-
Notifications
You must be signed in to change notification settings - Fork 90
/
Copy pathcluster_centers.py
121 lines (107 loc) · 5.64 KB
/
cluster_centers.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
# Author：MaXiao
# E-mail：maxiaoscut@aliyun.com
import time
import itertools
import numpy as np
from sklearn.cluster import KMeans, SpectralClustering, DBSCAN, Birch
from sklearn.metrics import calinski_harabasz_score
def timer(func):
def wrapper(*args, **kwargs):
start = time.time()
result = func(*args,**kwargs)
end = time.time()
print(func.__name__+' running time：{:.2f}s'.format(end-start))
return result
return wrapper
@timer
def get_cluster_centers(dataset, n_clusters='auto', cluster_algo='kmeans', params_grid='auto', random_state=2018):
if n_clusters == 'auto':
clusters_range = range(2, 8)
else:
assertion = 'n_clusters should be an integer greater than or equal to 2'
assert isinstance(n_clusters, int) and n_clusters>=2, assertion
clusters_range = [n_clusters]
# get_centers函数可根据聚类算法对无标签数据集的预测结果(cluster_pred)返回各聚类簇的中心
def get_centers(cluster_label, data=dataset):
centers = []
for label in np.unique(cluster_label):
subset = data[cluster_label==label]
center = np.mean(subset, axis=0)
centers.append(center)
return np.sort(centers, axis=0)
# 通过生成参数的笛卡尔积，寻求谱聚类算法的最优参数
if cluster_algo == 'spectral':
if params_grid == 'auto':
params_grid = {'n_clusters':clusters_range, 'gamma':np.linspace(0.5, 1.5, 3)}
params, score, y_pred_set = [], [], []
for i, j in itertools.product(params_grid['n_clusters'], params_grid['gamma']):
params.append((i, j))
spectral = SpectralClustering(n_clusters=i, gamma=j, n_jobs=-1, random_state=random_state)
y_pred_spectral = spectral.fit_predict(dataset)
y_pred_set.append(y_pred_spectral)
ch_score = calinski_harabasz_score(dataset, y_pred_spectral)
score.append(ch_score)
# 获取calinski_harabasz_score取最大值时对应的参数与预测聚类类标
best_param = params[np.argmax(score)]
y_pred = y_pred_set[np.argmax(score)]
return get_centers(y_pred), np.max(score)
# 通过生成参数的笛卡尔积，寻求Birch聚类算法的最优参数
elif cluster_algo == 'birch':
if params_grid == 'auto':
params_grid = {'n_clusters':clusters_range, 'branching_factor':range(2, 10),
'threshold':np.linspace(0, 0.8, num=10)}
params, score, y_pred_set = [], [], []
for i, j, k in itertools.product(params_grid['n_clusters'], params_grid['branching_factor'], params_grid['threshold']):
params.append((i, j, k))
birch = Birch(n_clusters=i, branching_factor=j, threshold=k)
y_pred_birch = birch.fit_predict(dataset)
y_pred_set.append(y_pred_birch)
ch_score = calinski_harabasz_score(dataset, y_pred_birch)
score.append(ch_score)
best_param = params[np.argmax(score)]
y_pred = y_pred_set[np.argmax(score)]
return get_centers(y_pred), np.max(score)
# 通过生成参数的笛卡尔积，寻求DBSCAN聚类算法的最优参数
elif cluster_algo == 'dbscan':
if params_grid == 'auto':
params_grid = {'eps':np.linspace(0.1, 10, num=50), 'min_samples':range(1, 10)}
params, unlabeled_set, y_pred_set, score = [], [], [], []
for i, j in itertools.product(params_grid['eps'], params_grid['min_samples']):
dbscan = DBSCAN(eps=i, min_samples=j, n_jobs=-1)
y_pred_dbscan = dbscan.fit_predict(dataset)
# DBSCAN视预测结果为-1的样本为噪声，因此需要将“噪音样本”予以排除
y_pred_new = y_pred_dbscan[y_pred_dbscan != -1]
dataset_new = dataset[y_pred_dbscan != -1]
# 计算剔除“噪音样本”后无标签样本的剩余比例以及聚类簇的数目
ratio = dataset_new.shape[0] / dataset.shape[0]
n_clusters = len(np.unique(y_pred_new))
# 剩余样本的聚类簇数以及剩余比例满足一定要求，才能对参数及预测结果予以保留
if n_clusters in range(2, 8) and ratio>=0.8:
params.append((i, j))
unlabeled_set.append(dataset_new)
y_pred_set.append(y_pred_new)
ch_score = calinski_harabasz_score(dataset_new, y_pred_new)
score.append(ch_score)
if len(score) > 0:
best_param = params[np.argmax(score)]
unlabeled_final = unlabeled_set[np.argmax(score)]
y_pred = y_pred_set[np.argmax(score)]
return get_centers(y_pred, data=unlabeled_final), np.max(score)
else:
descript = 'It is difficult for dbscan to determine the number of clusters in this dataset. \
Please switch to another clustering algorithm.'
print(descript)
# 寻求Kmeans聚类算法的最优参数
else:
if params_grid == 'auto': params_grid = {'n_clusters': clusters_range}
params, score, y_pred_set = [], [], []
for i in params_grid['n_clusters']:
params.append(i)
kmeans = KMeans(n_clusters=i, random_state=2018)
y_pred_kms = kmeans.fit_predict(dataset)
y_pred_set.append(y_pred_kms)
ch_score = calinski_harabasz_score(dataset, y_pred_kms)
score.append(ch_score)
best_param = params[np.argmax(score)]
y_pred = y_pred_set[np.argmax(score)]
return get_centers(y_pred), np.max(score)