def k_medoids_over_instances(self, dataset, cols, k, distance_metric, max_iters, n_inits=5, p=1):
# If we set it to default we use the pyclust package...
temp_dataset = dataset[cols]
if distance_metric == 'default':
km = pyclust.KMedoids(n_clusters=k, n_trials=n_inits)
km.fit(temp_dataset.values)
cluster_assignment = km.labels_
else:
self.p = p
cluster_assignment = []
best_silhouette = -1
# Compute all distances
D = self.compute_distance_matrix_instances(temp_dataset, distance_metric)
for it in range(0, n_inits):
# First select k random points as centers:
centers = random.sample(range(0, len(dataset.index)), k)
prev_centers = []
points_to_cluster = []
n_iter = 0
while (n_iter < max_iters) and not (centers == prev_centers):
n_iter += 1
prev_centers = centers
# Assign points to clusters.
points_to_centroid = D[centers].idxmin(axis=1)
new_centers = []
for i in range(0, k):
# And find the new center that minimized the sum of the differences.
best_center = D.loc[points_to_centroid == centers[i], points_to_centroid == centers[i]].sum().idxmin(axis=1)
new_centers.append(best_center)
centers = new_centers
# Convert centroids to cluster numbers:
points_to_centroid = D[centers].idxmin(axis=1)
current_cluster_assignment = []
for i in range(0, len(dataset.index)):
current_cluster_assignment.append(centers.index(points_to_centroid.iloc[i,:]))
silhouette_avg = silhouette_score(temp_dataset, np.array(current_cluster_assignment))
if silhouette_avg > best_silhouette:
cluster_assignment = current_cluster_assignment
best_silhouette = silhouette_avg
# And add the clusters and silhouette scores to the dataset.
dataset['cluster'] = cluster_assignment
silhouette_avg = silhouette_score(temp_dataset, np.array(cluster_assignment))
silhouette_per_inst = silhouette_samples(temp_dataset, np.array(cluster_assignment))
dataset['silhouette'] = silhouette_per_inst
return dataset
# input features
feature_matrix[obser_count] = [
month, date, hr, weekday, temp, cond_code
] + factility
# input truth
truth_vector[obser_count] = truth
obser_count += 1
return [feature_matrix, truth_vector]
print "reading data..."
train_feature_matrix, train_truth_vector = read_obser_set(
'trainset_2014_with_facility.txt', SIZE)
print "Standardizing..."
train_feature_matrix = StandardScaler().fit_transform(train_feature_matrix)
print "clustering..."
kmd = pyclust.KMedoids(n_clusters=300000, n_trials=50)
kmd.fit(train_feature_matrix)
# ## K-meoids
# # kmeans = KMeans(n_clusters=30000, init='k-means++', n_jobs=4, algorithm='auto').fit(train_feature_matrix)
# medoids, clusters = kMedoids(np.transpose(train_feature_matrix), 30000)
print "printing..."
centers = open('training_subsample.txt', 'a+')
for idx in medoids:
centers.write(
str(train_feature_matrix[idx]).replace("]", "").replace(
"[", "").replace('\'', "").replace(" ", ",") + "," +
str(train_truth_vector[0]) + "\n")
centers.close()
def kMedoids(X,k,d_metric='euclidean'):
km = pyclust.KMedoids(n_clusters=k,distance=d_metric)
return km.fit_predict(X)
def k_medoids(g_distance, number_clusters):
kmedoids = pyclust.KMedoids(n_clusters=number_clusters, n_trials=50, random_state = 0).fit_predict(g_distance)
tsne_model = TSNE(n_components=2, verbose=1, random_state=0)
tsne_kmedoids = tsne_model.fit_transform(g_distance)
return tsne_kmedoids, kmedoids
import numpy as np
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D
from sklearn import datasets
import pyclust
np.random.seed(5)
iris = datasets.load_iris()
X = iris.data
y = iris.target
print(X)
print('size of data:', X.shape)
print(y)
est = pyclust.KMedoids(n_clusters=3, n_trials=100)
est.fit(X)
print('estimated labels:')
print(est.labels_)
#######
fig = plt.figure(figsize=(4, 3))
plt.clf()
ax = Axes3D(fig, rect=[0, 0, .95, 1], elev=48, azim=134)
plt.cla() #clear current axes!!
