FarhadPiri-3-Clustering

Farhad Piri¶

Clustering algorithms in action¶

need the source data to try this for yourself? email me at farhad.peeri[at]gmail[dot]com¶

In [1]:

#!pip install zipcodes

In [2]:

#!pip install basemap

In [65]:

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import zipcodes as zcode
from sklearn.model_selection import train_test_split , GridSearchCV
from sklearn import metrics
from sklearn.preprocessing import StandardScaler
from sklearn.cluster import KMeans
from sklearn.metrics import silhouette_score
from warnings import filterwarnings
filterwarnings("ignore")

In [5]:

# Load the dataset
file_path = 'Customer_Data.csv'
data = pd.read_csv(file_path)

# Display the first few rows of the dataset to understand its structure
data.head()

Out[5]:

	CUST_ID	BALANCE	BALANCE_FREQUENCY	PURCHASES	ONEOFF_PURCHASES	INSTALLMENTS_PURCHASES	CASH_ADVANCE	PURCHASES_FREQUENCY	ONEOFF_PURCHASES_FREQUENCY	PURCHASES_INSTALLMENTS_FREQUENCY	CASH_ADVANCE_FREQUENCY	CASH_ADVANCE_TRX	PURCHASES_TRX	CREDIT_LIMIT	PAYMENTS	MINIMUM_PAYMENTS	PRC_FULL_PAYMENT	TENURE
0	C10001	40.900749	0.818182	95.40	0.00	95.4	0.000000	0.166667	0.000000	0.083333	0.000000	0	2	1000.0	201.802084	139.509787	0.000000	12
1	C10002	3202.467416	0.909091	0.00	0.00	0.0	6442.945483	0.000000	0.000000	0.000000	0.250000	4	0	7000.0	4103.032597	1072.340217	0.222222	12
2	C10003	2495.148862	1.000000	773.17	773.17	0.0	0.000000	1.000000	1.000000	0.000000	0.000000	0	12	7500.0	622.066742	627.284787	0.000000	12
3	C10004	1666.670542	0.636364	1499.00	1499.00	0.0	205.788017	0.083333	0.083333	0.000000	0.083333	1	1	7500.0	0.000000	NaN	0.000000	12
4	C10005	817.714335	1.000000	16.00	16.00	0.0	0.000000	0.083333	0.083333	0.000000	0.000000	0	1	1200.0	678.334763	244.791237	0.000000	12

Data Cleaning and Preprocessing:¶

In [10]:

# Checking for missing values and data types
missing_values = data.isnull().sum()
data_types = data.dtypes

# Summary statistics for detecting outliers
summary_statistics = data.describe()

missing_values, data_types, summary_statistics

Out[10]:

(CUST_ID                               0
 BALANCE                               0
 BALANCE_FREQUENCY                     0
 PURCHASES                             0
 ONEOFF_PURCHASES                      0
 INSTALLMENTS_PURCHASES                0
 CASH_ADVANCE                          0
 PURCHASES_FREQUENCY                   0
 ONEOFF_PURCHASES_FREQUENCY            0
 PURCHASES_INSTALLMENTS_FREQUENCY      0
 CASH_ADVANCE_FREQUENCY                0
 CASH_ADVANCE_TRX                      0
 PURCHASES_TRX                         0
 CREDIT_LIMIT                          1
 PAYMENTS                              0
 MINIMUM_PAYMENTS                    313
 PRC_FULL_PAYMENT                      0
 TENURE                                0
 dtype: int64,
 CUST_ID                              object
 BALANCE                             float64
 BALANCE_FREQUENCY                   float64
 PURCHASES                           float64
 ONEOFF_PURCHASES                    float64
 INSTALLMENTS_PURCHASES              float64
 CASH_ADVANCE                        float64
 PURCHASES_FREQUENCY                 float64
 ONEOFF_PURCHASES_FREQUENCY          float64
 PURCHASES_INSTALLMENTS_FREQUENCY    float64
 CASH_ADVANCE_FREQUENCY              float64
 CASH_ADVANCE_TRX                      int64
 PURCHASES_TRX                         int64
 CREDIT_LIMIT                        float64
 PAYMENTS                            float64
 MINIMUM_PAYMENTS                    float64
 PRC_FULL_PAYMENT                    float64
 TENURE                                int64
 dtype: object,
             BALANCE  BALANCE_FREQUENCY     PURCHASES  ONEOFF_PURCHASES  \
 count   8950.000000        8950.000000   8950.000000       8950.000000   
 mean    1564.474828           0.877271   1003.204834        592.437371   
 std     2081.531879           0.236904   2136.634782       1659.887917   
 min        0.000000           0.000000      0.000000          0.000000   
 25%      128.281915           0.888889     39.635000          0.000000   
 50%      873.385231           1.000000    361.280000         38.000000   
 75%     2054.140036           1.000000   1110.130000        577.405000   
 max    19043.138560           1.000000  49039.570000      40761.250000   
 
