import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
%matplotlib inline
import os
import csv
sns.set_style('darkgrid')
%config InlineBackend.figure_format = 'retina'
%precision %.2f  ## magic command to display 2 decimal places

# Clustering package
from sklearn.cluster import KMeans
from sklearn.preprocessing import StandardScaler, normalize
from sklearn.metrics import silhouette_score

import warnings
warnings.filterwarnings('ignore')

path = r"D:/UoW/Semester 1/Quantitative Studies -BSMM 8320/Excel Projects/Project 2/Output" 

# locate the data to be uploaded 
for i in os.listdir(path):
    if i.endswith(".csv"): # select ony the csv files
        print(i)

clustering_raw_data = pd.read_csv(path+'/clustering_raw_data.csv')

clustering_raw_data.columns

num = clustering_raw_data.select_dtypes(include=np.number)  # Get numeric columns
n = num.shape[1]  # Number of cols

fig, axes = plt.subplots(n, 1, figsize=(24/2.54, 70/2.54))  # create subplots with n rows and 1 column

for ax, col in zip(axes, num):  # For each column...
    sns.distplot(num[col], ax=ax)   # Plot histogaerm
    ax.set(ylabel= col)
    ax.axvline(num[col].mean(), c='k')  # Plot mean

plt.figure(figsize=(10,8))
sns.heatmap(clustering_raw_data.corr(),
            annot=True,
            linewidths=.5,
            center=0,
            cbar=False,
            cmap="YlGnBu")
plt.show()

### List columns that can be dropped
columns_drop = ['Policy Type', 'Vehicle Size']

clustering_raw_data.drop(columns=columns_drop, inplace = True)

g = sns.PairGrid(clustering_raw_data)
g.map(sns.scatterplot);

scaler = StandardScaler() 
scaled_df = scaler.fit_transform(clustering_raw_data) 
scaled_df

normalized_df = normalize(scaled_df) 
normalized_df = pd.DataFrame(normalized_df)
normalized_df

### The Elbow Method

sse = {}

for k in range(1, 16):
    kmeans = KMeans(n_clusters=k, init='k-means++',random_state= 0 ,max_iter=300).fit(normalized_df)
    sse[k] = kmeans.inertia_ # Inertia: Sum of distances of samples to their closest cluster center

    plt.figure()
plt.plot(list(sse.keys()), list(sse.values()),marker='*')
plt.title('Elbow Method')                               # Set plot title
plt.xlabel('Number of clusters')                        # Set x axis name
plt.ylabel('Within Cluster Sum of Squares (WCSS)') 
plt.show()

#### The Silhouette Coefficient Method

silhouette_scores = [] 

for n_cluster in range(2, 15):
    silhouette_scores.append( 
        silhouette_score(normalized_df, KMeans(n_clusters = n_cluster).fit_predict(normalized_df))) 
    
# Plotting a bar graph to compare the results 
k = [2, 3, 4, 5, 6,7, 8,9, 10,11, 12,13, 14] 
plt.bar(k, silhouette_scores) 
plt.xlabel('Number of clusters', fontsize = 10) 
plt.ylabel('Silhouette Score', fontsize = 10) 
plt.show() 

def cluster_summary(k,norm_data): ## k is no of cluster $ norm_data is normalized data
    final_clusters = KMeans(n_clusters = k).fit_predict(norm_data)
    final_clusters_data = pd.concat([pd.DataFrame(final_clusters),clustering_raw_data],axis=1)
    final_clusters_data.rename(columns={0:"Cluster_Label"}, inplace= True)
    metrics = final_clusters_data.groupby("Cluster_Label").size().reset_index(name = "Distribution")
#     metrics["Perc"] = metrics.groupby("Cluster_Label")["Distribution"].apply(lambda x: x/float(x.sum()))
    metrics["Perc"] = 100*metrics["Distribution"]/metrics["Distribution"].sum()
    return metrics
    
metrics = cluster_summary(5,normalized_df)   

metrics

kmeans1 = KMeans(n_clusters=5, init='k-means++',random_state= 0 ,max_iter=300).fit(normalized_df)
kmeans1.cluster_centers_