ydict = {0: 'setosa', 1: 'versicolor', 2: 'Virginica'}
with plt.style.context('seaborn-whitegrid'):
for lab, col in zip((0, 1, 2), ('blue', 'red', 'green')):
ax.scatter(X[est.labels_ == lab, 3],
X[est.labels_ == lab, 0],
X[est.labels_ == lab, 2],
c=col,
def kmedoid(kdata, n_clusters, n_trial):
kmd = pyclust.KMedoids(n_clusters=n_clusters, n_trials=n_trial)
kmd.fit(kdata.values)
return kmd.labels_
def k_medoids_over_instances(self,
dataset: pd.DataFrame,
cols: List[str],
k: int,
distance_metric: str,
max_iters: int,
n_inits: int = 5,
p: int = 1):
"""
Apply k-medoids clustering using the self implemented distance metrics.
:param dataset: DataFrame to apply clustering on.
:param cols: List of columns to use for clustering.
:param k: Number of clusters.
:param distance_metric: Distance metrics to use for clustering.
:param max_iters: Maximum number of iterations.
:param n_inits: Number of inits.
:param p: Optional parameter for Minkowski distance metrics.
:return: Original DataFrame with cluster and silhouette score columns added.
"""
# Select the appropriate columns
temp_dataset = dataset[cols]
# Use PyClust Package in case of default distance metric
if distance_metric == 'default':
km = pyclust.KMedoids(n_clusters=k, n_trials=n_inits)
km.fit(temp_dataset.values)
cluster_assignment = km.labels_
else:
self.p = p
cluster_assignment = []
best_silhouette = -1
# Compute all distances
D = self.compute_distance_matrix_instances(temp_dataset,
distance_metric)
for it in range(0, n_inits):
# Select k random points as centers first
centers = random.sample(range(0, len(dataset.index)), k)
prev_centers = []
n_iter = 0
while (n_iter < max_iters) and not (centers == prev_centers):
n_iter += 1
prev_centers = centers
# Assign points to clusters
points_to_centroid = D[centers].idxmin(axis=1)
new_centers = []
for i in range(0, k):
# Find the new center that minimized the sum of the differences
best_center = D.loc[points_to_centroid ==
centers[i]].sum().idxmin(axis=1)
new_centers.append(best_center)
centers = new_centers
# Convert centroids to cluster numbers:
points_to_centroid = D[centers].idxmin(axis=1)
current_cluster_assignment = []
for i in range(0, len(dataset.index)):
current_cluster_assignment.append(
centers.index(points_to_centroid.iloc[i]))
silhouette_avg = silhouette_score(
temp_dataset, np.array(current_cluster_assignment))
if silhouette_avg > best_silhouette:
cluster_assignment = current_cluster_assignment
best_silhouette = silhouette_avg
# Add the clusters and silhouette scores to the dataset
dataset['cluster'] = cluster_assignment
silhouette_per_inst = silhouette_samples(temp_dataset,
np.array(cluster_assignment))
dataset['silhouette'] = silhouette_per_inst
return dataset
membs,
n_clusters=2,
distance='euclidean')
print(cent_upd)
rng = np.random.RandomState(1234)
print(
pyclust._kmedoids._kmedoids_run(d,
n_clusters=2,
distance='euclidean',
max_iter=20,
tol=0.001,
rng=rng))
kmd = pyclust.KMedoids(n_clusters=2)
kmd.fit(d)
print("Centers: ", kmd.centers_)
print("Labels: ", kmd.labels_)
print("SSE: ", kmd.sse_arr_)
print("N_ITER: ", kmd.n_iter_)
print('\n\n*** Testing RandomState: ***')
for i in range(3):
kmd2 = pyclust.KMedoids(n_clusters=2, random_state=123)
kmd2.fit(d)
print("Centers: ", kmd2.centers_)
print("Labels: ", kmd2.labels_)
print("SSE: ", kmd2.sse_arr_)
