• 机器学习sklearn(63):算法实例(二十)聚类(三)KMeans (二) sklearn.cluster.KMeans


    class sklearn.cluster.KMeans (n_clusters=8, init=’k-means++’, n_init=10, max_iter=300, tol=0.0001,precompute_distances=’auto’, verbose=0, random_state=None, copy_x=True, n_jobs=None, algorithm=’auto’)

    1 重要参数n_clusters

    n_clusters是KMeans中的k,表示着我们告诉模型我们要分几类。这是KMeans当中唯一一个必填的参数,默认为8类,但通常我们的聚类结果会是一个小于8的结果。通常,在开始聚类之前,我们并不知道n_clusters究竟是多少,因此我们要对它进行探索。
    1.1 先进行一次聚类看看吧
    当我们拿到一个数据集,如果可能的话,我们希望能够通过绘图先观察一下这个数据集的数据分布,以此来为我们聚类时输入的n_clusters做一个参考。
    首先,我们来自己创建一个数据集。这样的数据集是我们自己创建,所以是有标签的。
    from sklearn.datasets import make_blobs
    import matplotlib.pyplot as plt
    #自己创建数据集
    X, y = make_blobs(n_samples=500,n_features=2,centers=4,random_state=1)
    fig, ax1 = plt.subplots(1)
    ax1.scatter(X[:, 0], X[:, 1]
               ,marker='o' #点的形状
               ,s=8 #点的大小
               )
    plt.show()
    #如果我们想要看见这个点的分布,怎么办?
    color = ["red","pink","orange","gray"]
    fig, ax1 = plt.subplots(1)
    for i in range(4):
        ax1.scatter(X[y==i, 0], X[y==i, 1]
               ,marker='o' #点的形状
               ,s=8 #点的大小
               ,c=color[i]
               )
    plt.show()
    基于这个分布,我们来使用Kmeans进行聚类。首先,我们要猜测一下,这个数据中有几簇?
    from sklearn.cluster import KMeans
    n_clusters = 3
    cluster = KMeans(n_clusters=n_clusters, random_state=0).fit(X)
    y_pred = cluster.labels_
    y_pred
    pre = cluster.fit_predict(X)
    pre == y_pred
    cluster_smallsub
    = KMeans(n_clusters=n_clusters, random_state=0).fit(X[:200]) y_pred_ = cluster_smallsub.predict(X) y_pred == y_pred_
    centroid
    = cluster.cluster_centers_ centroid
    centroid.shape
    inertia
    = cluster.inertia_ inertia
    color
    = ["red","pink","orange","gray"] fig, ax1 = plt.subplots(1)
    for i in range(n_clusters):    ax1.scatter(X[y_pred==i, 0], X[y_pred==i, 1]           ,marker='o'           ,s=8           ,c=color[i]           ) ax1.scatter(centroid[:,0],centroid[:,1]           ,marker="x"           ,s=15           ,c="black") plt.show()
    n_clusters
    = 4 cluster_ = KMeans(n_clusters=n_clusters, random_state=0).fit(X) inertia_ = cluster_.inertia_ inertia_
    n_clusters
    = 5 cluster_ = KMeans(n_clusters=n_clusters, random_state=0).fit(X) inertia_ = cluster_.inertia_ inertia_
    n_clusters
    = 6 cluster_ = KMeans(n_clusters=n_clusters, random_state=0).fit(X) inertia_ = cluster_.inertia_ inertia_
    1.2 聚类算法的模型评估指标

     

     

     

    1.2.1 当真实标签已知的时候

     

    1.2.2 当真实标签未知的时候:轮廓系数

     

    from sklearn.metrics import silhouette_score
    from sklearn.metrics import silhouette_samples
    X
    y_pred
    silhouette_score(X,y_pred)
    silhouette_score(X,cluster_.labels_)
    silhouette_samples(X,y_pred)

     