        INSTALLMENTS_PURCHASES  CASH_ADVANCE  PURCHASES_FREQUENCY  \
 count             8950.000000   8950.000000          8950.000000   
 mean               411.067645    978.871112             0.490351   
 std                904.338115   2097.163877             0.401371   
 min                  0.000000      0.000000             0.000000   
 25%                  0.000000      0.000000             0.083333   
 50%                 89.000000      0.000000             0.500000   
 75%                468.637500   1113.821139             0.916667   
 max              22500.000000  47137.211760             1.000000   
 
        ONEOFF_PURCHASES_FREQUENCY  PURCHASES_INSTALLMENTS_FREQUENCY  \
 count                 8950.000000                       8950.000000   
 mean                     0.202458                          0.364437   
 std                      0.298336                          0.397448   
 min                      0.000000                          0.000000   
 25%                      0.000000                          0.000000   
 50%                      0.083333                          0.166667   
 75%                      0.300000                          0.750000   
 max                      1.000000                          1.000000   
 
        CASH_ADVANCE_FREQUENCY  CASH_ADVANCE_TRX  PURCHASES_TRX  CREDIT_LIMIT  \
 count             8950.000000       8950.000000    8950.000000   8949.000000   
 mean                 0.135144          3.248827      14.709832   4494.449450   
 std                  0.200121          6.824647      24.857649   3638.815725   
 min                  0.000000          0.000000       0.000000     50.000000   
 25%                  0.000000          0.000000       1.000000   1600.000000   
 50%                  0.000000          0.000000       7.000000   3000.000000   
 75%                  0.222222          4.000000      17.000000   6500.000000   
 max                  1.500000        123.000000     358.000000  30000.000000   
 
            PAYMENTS  MINIMUM_PAYMENTS  PRC_FULL_PAYMENT       TENURE  
 count   8950.000000       8637.000000       8950.000000  8950.000000  
 mean    1733.143852        864.206542          0.153715    11.517318  
 std     2895.063757       2372.446607          0.292499     1.338331  
 min        0.000000          0.019163          0.000000     6.000000  
 25%      383.276166        169.123707          0.000000    12.000000  
 50%      856.901546        312.343947          0.000000    12.000000  
 75%     1901.134317        825.485459          0.142857    12.000000  
 max    50721.483360      76406.207520          1.000000    12.000000  )

In [151]:

data.drop('CUST_ID', axis = 1, inplace= True)
data.head()

Out[151]:

	BALANCE	BALANCE_FREQUENCY	PURCHASES	ONEOFF_PURCHASES	INSTALLMENTS_PURCHASES	CASH_ADVANCE	PURCHASES_FREQUENCY	ONEOFF_PURCHASES_FREQUENCY	PURCHASES_INSTALLMENTS_FREQUENCY	CASH_ADVANCE_FREQUENCY	CASH_ADVANCE_TRX	PURCHASES_TRX	CREDIT_LIMIT	PAYMENTS	MINIMUM_PAYMENTS	PRC_FULL_PAYMENT	TENURE
0	40.900749	0.818182	95.40	0.00	95.4	0.000000	0.166667	0.000000	0.083333	0.000000	0	2	1000.0	201.802084	139.509787	0.000000	12
1	3202.467416	0.909091	0.00	0.00	0.0	6442.945483	0.000000	0.000000	0.000000	0.250000	4	0	7000.0	4103.032597	1072.340217	0.222222	12
2	2495.148862	1.000000	773.17	773.17	0.0	0.000000	1.000000	1.000000	0.000000	0.000000	0	12	7500.0	622.066742	627.284787	0.000000	12
3	1666.670542	0.636364	1499.00	1499.00	0.0	205.788017	0.083333	0.083333	0.000000	0.083333	1	1	7500.0	0.000000	NaN	0.000000	12
4	817.714335	1.000000	16.00	16.00	0.0	0.000000	0.083333	0.083333	0.000000	0.000000	0	1	1200.0	678.334763	244.791237	0.000000	12

