Clustering con Scikit Learn

import pandas as pd
import matplotlib
import matplotlib.pyplot as plt
import numpy as np

from sklearn import metrics
from sklearn.cluster import KMeans

iris_df = pd.read_csv('datasets/iris.csv', 
                      skiprows=1, 
                      names = ['sepal-length',
                               'sepal-width',
                               'petal-length',
                               'petal-width',
                               'class'])

iris_df.head()

iris_df = iris_df.sample(frac=1).reset_index(drop=True)

iris_df.head()

iris_df.shape

iris_df[iris_df.isnull().any(axis=1)]

iris_df.describe()

iris_df['class'].unique()

from sklearn import preprocessing

label_encoding = preprocessing.LabelEncoder()

iris_df['class'] = label_encoding.fit_transform(iris_df['class'].astype(str))

iris_df.head()

iris_df.describe()

fig, ax = plt.subplots(figsize=(12, 8))

plt.scatter(iris_df['sepal-length'], iris_df['sepal-width'], s=250)

plt.xlabel('sepal-length')
plt.ylabel('sepal-width')

plt.show()

fig, ax = plt.subplots(figsize=(12, 8))

plt.scatter(iris_df['petal-width'], iris_df['petal-length'], s=250)

plt.xlabel('petal-width')
plt.ylabel('petal-length')

plt.show()

fig, ax = plt.subplots(figsize=(12, 8))

plt.scatter(iris_df['sepal-length'], iris_df['petal-length'], s=250)

plt.xlabel('sepal-length')
plt.ylabel('petal-length')

plt.show()

iris_2D = iris_df[['sepal-length', 'petal-length']]

iris_2D.sample(5)

iris_2D.shape

iris_2D = np.array(iris_2D)

kmeans_model_2D = KMeans(n_clusters=3, max_iter=1000).fit(iris_2D)

kmeans_model_2D.labels_

centroids_2D = kmeans_model_2D.cluster_centers_

centroids_2D

fig, ax = plt.subplots(figsize=(12, 8))

plt.scatter(centroids_2D[:,0], centroids_2D[:,1], c='r', s=250, marker='s')

for i in range(len(centroids_2D)):
    plt.annotate(i, (centroids_2D[i][0], centroids_2D[i][1]), fontsize=30)

iris_labels = iris_df['class']

print("Homogeneity_score: ", metrics.homogeneity_score(iris_labels, kmeans_model_2D.labels_))

print("Completeness_score: ", metrics.completeness_score(iris_labels, kmeans_model_2D.labels_))

print("v_measure_score: ", metrics.v_measure_score(iris_labels, kmeans_model_2D.labels_))

print("Adjusted_rand_score: ", metrics.adjusted_rand_score(iris_labels, kmeans_model_2D.labels_))

print("Adjusted_mutual_info_score: ", metrics.adjusted_mutual_info_score(iris_labels,  kmeans_model_2D.labels_))

print("Silhouette_score: ", metrics.silhouette_score(iris_2D, kmeans_model_2D.labels_))

colors = ['yellow','blue','green']

plt.figure(figsize=(12, 8))

plt.scatter(iris_df['sepal-length'], iris_df['petal-length'], c=iris_df['class'], s=200,
            cmap=matplotlib.colors.ListedColormap(colors), alpha=0.5)

plt.scatter(centroids_2D[:,0], centroids_2D[:,1], c='r', s=250, marker='s')

for i in range(len(centroids_2D)):
    plt.annotate( i, (centroids_2D[i][0], centroids_2D[i][1]), fontsize=30)

iris_features = iris_df.drop('class', axis=1)

iris_features.head()

iris_labels = iris_df['class']

iris_labels.sample(5)

kmeans_model = KMeans(n_clusters=3).fit(iris_features)

kmeans_model.labels_

kmeans_model.cluster_centers_

print("Homogeneity_score: ", metrics.homogeneity_score(iris_labels, kmeans_model.labels_))
print("Completeness_score: ", metrics.completeness_score(iris_labels, kmeans_model.labels_))
print("v_measure_score: ", metrics.v_measure_score(iris_labels, kmeans_model.labels_))
print("Adjusted_rand_score: ", metrics.adjusted_rand_score(iris_labels, kmeans_model.labels_))
print("Adjusted_mutual_info_score: ", metrics.adjusted_mutual_info_score(iris_labels,  kmeans_model.labels_))
print("Silhouette_score: ", metrics.silhouette_score(iris_features, kmeans_model.labels_))

from sklearn import metrics

from sklearn.cluster import KMeans
from sklearn.cluster import AgglomerativeClustering
from sklearn.cluster import DBSCAN
from sklearn.cluster import MeanShift
from sklearn.cluster import Birch
from sklearn.cluster import AffinityPropagation
from sklearn.cluster import MiniBatchKMeans

import warnings
warnings.filterwarnings("ignore")

def build_model(clustering_model, data, labels):
    
    model = clustering_model(data)

    print('homo\tcompl\tv-meas\tARI\tAMI\tsilhouette')
    print(50 * '-')
    
