Kmeans
1.类别是人为给定的
如何确定最佳的类别数目，可以通过基于簇内误差平方和，使用肘方法确定簇的最佳数量，肘方法的基本理念就是找出聚类偏差骤增是的k值，通过画出不同k值对应的聚类偏差图。
DBSCAN
基于密度聚类。
密度：样本的紧密程度。使用半径和最小样本量进行评估，如果在指定的半径领域内，实际样本量超过给定的最小期望样本量。
K-means和DBSCAN对比：
优点：
DBSCAN不需要事先要形成的簇类的数量；
DBSCAN可以发现任意形状的簇类；
DBSCAN能够识别出噪声点；
DBSCAN对数据库中样本的顺序不敏感，但对簇类之间的边界样本有所摆动；
缺点：
DBSCAN不能很好反映高维数据；
如果样本集的密度不均匀、聚类间距差很大时，聚类质量较差。
评价指标
（1）sihouette_score 轮廓系数
[TencentCloudSDKException] code:FailedOperation.ServiceIsolate message:service is stopped due to arrears, please recharge your account in Tencent Cloud requestId:579b0557-0bb8-488d-9153-f45a823c2972
[En]
[TencentCloudSDKException] code:FailedOperation.ServiceIsolate message:service is stopped due to arrears, please recharge your account in Tencent Cloud requestId:14d7cbce-b512-46f9-98ce-94db8976d392
silhouette_score 轮廓系数：结合了凝聚度和分离度
[TencentCloudSDKException] code:FailedOperation.ServiceIsolate message:service is stopped due to arrears, please recharge your account in Tencent Cloud requestId:7b84c664-38cf-4349-a40e-b9f531261e0d
[En]
[TencentCloudSDKException] code:FailedOperation.ServiceIsolate message:service is stopped due to arrears, please recharge your account in Tencent Cloud requestId:bddcb525-b756-4d40-a589-292df38d63fd
[TencentCloudSDKException] code:FailedOperation.ServiceIsolate message:service is stopped due to arrears, please recharge your account in Tencent Cloud requestId:c099668a-191a-4a46-af70-7ed187b62849
[En]
[TencentCloudSDKException] code:FailedOperation.ServiceIsolate message:service is stopped due to arrears, please recharge your account in Tencent Cloud requestId:5215a599-ba61-4bb2-b279-7be17a83c168
取值为[-1,1]，值越大越好；
[TencentCloudSDKException] code:FailedOperation.ServiceIsolate message:service is stopped due to arrears, please recharge your account in Tencent Cloud requestId:47c2c4bd-0ae8-4610-ba06-0b73f8fdca67
[En]
[TencentCloudSDKException] code:FailedOperation.ServiceIsolate message:service is stopped due to arrears, please recharge your account in Tencent Cloud requestId:3f670ffd-d5f4-4e46-90b1-9dda967e9365
当值接近0时，表明聚类结果有重叠的情况。
（2）inertias
K-Means模型对象的属性，作为没有真实分类结果标签下的非监督式评估指标。表示样本到最近聚类中心的距离总和。
[TencentCloudSDKException] code:FailedOperation.ServiceIsolate message:service is stopped due to arrears, please recharge your account in Tencent Cloud requestId:847e6c0c-95de-48e1-9c13-7902e3ccaa39
[En]
[TencentCloudSDKException] code:FailedOperation.ServiceIsolate message:service is stopped due to arrears, please recharge your account in Tencent Cloud requestId:ea728289-f2d1-46d0-a4e9-3099973c0d1e
显而易见的，类的数量越大，inertias会趋向越小。
（3）兰德指数
兰德指数需要给定 实际类别信息C，假设K是聚类结果，a表示在C和K中都是同类别的元素对数，
b表示在C和K中都是不同类别的元素对数，计算公式略。
取值范围为[0,1]，值越大意味着聚类结果与真实情况越吻合。
（4）互信息 MI
[TencentCloudSDKException] code:FailedOperation.ServiceIsolate message:service is stopped due to arrears, please recharge your account in Tencent Cloud requestId:7e244f4a-4ae7-4121-bd39-c53ddc9f8786
[En]
[TencentCloudSDKException] code:FailedOperation.ServiceIsolate message:service is stopped due to arrears, please recharge your account in Tencent Cloud requestId:5912e7f8-c9ac-4ad4-a0f5-3b415e8aba9f
（5）同质化得分、完整性得分、v_meansure_score
import numpy as np
import matplotlib.pyplot as plt
from sklearn import metrics
from sklearn import datasets
%matplotlib inline
X1, y1 = datasets.make_moons(n_samples=1000,
noise=0.1,
random_state=16)
X2, y2 = datasets.make_blobs(n_samples=1000,
n_features=2,
centers=[[1.2,1.2]],
cluster_std=[[0.1]],
random_state=16)
X = np.concatenate((X1, X2))
plt.figure(figsize=(10,7))
plt.title('origin')
plt.plot(X[:,0], X[:,1], 'o', markersize=6)
plt.show()
from sklearn.cluster import KMeans, DBSCAN
使用KMeans
y_pred = KMeans(n_clusters=3, random_state=9).fit_predict(X)
plt.figure(figsize=(10,7))
plt.scatter(X[:,0], X[:,1], s=25, c=y_pred)
plt.title('k-means:k=3')
plt.show()
print(metrics.silhouette_score(X, y_pred))
使用DBSCAN
y_pred = DBSCAN(eps=0.15, min_samples=10).fit_predict(X) # eps表示半径，min_samples表示簇最小样本数
plt.figure(figsize=(10, 7))
plt.scatter(X[:,0], X[:,1], s=25, c=y_pred)
plt.title('DBSCAN')
plt.show()
print(metrics.silhouette_score(X, y_pred))
Original: https://blog.csdn.net/jinselizhi/article/details/114397507
Author: 谁怕平生太急
Title: Kmeans和DBSCAN
原创文章受到原创版权保护。转载请注明出处：https://www.johngo689.com/561627/
转载文章受原作者版权保护。转载请注明原作者出处！