Missing Values:

For CREDIT_LIMIT, since only one value is missing, we can impute it with a central tendency measure (mean or median).

For MINIMUM_PAYMENTS, given the higher number of missing values, we need to decide whether to impute them (using mean, median, or a more complex imputation method) or to remove these records.

In [159]:

# Imputing missing value in CREDIT_LIMIT with the median
data['CREDIT_LIMIT'].fillna(data['CREDIT_LIMIT'].median(), inplace=True)

# Removing records with missing MINIMUM_PAYMENTS
data_cleaned = data.dropna(subset=['MINIMUM_PAYMENTS'])

# Checking the shape of the data before and after removing missing values
original_shape = data.shape
cleaned_shape = data_cleaned.shape

original_shape, cleaned_shape

Out[159]:

((8950, 17), (8637, 17))

In [163]:

data_cleaned['TENURE'].value_counts()
data_cleaned.head()

Out[163]:

	BALANCE	BALANCE_FREQUENCY	PURCHASES	ONEOFF_PURCHASES	INSTALLMENTS_PURCHASES	CASH_ADVANCE	PURCHASES_FREQUENCY	ONEOFF_PURCHASES_FREQUENCY	PURCHASES_INSTALLMENTS_FREQUENCY	CASH_ADVANCE_FREQUENCY	CASH_ADVANCE_TRX	PURCHASES_TRX	CREDIT_LIMIT	PAYMENTS	MINIMUM_PAYMENTS	PRC_FULL_PAYMENT	TENURE
0	40.900749	0.818182	95.40	0.00	95.40	0.000000	0.166667	0.000000	0.083333	0.00	0	2	1000.0	201.802084	139.509787	0.000000	12
1	3202.467416	0.909091	0.00	0.00	0.00	6442.945483	0.000000	0.000000	0.000000	0.25	4	0	7000.0	4103.032597	1072.340217	0.222222	12
2	2495.148862	1.000000	773.17	773.17	0.00	0.000000	1.000000	1.000000	0.000000	0.00	0	12	7500.0	622.066742	627.284787	0.000000	12
4	817.714335	1.000000	16.00	16.00	0.00	0.000000	0.083333	0.083333	0.000000	0.00	0	1	1200.0	678.334763	244.791237	0.000000	12
5	1809.828751	1.000000	1333.28	0.00	1333.28	0.000000	0.666667	0.000000	0.583333	0.00	0	8	1800.0	1400.057770	2407.246035	0.000000	12

In [46]:

# Selecting numerical columns for plotting
numerical_columns = data_cleaned.select_dtypes(include=['float64', 'int64']).columns

# Plotting boxplots for each numerical column to visualize outliers
plt.figure(figsize=(20, 15))
for i, col in enumerate(numerical_columns):
    plt.subplot(len(numerical_columns) // 3 + 1, 3, i + 1)
    sns.boxplot(y=data_cleaned[col])
    plt.title(col)

plt.tight_layout()
plt.show()

There are no issues in the data as noises

Let`s Normalize the data¶

In [165]:

scaler = StandardScaler()
scaled_data = scaler.fit_transform(data_cleaned)