     

    from sklearn.metrics import calinski_harabaz_score
    X
    y_pred
    calinski_harabaz_score(X, y_pred)
    虽然calinski-Harabaz指数没有界,在凸型的数据上的聚类也会表现虚高。但是比起轮廓系数,它有一个巨大的优点,就是计算非常快速。之前我们使用过魔法命令%%timeit来计算一个命令的运算时间,今天我们来选择另一种方法:时间戳计算运行时间。 
    from time import time
    t0 = time()
    calinski_harabaz_score(X, y_pred)
    time() - t0
    t0 = time()
    silhouette_score(X,y_pred)
    time() - t0
    import datetime
    datetime.datetime.fromtimestamp(t0).strftime("%Y-%m-%d %H:%M:%S")
    可以看得出,calinski-harabaz指数比轮廓系数的计算块了一倍不止。想想看我们使用的数据量,如果是一个以万计的数据,轮廓系数就会大大拖慢我们模型的运行速度了。 
     
    1.3 案例:基于轮廓系数来选择n_clusters 
    我们通常会绘制轮廓系数分布图和聚类后的数据分布图来选择我们的最佳n_clusters。 
    from sklearn.cluster import KMeans
    from sklearn.metrics import silhouette_samples, silhouette_score
    import matplotlib.pyplot as plt
    import matplotlib.cm as cm
    import numpy as np
    n_clusters = 4
    fig, (ax1, ax2) = plt.subplots(1, 2)
    fig.set_size_inches(18, 7)
    ax1.set_xlim([-0.1, 1])
    ax1.set_ylim([0, X.shape[0] + (n_clusters + 1) * 10])
    clusterer = KMeans(n_clusters=n_clusters, random_state=10).fit(X)
    cluster_labels = clusterer.labels_
    silhouette_avg = silhouette_score(X, cluster_labels)
    print("For n_clusters =", n_clusters,
          "The average silhouette_score is :", silhouette_avg)
    sample_silhouette_values = silhouette_samples(X, cluster_labels)
    y_lower = 10
    for i in range(n_clusters):
        ith_cluster_silhouette_values = sample_silhouette_values[cluster_labels == i]
        ith_cluster_silhouette_values.sort()
        size_cluster_i = ith_cluster_silhouette_values.shape[0]
        y_upper = y_lower + size_cluster_i
        color = cm.nipy_spectral(float(i)/n_clusters)
        ax1.fill_betweenx(np.arange(y_lower, y_upper)
                         ,ith_cluster_silhouette_values
                         ,facecolor=color
                         ,alpha=0.7
                         )
        ax1.text(-0.05
                 , y_lower + 0.5 * size_cluster_i
                 , str(i))
        y_lower = y_upper + 10
    ax1.set_title("The silhouette plot for the various clusters.")
    ax1.set_xlabel("The silhouette coefficient values")
    ax1.set_ylabel("Cluster label")
    ax1.axvline(x=silhouette_avg, color="red", linestyle="--")
    ax1.set_yticks([])
    ax1.set_xticks([-0.1, 0, 0.2, 0.4, 0.6, 0.8, 1])
    colors = cm.nipy_spectral(cluster_labels.astype(float) / n_clusters)
    ax2.scatter(X[:, 0], X[:, 1]
               ,marker='o'
               ,s=8
               ,c=colors
               )
    centers = clusterer.cluster_centers_
    # Draw white circles at cluster centers
    ax2.scatter(centers[:, 0], centers[:, 1], marker='x',
                c="red", alpha=1, s=200)
    ax2.set_title("The visualization of the clustered data.")
    ax2.set_xlabel("Feature space for the 1st feature")
    ax2.set_ylabel("Feature space for the 2nd feature")
    plt.suptitle(("Silhouette analysis for KMeans clustering on sample data "
                  "with n_clusters = %d" % n_clusters),
                 fontsize=14, fontweight='bold')
    plt.show()
    将上述过程包装成一个循环,可以得到:
    from sklearn.cluster import KMeans
    from sklearn.metrics import silhouette_samples, silhouette_score
    import matplotlib.pyplot as plt
    import matplotlib.cm as cm
    import numpy as np
    for n_clusters in [2,3,4,5,6,7]:
        n_clusters = n_clusters
        fig, (ax1, ax2) = plt.subplots(1, 2)
        fig.set_size_inches(18, 7)
        ax1.set_xlim([-0.1, 1])
        ax1.set_ylim([0, X.shape[0] + (n_clusters + 1) * 10])
        clusterer = KMeans(n_clusters=n_clusters, random_state=10).fit(X)
        cluster_labels = clusterer.labels_
        silhouette_avg = silhouette_score(X, cluster_labels)
        print("For n_clusters =", n_clusters,
              "The average silhouette_score is :", silhouette_avg)
        sample_silhouette_values = silhouette_samples(X, cluster_labels)
        y_lower = 10
        for i in range(n_clusters):
            ith_cluster_silhouette_values = sample_silhouette_values[cluster_labels == i]
            ith_cluster_silhouette_values.sort()
            size_cluster_i = ith_cluster_silhouette_values.shape[0]
            y_upper = y_lower + size_cluster_i
            color = cm.nipy_spectral(float(i)/n_clusters)
            ax1.fill_betweenx(np.arange(y_lower, y_upper)
                             ,ith_cluster_silhouette_values
                             ,facecolor=color
                             ,alpha=0.7
                             )
            ax1.text(-0.05
                     , y_lower + 0.5 * size_cluster_i
                     , str(i))
            y_lower = y_upper + 10
        ax1.set_title("The silhouette plot for the various clusters.")
        ax1.set_xlabel("The silhouette coefficient values")
        ax1.set_ylabel("Cluster label")
        ax1.axvline(x=silhouette_avg, color="red", linestyle="--")
        ax1.set_yticks([])
        ax1.set_xticks([-0.1, 0, 0.2, 0.4, 0.6, 0.8, 1])
        colors = cm.nipy_spectral(cluster_labels.astype(float) / n_clusters)
        ax2.scatter(X[:, 0], X[:, 1]
                   ,marker='o'
                   ,s=8
                   ,c=colors
                   )
        centers = clusterer.cluster_centers_
        # Draw white circles at cluster centers
        ax2.scatter(centers[:, 0], centers[:, 1], marker='x',
                    c="red", alpha=1, s=200)
        
