem.height'], doesbruise0['stem.height'])
print(f" Kolmogorov-Smirnov Test for different values of does.bruise.or.bleed on variable stem.height: statistic={stat:.4f}, p-value={p_value:.4f} \n \n")
```

```{r}
data_r$class = as.integer(data_r$class)
data_r$cap.diameter = as.integer(data_r$cap.diameter)
data_r$does.bruise.or.bleed = as.integer(data_r$does.bruise.or.bleed)
data_r$stem.height = as.integer(data_r$stem.height)
pca <- prcomp(data_r, center = TRUE, scale = TRUE)

summary(pca)
```


```{r}
pca$rotation
```

```{r}
#install.packages("factoextra")
library(factoextra)

fviz_eig(pca, barfill = "mediumturquoise", linecolor = "darkcyan", ggtheme = theme_gray(), main = "Principal components explained variance")
```


```{r}
fviz_pca_ind(pca,
             col.ind = "cos2", # Color by the quality of representation
             gradient.cols = c("#00AFBB", "#E7B800", "#FC4E07"),
             repel = TRUE # Avoid text overlapping
             )
```

```{r}
fviz_pca_var(pca,
             col.var = "contrib", # Color by contributions to the PC
             gradient.cols = c("#00AFBB", "#E7B800", "#FC4E07"),
             repel = TRUE     # Avoid text overlapping
             )
```


```{r}
fviz_pca_biplot(pca, repel = TRUE,
                col.var = "#00AFBB", # Variables color
                col.ind = "#FC4E07"  # Individuals color
                )
```

```{r}
groups <- as.factor(data_r$class)
fviz_pca_ind(pca,
             col.ind = groups, # color by groups
             palette = c("#00AFBB",  "#FC4E07"),
             addEllipses = TRUE, # Concentration ellipses
             ellipse.type = "confidence",
             legend.title = "Groups",
             repel = TRUE
             )
```
```{r}
groups <- as.factor(data_r$does.bruise.or.bleed)
fviz_pca_ind(pca,
             col.ind = groups, # color by groups
             palette = c("#00AFBB",  "#FC4E07"),
             addEllipses = TRUE, # Concentration ellipses
             ellipse.type = "confidence",
             legend.title = "Groups",
             repel = TRUE
             )
```

```{r}
pca.coord <- pca$x
pca.coord = pca.coord[, 1:2]
```

```{r}
class_pca <- as.factor(data_r$class)
library(car)
scatterplot(pca.coord[, 1],
            pca.coord[, 2],
            group = class_pca,
            regLine=TRUE,
            legend=TRUE,
            ellipse=TRUE,
            grid=TRUE,
            smooth=list(smoother=loessLine, style="line", alpha=0.5, border=TRUE, vertical=TRUE, col.var = "grey"),
            boxplots = TRUE,
            ylab='PC2',
            xlab='PC1',
            main = "Class differentianting in PC's",
            col=c("#00AFBB",  "#FC4E07"))
```

```{r}
bruised_pca <- as.factor(data_r$does.bruise.or.bleed)
scatterplot(pca.coord[, 1],
            pca.coord[, 2],
            group = bruised_pca,
            regLine=TRUE,
            legend=TRUE,
            ellipse=TRUE,
            grid=TRUE,
            smooth=list(smoother=loessLine, style="line", alpha=0.5, border=TRUE, vertical=TRUE, col.var = "grey"),
            boxplots = TRUE,
            ylab='PC2',
            xlab='PC1',
            main = "Bruised differentianting in PC's",
            col=c("#00AFBB",  "#FC4E07"))
```
```{r}
components <- data.frame(pca.coord)
#components['index'] = seq(1, length(components), 1)
```

```{r}
colnames(components) <- c("PC1", "PC2")
#library(ggplot2)
#ggplot(components, aes(x= PC1, y = PC2))+geom_point()+theme_minimal()
```

```{r}
library(stats)
#Метод сингл 
dist_obs = dist(components, method = "euclidean")
```

```{r}
hc_obs = hclust(dist_obs, method = "single") 
```

```{r}
H <- data.frame (first_column  = c(0, 1),
                  second_column = c(0.37, 0.37))
#Раскрасим кластеры на дендрограмме
library(dendextend)
dend_stock <- as.dendrogram(hc_obs)
dend_stock <- color_branches(dend_stock, h = 0.5)
plot(dend_stock, main = "Cluster dendrogram | Single method", xlab = "Euclidian distances", ylab = "Height")
lines(H, cex = .5, col = "dark red")
#Не раскрашивает из-за большой размерности данных:(
```