# Checking the first few rows of the scaled data
scaled_data[:5]

Out[165]:

array([[-0.74455592, -0.3693992 , -0.42914833, -0.35913681, -0.35478902,
        -0.46863463, -0.82060062, -0.68622114, -0.71707755, -0.68200785,
        -0.47942277, -0.51757308, -0.96257337, -0.54389103, -0.30548157,
        -0.53768681,  0.35532361],
       [ 0.76425857,  0.0680405 , -0.47317466, -0.35913681, -0.45880689,
         2.56872771, -1.23595801, -0.68622114, -0.9264219 ,  0.55703709,
         0.09930182, -0.59700728,  0.67728465,  0.79695187,  0.08773463,
         0.21245101,  0.35532361],
       [ 0.42670046,  0.5054802 , -0.11636295,  0.0999561 , -0.45880689,
        -0.46863463,  1.25618136,  2.64681173, -0.9264219 , -0.68200785,
        -0.47942277, -0.12040209,  0.81393948, -0.39944714, -0.09986968,
        -0.53768681,  0.35532361],
       [-0.37383225,  0.5054802 , -0.46579079, -0.34963633, -0.45880689,
        -0.46863463, -1.02828056, -0.40846951, -0.9264219 , -0.68200785,
        -0.47942277, -0.55729018, -0.90791144, -0.38010797, -0.26110226,
        -0.53768681,  0.35532361],
       [ 0.0996408 ,  0.5054802 ,  0.14212329, -0.35913681,  0.99491363,
        -0.46863463,  0.42546907, -0.68622114,  0.53899358, -0.68200785,
        -0.47942277, -0.27927048, -0.74392564, -0.13205363,  0.65043775,
        -0.53768681,  0.35532361]])

In [117]:

# Selecting a subset of the dataset for visualization purposes
sampled_data = scaled_data.sample(n=100, random_state=42)

# Plotting scatter plots for each pair of features
# We are using pairplot from seaborn which automatically plots pairwise relationships in the dataset
sns.pairplot(sampled_data, diag_kind='kde', plot_kws={'alpha':0.6, 's':20, 'edgecolor':'k'}, height=2)

plt.show()

we take 13 and 14 as our x, y to show in 2d

Model¶

K-Means

In [118]:

kmeans_set = {'init':'random', 'n_init':10, 'max_iter':300, 'random_state':42}

In [119]:

# Finding the optimal number of clusters using the elbow method
inertia = []
for i in range(1, 11):
    kmeans = KMeans(n_clusters=i, **kmeans_set)
    kmeans.fit(scaled_data)
    inertia.append(kmeans.inertia_)

In [120]:

# Plotting the elbow method graph
plt.style.use('fivethirtyeight')
plt.figure(figsize=(10, 6))
plt.plot(range(1, 11), inertia, marker='o')
plt.title('Elbow Method')
plt.xlabel('Number of clusters')
plt.ylabel('Inertia')
plt.grid(True)
plt.show()

In [121]:

# Let's also evaluate the silhouette score for different number of clusters
silhouette_scores = []
for i in range(2, 11):
    kmeans = KMeans(n_clusters=i, **kmeans_set)
    kmeans.fit(scaled_data)
    score = silhouette_score(scaled_data, kmeans.labels_)
    silhouette_scores.append(score)

In [122]:

# Plotting the silhouette scores
plt.figure(figsize=(10, 6))
plt.plot(range(2, 11), silhouette_scores, marker='o')
plt.title('Silhouette Scores for Different Number of Clusters')
plt.xlabel('Number of clusters')
plt.ylabel('Silhouette Score')
plt.grid(True)
plt.show()

In [123]:

pip install kneed

Defaulting to user installation because normal site-packages is not writeable
Looking in links: /usr/share/pip-wheels
Requirement already satisfied: kneed in ./.local/lib/python3.11/site-packages (0.8.5)
Requirement already satisfied: numpy>=1.14.2 in /opt/conda/envs/anaconda-panel-2023.05-py310/lib/python3.11/site-packages (from kneed) (1.24.3)
Requirement already satisfied: scipy>=1.0.0 in /opt/conda/envs/anaconda-panel-2023.05-py310/lib/python3.11/site-packages (from kneed) (1.11.1)
Note: you may need to restart the kernel to use updated packages.

