In [1]:
import pandas as pd
import numpy as np
import math
from time import gmtime, strftime #DateTime Formatting
import matplotlib.pyplot as plt #Basic Plotting
import seaborn as sns #Advanced Plotting
import datetime
import warnings

from sklearn.cluster import KMeans
from sklearn import metrics
from scipy.spatial.distance import cdist
import numpy as np
import matplotlib.pyplot as plt

#pre-processing
from sklearn.preprocessing import OneHotEncoder
from sklearn.preprocessing import StandardScaler

warnings.filterwarnings('ignore')
pd.set_option('display.max_columns', None)
pd.set_option('display.max_rows', None)
pd.options.display.float_format = '{:.3f}'.format
In [2]:
import os
orig_dir = os.getcwd()
os.chdir("C:\\Users\\gusahil\\Downloads")
os.getcwd()
Out[2]:
'C:\\Users\\gusahil\\Downloads'
In [7]:
wine = pd.read_csv('wine_data_train.csv', sep=",").reset_index(drop = True).drop('ID', axis=1)
wine.head(3)
Out[7]:
fixed acidity volatile acidity citric acid residual sugar chlorides free sulfur dioxide total sulfur dioxide density pH sulphates alcohol quality color
0 7.400 0.700 0.000 1.900 0.076 11.000 34.000 0.998 3.510 0.560 9.400 5 R
1 7.800 0.880 0.000 2.600 0.098 25.000 67.000 0.997 3.200 0.680 9.800 5 R
2 7.800 0.760 0.040 2.300 0.092 15.000 54.000 0.997 3.260 0.650 9.800 5 R
In [12]:
wine.describe().T
Out[12]:
count mean std min 25% 50% 75% max
fixed acidity 5997.000 7.212 1.301 3.800 6.400 7.000 7.700 15.900
volatile acidity 5997.000 0.339 0.164 0.080 0.230 0.290 0.400 1.580
citric acid 5997.000 0.319 0.145 0.000 0.250 0.310 0.390 1.660
residual sugar 5997.000 5.445 4.757 0.600 1.800 3.000 8.100 65.800
chlorides 5997.000 0.056 0.034 0.009 0.038 0.047 0.064 0.610
free sulfur dioxide 5997.000 30.695 17.913 1.000 17.000 29.000 42.000 289.000
total sulfur dioxide 5997.000 116.313 56.619 6.000 78.000 119.000 156.000 440.000
density 5997.000 0.995 0.003 0.987 0.992 0.995 0.997 1.039
pH 5997.000 3.219 0.160 2.720 3.110 3.210 3.320 4.010
sulphates 5997.000 0.532 0.150 0.220 0.430 0.510 0.600 2.000
alcohol 5997.000 10.491 1.199 0.099 9.500 10.300 11.300 14.900
quality 5997.000 5.322 1.957 -1.000 5.000 6.000 6.000 9.000
In [10]:
np.unique(wine['quality'], return_counts = True)
Out[10]:
(array([-1,  3,  4,  5,  6,  7,  8,  9], dtype=int64),
 array([ 436,   26,  178, 1825, 2454,  916,  158,    4], dtype=int64))
In [11]:
np.unique(wine['color'], return_counts = True)
Out[11]:
(array(['R', 'W'], dtype=object), array([1458, 4539], dtype=int64))
In [14]:
#wine.isnull()
In [15]:
# Feature PreProcessing
In [17]:
wine.loc[wine['quality'] == -1, 'quality'] = np.median(wine.loc[wine['quality'] != -1, 'quality'])
In [18]:
np.unique(wine['quality'], return_counts = True)
Out[18]:
(array([3., 4., 5., 6., 7., 8., 9.]),
 array([  26,  178, 1825, 2890,  916,  158,    4], dtype=int64))
In [21]:
# use pd.concat to join the new columns with your original dataframe
wine = pd.concat([wine,pd.get_dummies(wine['color'], prefix='color')],axis=1).drop(['color','color_R'], axis=1)
wine.head()
Out[21]:
fixed acidity volatile acidity citric acid residual sugar chlorides free sulfur dioxide total sulfur dioxide density pH sulphates alcohol quality color_W
0 7.400 0.700 0.000 1.900 0.076 11.000 34.000 0.998 3.510 0.560 9.400 5.000 0
1 7.800 0.880 0.000 2.600 0.098 25.000 67.000 0.997 3.200 0.680 9.800 5.000 0
2 7.800 0.760 0.040 2.300 0.092 15.000 54.000 0.997 3.260 0.650 9.800 5.000 0
3 11.200 0.280 0.560 1.900 0.075 17.000 60.000 0.998 3.160 0.580 9.800 6.000 0
4 7.400 0.700 0.000 1.900 0.076 11.000 34.000 0.998 3.510 0.560 9.400 5.000 0
In [23]:
# Create scalar
scaler = StandardScaler()