    print('%.3f\t%.3f\t%.3f\t%.3f\t%.3f\t%.3f'
          %(metrics.homogeneity_score(labels, model.labels_),
            metrics.completeness_score(labels, model.labels_),
            metrics.v_measure_score(labels, model.labels_),
            metrics.adjusted_rand_score(labels, model.labels_),
            metrics.adjusted_mutual_info_score(labels,  model.labels_),
            metrics.silhouette_score(data, model.labels_)))

def k_means(data, n_clusters=3, max_iter=1000):
    model = KMeans(n_clusters=n_clusters, max_iter=max_iter).fit(data)
    
    return model

build_model(k_means, iris_features, iris_labels)

def agglomerative_fn(data, n_clusters=3):
    model = AgglomerativeClustering(n_clusters = n_clusters).fit(data)
    
    return model

build_model(agglomerative_fn, iris_features, iris_labels)

def dbscan_fn(data, eps=0.45, min_samples=4):
    model = DBSCAN(eps=eps, min_samples=min_samples).fit(data)
    return model

build_model(dbscan_fn, iris_features, iris_labels)

def mean_shift_fn(data, bandwidth=0.85):
    model = MeanShift(bandwidth=bandwidth).fit(data)
    return model

build_model(mean_shift_fn, iris_features, iris_labels)

def birch_fn(data, n_clusters=3):
    model = Birch(n_clusters=n_clusters).fit(data)
    return model

build_model(birch_fn, iris_features, iris_labels)

def affinity_propagation_fn(data, damping=0.6, max_iter=1000):
    model = AffinityPropagation(damping=damping, max_iter=max_iter).fit(data)
    return model

build_model(affinity_propagation_fn, iris_features, iris_labels)

def mini_batch_kmeans_fn(data, n_clusters=3, max_iter=1000):
    model = MiniBatchKMeans(n_clusters=n_clusters, max_iter=max_iter, batch_size=20).fit(data)
    return model

build_model(mini_batch_kmeans_fn, iris_features, iris_labels)

from sklearn import metrics
from sklearn.metrics import silhouette_score

from sklearn.model_selection import ParameterGrid

from sklearn.cluster import KMeans
from sklearn.cluster import DBSCAN
from sklearn.cluster import MeanShift

drivers_df = pd.read_csv('datasets/driver_details.csv')

drivers_df.head()

drivers_df.shape

drivers_df[drivers_df.isnull().any(axis=1)]

drivers_df.describe()

drivers_features = drivers_df.drop('Driver_ID', axis=1)

parameters = {'n_clusters': [2, 3, 4, 5, 10, 20, 30]}

parameter_grid = ParameterGrid(parameters)

list(parameter_grid)

best_score = -1
model = KMeans()

for g in parameter_grid:
    model.set_params(**g)
    model.fit(drivers_features)

    ss = metrics.silhouette_score(drivers_features, model.labels_)
    print('Parametro: ', g, 'Score: ', ss)
    if ss > best_score:
        best_score = ss
        best_grid = g

best_grid

fig, ax = plt.subplots(figsize=(12, 8))

plt.scatter(drivers_features['Distance_Feature'], 
            drivers_features['Speeding_Feature'], s=250)

plt.xlabel('Distance_Feature')
plt.ylabel('Speeding_Feature')

plt.show()

parameters = {'eps': [0.9, 1.0, 5.0, 10.0, 12.0, 14.0, 20.0],
              'min_samples': [5, 7, 10, 12]}

parameter_grid = ParameterGrid(parameters)

list(parameter_grid)

model = DBSCAN()
best_score = -1

for g in parameter_grid:
    model.set_params(**g)
    model.fit(drivers_features)

    ss = metrics.silhouette_score(drivers_features, model.labels_)
    print('Parametro: ', g, 'Score: ', ss)
    if ss > best_score:
        best_score = ss
        best_grid = g

best_grid

model.set_params(**best_grid)
model.fit(drivers_features)

len(model.labels_)

n_clusters = len(set(model.labels_)) - (1 if -1 in model.labels_ else 0)

n_clusters

n_noise = list(model.labels_).count(-1)

n_noise

another_grid = {'eps': 5.0, 'min_samples': 5}

model.set_params(**another_grid)
model.fit(drivers_features)

n_clusters = len(set(model.labels_)) - (1 if -1 in model.labels_ else 0)
print('Numero de clusters: ', n_clusters)

n_noise = list(model.labels_).count(-1)
print('Puntos ruido: ', n_noise)

another_grid = {'eps': 5.0, 'min_samples': 7}

model.set_params(**another_grid)
model.fit(drivers_features)

n_clusters = len(set(model.labels_)) - (1 if -1 in model.labels_ else 0)
print('Numero de Clusters: ', n_clusters)

n_noise = list(model.labels_).count(-1)
print('Puntos Ruido: ', n_noise)

from sklearn.cluster import estimate_bandwidth
estimate_bandwidth(drivers_features)

model = MeanShift(bandwidth=estimate_bandwidth(drivers_features)).fit(drivers_features)

metrics.silhouette_score(drivers_features, model.labels_)