In [124]:

from kneed import KneeLocator
kl = KneeLocator(range(1, 11), inertia, curve ='convex', direction ='decreasing')
kl.elbow

Out[124]:

In [125]:

# Plotting the elbow method graph
plt.style.use('fivethirtyeight')
plt.figure(figsize=(10, 6))
plt.plot(range(1, 11), inertia, marker='o')
plt.title('Elbow Method')
plt.xlabel('Number of clusters')
plt.ylabel('Inertia')
plt.axvline(x = kl.elbow, color = 'b', label = 'axvline - full height', ls ='--')
plt.grid(True)
plt.show()

In [179]:

kmeans =KMeans(n_clusters =4).fit(data_cleaned)
centroids = kmeans.cluster_centers_
print(centroids)

plt.scatter(data_cleaned['MINIMUM_PAYMENTS'], data_cleaned['PAYMENTS'], c= kmeans.labels_.astype(float), s=50, alpha = 0.5)
plt.scatter(centroids[:, 0], centroids[:, 1], c = 'red', s=50)
plt.show()

[[5.98784744e+03 9.58217937e-01 2.65268750e+03 1.61459108e+03
  1.03823369e+03 4.76390092e+03 5.24534631e-01 3.10867975e-01
  4.10617546e-01 3.35048065e-01 1.09151235e+01 3.36203704e+01
  1.11496984e+04 6.79839497e+03 3.31890089e+03 1.04418332e-01
  1.17453704e+01]
 [8.38945378e+02 8.75314373e-01 5.38521666e+02 2.65495091e+02
  2.73325861e+02 4.97813132e+02 4.59926188e-01 1.37401617e-01
  3.52241430e-01 1.11756603e-01 2.34466287e+00 9.96490906e+00
  2.28446394e+03 9.58590647e+02 6.04084650e+02 1.54811013e-01
  1.14131913e+01]
 [2.08245987e+03 9.20474871e-01 1.40796958e+03 8.75310686e+02
  5.33042982e+02 1.09122637e+03 5.62395976e-01 3.20654943e-01
  3.90365676e-01 1.43416961e-01 3.45955591e+00 2.01153846e+01
  7.53784154e+03 2.02917440e+03 7.72034605e+02 1.79689826e-01
  1.17359239e+01]
 [5.40533094e+03 9.57954542e-01 2.72763637e+04 2.18771029e+04
  5.39926083e+03 1.55037839e+03 9.09027750e-01 8.42361167e-01
  7.20833292e-01 5.90277500e-02 2.50000000e+00 1.27958333e+02
  1.60833333e+04 2.79256345e+04 3.26667104e+03 5.11205792e-01
  1.19166667e+01]]

In [256]:

data_cleaned

Out[256]:

	BALANCE	BALANCE_FREQUENCY	PURCHASES	ONEOFF_PURCHASES	INSTALLMENTS_PURCHASES	CASH_ADVANCE	PURCHASES_FREQUENCY	ONEOFF_PURCHASES_FREQUENCY	PURCHASES_INSTALLMENTS_FREQUENCY	CASH_ADVANCE_FREQUENCY	CASH_ADVANCE_TRX	PURCHASES_TRX	CREDIT_LIMIT	PAYMENTS	MINIMUM_PAYMENTS	PRC_FULL_PAYMENT	TENURE
0	40.900749	0.818182	95.40	0.00	95.40	0.000000	0.166667	0.000000	0.083333	0.000000	0	2	1000.0	201.802084	139.509787	0.000000	12
1	3202.467416	0.909091	0.00	0.00	0.00	6442.945483	0.000000	0.000000	0.000000	0.250000	4	0	7000.0	4103.032597	1072.340217	0.222222	12
2	2495.148862	1.000000	773.17	773.17	0.00	0.000000	1.000000	1.000000	0.000000	0.000000	0	12	7500.0	622.066742	627.284787	0.000000	12
4	817.714335	1.000000	16.00	16.00	0.00	0.000000	0.083333	0.083333	0.000000	0.000000	0	1	1200.0	678.334763	244.791237	0.000000	12
5	1809.828751	1.000000	1333.28	0.00	1333.28	0.000000	0.666667	0.000000	0.583333	0.000000	0	8	1800.0	1400.057770	2407.246035	0.000000	12
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
8943	5.871712	0.500000	20.90	20.90	0.00	0.000000	0.166667	0.166667	0.000000	0.000000	0	1	500.0	58.644883	43.473717	0.000000	6
8945	28.493517	1.000000	291.12	0.00	291.12	0.000000	1.000000	0.000000	0.833333	0.000000	0	6	1000.0	325.594462	48.886365	0.500000	6
8947	23.398673	0.833333	144.40	0.00	144.40	0.000000	0.833333	0.000000	0.666667	0.000000	0	5	1000.0	81.270775	82.418369	0.250000	6
8948	13.457564	0.833333	0.00	0.00	0.00	36.558778	0.000000	0.000000	0.000000	0.166667	2	0	500.0	52.549959	55.755628	0.250000	6
8949	372.708075	0.666667	1093.25	1093.25	0.00	127.040008	0.666667	0.666667	0.000000	0.333333	2	23	1200.0	63.165404	88.288956	0.000000	6