```{r}
library(dplyr) 
clust_hc_obs = cutree(hc_obs, h = 0.37) 
data = mutate(components, clusters = clust_hc_obs)
```


```{r}
data %>%
count(clusters) 
#4 кластера
```


```{r}
ggplot(data, aes(x=PC1, y = PC2, color = factor(clusters))) +geom_point()+theme_grey() + ggtitle("Clusters in PC's") 
```


```{r}
#Выведем средние значения по всем кластерам
data %>% 
  group_by(clusters) %>% 
  summarise_all(list(mean))
```

```{r }
#Применяем метод k-means и рисуем график локтя
components1 <- components[,-1]
library(purrr)

tot_withinss <- map_dbl(1:10, function(k){
model <- kmeans(x = components1, centers = k)
model$tot.withinss
})
```

```{r}
elbow_df <- data.frame(
k = 1:10,
tot_withinss = tot_withinss
)

```

```{r}
ggplot(elbow_df, aes(x = k, y = tot_withinss)) +
geom_line(colour = "darkcyan") + 
scale_x_continuous(breaks = 1:10) +
  ggtitle("График локтя") + 
  xlab("Количество кластеров")+
  ylab("Совокупное расстояние между наблюдениями в кластере")

# Согласно графику локтя после добавления 4-го кластера идет не радикальное снижение расстояний между наблюдениями в кластере, поэтому необходимо выделить 4 кластера
```

```{r}
set.seed(42)
components_kmeans <- kmeans(components1, centers = 4)
```

```{r}
data1 = mutate(components, clusters = components_kmeans$cluster)
ggplot(data1, aes(x=PC1, y = PC2, color = factor(clusters))) +geom_point()+theme_grey() + ggtitle("Clusters in PC's | Kmeans method")
```

```{r}
#Выведем средние значения по всем кластерам
data1 %>% 
  group_by(clusters) %>% 
  summarise_all(list(mean))
```
```{r}
#install.packages('boot',dep=TRUE)
library(boot)
```

```{r}
components2 <- data.frame(components[,-1])
components3 <- data.frame(components[,-1])

function_1 <- function(data, i){
  d2 <- data[i,] 
  set.seed(42)
  components_kmeans <- kmeans(d2, centers = 4)
  components2 = mutate(components2, cluster = components_kmeans$cluster)
  return(components_kmeans$cluster)
}
```

```{r}
set.seed(1)
bootstrap_clusters <- boot(components3,function_1,R=10)
```

```{r}
summary(bootstrap_clusters)
```

```{r}
boot_samples_clusters = data.frame(bootstrap_clusters$t)
boot_samples_clusters = t(boot_samples_clusters)
```

```{r}
boot_samples_clusters = data.frame(boot_samples_clusters) 
colnames(boot_samples_clusters) = c("Sample1", "Sample2", "Sample3", "Sample4", "Sample5", "Sample6", "Sample7", "Sample8", "Sample9", "Sample10")
```

```{r}
boot_samples_clusters %>%
    group_by(Sample1) %>%
    count()

boot_samples_clusters %>%
    group_by(Sample2) %>%
    count()

boot_samples_clusters %>%
    group_by(Sample3) %>%
    count()

boot_samples_clusters %>%
    group_by(Sample4) %>%
    count()

boot_samples_clusters %>%
    group_by(Sample5) %>%
    count()

boot_samples_clusters %>%
    group_by(Sample6) %>%
    count()

boot_samples_clusters %>%
    group_by(Sample7) %>%
    count()

boot_samples_clusters %>%
    group_by(Sample8) %>%
    count()

boot_samples_clusters %>%
    group_by(Sample9) %>%
    count()

boot_samples_clusters %>%
    group_by(Sample10) %>%
    count()
#4 кластера
```

```{r}
data_boot = mutate(components, clusters = bootstrap_clusters$t0)
ggplot(data_boot, aes(x=PC1, y = PC2, color = factor(clusters))) +geom_point()+theme_grey() + ggtitle("Original clusters in PC's | bootstrap version")
```