# Fit on training data
wine_normalize = scaler.fit_transform(wine)
wine_normalize_df = pd.DataFrame(wine_normalize, columns = wine.columns)
                            
wine_normalize_df.head()
Out[23]:
fixed acidity volatile acidity citric acid residual sugar chlorides free sulfur dioxide total sulfur dioxide density pH sulphates alcohol quality color_W
0 0.144 2.199 -2.193 -0.745 0.592 -1.100 -1.454 1.035 1.819 0.190 -0.910 -0.995 -1.764
1 0.452 3.295 -2.193 -0.598 1.236 -0.318 -0.871 0.703 -0.117 0.992 -0.576 -0.995 -1.764
2 0.452 2.565 -1.918 -0.661 1.060 -0.876 -1.101 0.769 0.257 0.792 -0.576 -0.995 -1.764
3 3.065 -0.359 1.659 -0.745 0.563 -0.765 -0.995 1.102 -0.367 0.324 -0.576 0.203 -1.764
4 0.144 2.199 -2.193 -0.745 0.592 -1.100 -1.454 1.035 1.819 0.190 -0.910 -0.995 -1.764
In [25]:
kmeans = KMeans(n_clusters=5, random_state=0).fit(wine_normalize_df)
In [27]:
np.unique(kmeans.labels_, return_counts=True)
#kmeans.cluster_centers_
Out[27]:
(array([0, 1, 2, 3, 4]), array([1572,  882, 1513,  567, 1463], dtype=int64))
In [28]:
kmeans.cluster_centers_
Out[28]:
array([[-2.12365378e-01, -3.82860071e-01,  1.85145569e-05,
        -3.43165071e-01, -1.98305146e-01, -1.55977855e-01,
         2.18141023e-01, -3.45851269e-01, -9.20564940e-02,
        -2.98818894e-01, -2.44048260e-01, -4.32977724e-01,
         5.62307678e-01],
       [ 1.26301966e-01,  1.67003687e+00, -1.21319547e+00,
        -6.33714276e-01,  7.28073009e-01, -7.93156267e-01,
        -1.18934934e+00,  5.32953899e-01,  9.88569665e-01,
         4.74683880e-01, -2.13600393e-01, -4.59961888e-01,
        -1.74587394e+00],
       [-4.43835014e-01, -3.88821859e-01,  4.25562715e-02,
        -3.92072979e-01, -5.69245624e-01,  5.09618999e-02,
        -2.37258214e-02, -1.16143157e+00, -4.60975437e-02,
        -2.75051035e-01,  1.14766456e+00,  8.93839357e-01,
         5.46742611e-01],
       [ 2.01609498e+00,  4.26959534e-01,  9.91087786e-01,
        -5.90955413e-01,  1.23390936e+00, -9.10575725e-01,
        -1.33492171e+00,  9.50530406e-01, -4.50467505e-02,
         1.44545797e+00,  1.23698281e-01,  1.39520994e-01,
        -1.75622345e+00],
       [-1.72735172e-01, -3.57090194e-01,  3.00221314e-01,
         1.38525838e+00, -1.15805480e-01,  9.45987694e-01,
         1.02489352e+00,  8.83040675e-01, -4.30551791e-01,
        -2.42184312e-01, -8.45236024e-01, -2.37652608e-01,
         5.63572476e-01]])
In [30]:
distortions = []
inertias = []
mapping1 = {}
mapping2 = {}
K = range(1, 10)

X = wine_normalize_df
 
for k in K:
    # Building and fitting the model
    kmeanModel = KMeans(n_clusters=k).fit(X)
    kmeanModel.fit(X)
 
    distortions.append(sum(np.min(cdist(X, kmeanModel.cluster_centers_,
                                        'euclidean'), axis=1)) / X.shape[0])
    inertias.append(kmeanModel.inertia_)
 
    mapping1[k] = sum(np.min(cdist(X, kmeanModel.cluster_centers_,
                                   'euclidean'), axis=1)) / X.shape[0]
    mapping2[k] = kmeanModel.inertia_
In [31]:
# It is calculated as the average of the squared distances from the cluster centers of the respective clusters. Typically, the Euclidean distance metric is used.

plt.plot(K, distortions, 'bx-')
plt.xlabel('Values of K')
plt.ylabel('Distortion')
plt.title('The Elbow Method using Distortion')
plt.show()
In [32]:
for key, val in mapping1.items():
    print(f'{key} : {val}')

1 : 3.398663442230733
2 : 2.917028762948958
3 : 2.6162602589705894
4 : 2.5006083589797243
5 : 2.3995061784912033
6 : 2.37215372581441
7 : 2.3174643113438664
8 : 2.2818834977325357
9 : 2.236834289199079
In [33]:
cluster_means = pd.DataFrame(kmeans.cluster_centers_, columns = wine.columns)
cluster_means
Out[33]:
fixed acidity volatile acidity citric acid residual sugar chlorides free sulfur dioxide total sulfur dioxide density pH sulphates alcohol quality color_W
0 -0.212 -0.383 0.000 -0.343 -0.198 -0.156 0.218 -0.346 -0.092 -0.299 -0.244 -0.433 0.562
1 0.126 1.670 -1.213 -0.634 0.728 -0.793 -1.189 0.533 0.989 0.475 -0.214 -0.460 -1.746
2 -0.444 -0.389 0.043 -0.392 -0.569 0.051 -0.024 -1.161 -0.046 -0.275 1.148 0.894 0.547
3 2.016 0.427 0.991 -0.591 1.234 -0.911 -1.335 0.951 -0.045 1.445 0.124 0.140 -1.756
4 -0.173 -0.357 0.300 1.385 -0.116 0.946 1.025 0.883 -0.431 -0.242 -0.845 -0.238 0.564