8637 rows × 17 columns

In [232]:

conda install tkinter

Channels:
 - defaults
Platform: linux-64
Collecting package metadata (repodata.json): done
Solving environment: failed

PackagesNotFoundError: The following packages are not available from current channels:

  - tkinter

Current channels:

  - defaults

To search for alternate channels that may provide the conda package you're
looking for, navigate to

    https://anaconda.org

and use the search bar at the top of the page.



Note: you may need to restart the kernel to use updated packages.

In [ ]:

import tkinter as tk
from tkinter import filedialog
from matplotlib.backends.backend_tkagg import FigureCanvasTkAgg

kmeans = KMeans(n_clusters = 4).fit(data_cleaned)
centroids = kmeans.cluster_centers_

root = tk.Tk()

canvas1 = tk.Canvas(root, width = 100, height=100)
canvas1.pack()

label1 = tk.Label(root, text = centroids, justify='center')

canvas1.create_window(70, 50, window = label1)


figure1 = plt.Figure(figsize= (5, 8), dpi = 100)

ax1 = figure1.add_subplot(111)
ax1.scatter(data_cleaned['MINIMUM_PAYMENTS'], data_cleaned['PAYMENTS'], c = kmeans.labels_.astype(float), s=50, alpha = 0.5)
ax1.scatter(centroids[:, 0], centroids[:, 1], c='red', s = 50)
scatter1 = FigureCanvasTkAgg(figure1, root)
scatter1.get_tk_widget().pack(side=tk.LEFT, fill = tk.BOTH)

root.mainloop()

In [ ]:

root = tk.Tk()

canvas1 = tk.Canvas(root, width = 400, height=300, relief = 'raised')
canvas1.pack()

label1 = tk.Label(root, text = 'K-Means Clustering')
label1.config(font=('helvetica', 14))
canvas1.create_window(200, 25, window = label1)


label2=tk.Label(root, text ='Type Number of Clusters:')
label2.config(font = ('helvetica', 8))
canvas1.create_window(200, 120, window = label2)

entry1=tk.Entry(root)
canvas1.create_window(200, 140, window = entry1)

In [ ]:

def getKMeans():
    global data_cleaned
    global number_of_clusters
    number_of_clusters = int(entry.get())
    
   
    
    label3 = tk.Label(root, text = centroids)
    canvas1.create_window(200, 200, window =label3)
    
    figure1 = plt.Figure(figsize=(4, 3), dpi = 100)
    ax1 = figure1.add_subplot(111)
    ax1.scatter(data_cleaned['MINIMUM_PAYMENTS'], data_cleaned['PAYMENTS'], c = kmeans.labels.astype(float), s=50, alpha = 0.5)
    ax1.scatter(centroids[:, 0], centroids[:, 1], c='red', s = 50)
    scatter1.get_tk_widget().pack(side=tk.RIGHT, fill = tk.BOTH)
processButton = tk.Button(text = ' Process k-Means ',command = getKMeans, bg='brown', fg='white' , font = ('helvetica', 10, 'bold'))
canvas1.create_window(200, 170, window=processButton)
canvas1.create_window(200, 170, window = processButton)

root.mainloop()

In [287]:

from sklearn.metrics import silhouette_score, calinski_harabasz_score, davies_bouldin_score

def evaluate_clustering(labels, data):
    silhouette = silhouette_score(data, labels)
    calinski = calinski_harabasz_score(data, labels)
    davies = davies_bouldin_score(data, labels)
    return silhouette, calinski, davies

In [ ]:

kmeans_scores = evaluate_clustering(kmeans.labels_, scaled_data)
kmeans_scores

Affinity Propagation¶

In [324]:

from sklearn.cluster import AffinityPropagation

affinity_propagation = AffinityPropagation(random_state=42)
affinity_labels = affinity_propagation.fit_predict(data_cleaned)

In [325]:

evaluate_clustering(affinity_propagation.labels_, data_cleaned)

Out[325]:

(0.1588224370367602, 545.3048259421168, 1.0123695722157848)

Mean Shift¶

In [323]:

from sklearn.cluster import MeanShift, estimate_bandwidth

# The following bandwidth can be automatically detected using
bandwidth = estimate_bandwidth(data_cleaned, quantile=0.2, n_samples=100)

ms = MeanShift(bandwidth=bandwidth, bin_seeding=True)
ms.fit(data_cleaned)
labels = ms.labels_
cluster_centers = ms.cluster_centers_

labels_unique = np.unique(labels)
n_clusters_ = len(labels_unique)

print("number of estimated clusters : %d" % n_clusters_)

number of estimated clusters : 154

In [ ]:

import matplotlib.pyplot as plt

plt.figure(1)
plt.clf()

colors = ["#dede00", "#377eb8", "#f781bf"]
markers = ["x", "o", "^"]
labels = make_shift.labels_

for k, col in zip(range(n_clusters_), colors):
    my_members = labels == k
    cluster_center = cluster_centers[k]
    plt.plot(data_cleaned[my_members, 0], data_cleaned[my_members, 1], markers[k], color=col)
    plt.plot(
        cluster_center[0],
        cluster_center[1],
        markers[k],
        markerfacecolor=col,
        markeredgecolor="k",
        markersize=14,
    )
plt.title("Estimated number of clusters: %d" % n_clusters_)
plt.show()

Spectral Clustering¶

In [ ]:

from sklearn.cluster import SpectralClustering

spectral_clustering = SpectralClustering(n_clusters=4, random_state=42)
spectral_labels = spectral_clustering.fit_predict(scaled_data)

In [ ]:

evaluate_clustering(spectral_labels, spectral_clustering)

Hierarchical Clustering¶

In [ ]:

from sklearn.cluster import AgglomerativeClustering

hierarchical = AgglomerativeClustering(n_clusters=4)
hierarchical_labels = hierarchical.fit_predict(scaled_data)
evaluate_clustering(hierarchical_labels, hierarchical)

DBSCAN¶

Determining Optimal Parameters for DBSCAN¶

In [267]:

from sklearn.neighbors import NearestNeighbors
import numpy as np
scaled_features = scaled_data
# We'll use the scaled features 'PAYMENTS' and 'MINIMUM_PAYMENTS'
neighbors = NearestNeighbors(n_neighbors=2)
neighbors_fit = neighbors.fit(scaled_features)
distances, indices = neighbors_fit.kneighbors(scaled_features)

# Sort and plot distances
distances = np.sort(distances, axis=0)
distances = distances[:, 1]
plt.plot(distances)
plt.title("Distance to Nearest Neighbor")
plt.xlabel("Data points sorted by distance")
plt.ylabel("Epsilon")
plt.show()

look for the point where the increase in distance begins to slow down (known as the "elbow"), as it can be a good estimate for eps.

Clustering with DBSCAN

In [311]:

from sklearn.cluster import DBSCAN

# Use the estimated eps value and a suitable min_samples
dbscan = DBSCAN(eps=0.5, min_samples=2000) 
dbscan_labels = dbscan.fit_predict(scaled_features)

Evaluating the Results¶

DBSCAN doesn't have a set number of clusters like K-means. Instead, it forms clusters based on the density, and points that don't belong to any cluster are marked as noise. You can evaluate the results by:

Counting the number of clusters and noise points.

Checking the spread and size of the clusters.

Using silhouette score, although it might not be as informative for DBSCAN as for other algorithms.

In [321]:

evaluate_clustering(dbscan.labels_, data_cleaned)

Out[321]:

(0.4868922360554868, 1768.8963154477767, 1.7107184815587702)

In [315]:

dbscan_labels

Out[315]:

array([ 0, -1,  0, ...,  0,  0,  0])

Visualization¶

In [275]:

# Plotting
plt.figure(figsize=(10, 6))
unique_labels = set(dbscan_labels)
colors = [plt.cm.Spectral(each) for each in np.linspace(0, 1, len(unique_labels))]

for k, col in zip(unique_labels, colors):
    if k == -1:
        # Black used for noise
        col = [0, 0, 0, 1]

    class_member_mask = (dbscan_labels == k)

    xy = scaled_features[class_member_mask]
    plt.plot(xy[:, 0], xy[:, 1], 'o', markerfacecolor=tuple(col), markeredgecolor='k', markersize=6)

plt.title("DBSCAN Clustering")
plt.xlabel("PAYMENTS")
plt.ylabel("MINIMUM_PAYMENTS")
plt.show()

HDBSCAN¶

In [ ]:

import hdbscan

hdbscan_cluster = hdbscan.HDBSCAN(min_cluster_size=5)
hdbscan_labels = hdbscan_cluster.fit_predict(scaled_data)
evaluate_clustering(hdbscan_labels, hdbscan_cluster)

OPTICS¶

In [ ]:

from sklearn.cluster import OPTICS

optics = OPTICS(min_samples=5)
optics_labels = optics.fit_predict(scaled_data)

In [ ]:

evaluate_clustering(optics_labels, optics)

BIRCH¶

In [ ]:

from sklearn.cluster import BIRCH

birch = BIRCH(n_clusters=4)
birch_labels = birch.fit_predict(scaled_data)

In [ ]:

evaluate_clustering(birch_labels, birch)

Evaluation¶

In [277]:

# Extracting and scaling the specific features
features = data_cleaned[['PAYMENTS', 'MINIMUM_PAYMENTS']]
scaler = StandardScaler()
scaled_features = scaler.fit_transform(features)

# Apply K-means
kmeans = KMeans(n_clusters=3)
kmeans_labels = kmeans.fit_predict(scaled_features)
centroids = kmeans.cluster_centers_

# Plotting
plt.figure(figsize=(10, 6))
plt.scatter(scaled_features[:, 0], scaled_features[:, 1], c=kmeans_labels, cmap='viridis', marker='o', edgecolor='k', s=50, alpha=0.7)
plt.scatter(centroids[:, 0], centroids[:, 1], c='red', marker='x', s=200)

# Decision boundaries (optional, for visualization)
x_min, x_max = scaled_features[:, 0].min() - 1, scaled_features[:, 0].max() + 1
y_min, y_max = scaled_features[:, 1].min() - 1, scaled_features[:, 1].max() + 1
xx, yy = np.meshgrid(np.arange(x_min, x_max, 0.01), np.arange(y_min, y_max, 0.01))
Z = kmeans.predict(np.c_[xx.ravel(), yy.ravel()])
Z = Z.reshape(xx.shape)

plt.contourf(xx, yy, Z, alpha=0.2, cmap='viridis')
plt.xlim(xx.min(), xx.max())
plt.ylim(yy.min(), yy.max())
plt.title("K-means Clustering on 'PAYMENTS' and 'MINIMUM_PAYMENTS'")
plt.xlabel('PAYMENTS (scaled)')
plt.ylabel('MINIMUM_PAYMENTS (scaled)')
plt.show()

In [ ]:

How to use and evaluate clustering algorithms?

Farhad Piri¶

Clustering algorithms in action¶

need the source data to try this for yourself? email me at farhad.peeri[at]gmail[dot]com¶

Data Cleaning and Preprocessing:¶

Let`s Normalize the data¶

Model¶

Affinity Propagation¶

Mean Shift¶

Spectral Clustering¶

Hierarchical Clustering¶

DBSCAN¶

Determining Optimal Parameters for DBSCAN¶

Evaluating the Results¶

Visualization¶

HDBSCAN¶

OPTICS¶

BIRCH¶

Evaluation¶

Leave a ReplyCancel Reply