        ax2.set_title("The visualization of the clustered data.")
        ax2.set_xlabel("Feature space for the 1st feature")
        ax2.set_ylabel("Feature space for the 2nd feature")
        plt.suptitle(("Silhouette analysis for KMeans clustering on sample data "
                      "with n_clusters = %d" % n_clusters),
                     fontsize=14, fontweight='bold')
        plt.show()

    2 重要参数init & random_state & n_init:初始质心怎么放好?

     

    X
    y plus
    = KMeans(n_clusters = 10).fit(X) plus.n_iter_ random = KMeans(n_clusters = 10,init="random",random_state=420).fit(X) random.n_iter_

    3 重要参数max_iter & tol:让迭代停下来 

    random = KMeans(n_clusters = 10,init="random",max_iter=10,random_state=420).fit(X)
    y_pred_max10 = random.labels_
    silhouette_score(X,y_pred_max10)
    random = KMeans(n_clusters = 10,init="random",max_iter=20,random_state=420).fit(X)
    y_pred_max20 = random.labels_
    silhouette_score(X,y_pred_max20)

    4 重要属性与重要接口 

     

    5 函数cluster.k_means

    sklearn.cluster.k_means (X, n_clusters, sample_weight=None, init=’k-means++’, precompute_distances=’auto’,n_init=10, max_iter=300, verbose=False, tol=0.0001, random_state=None, copy_x=True, n_jobs=None,algorithm=’auto’, return_n_iter=False) 
     
    函数k_means的用法其实和类非常相似,不过函数是输入一系列值,而直接返回结果。一次性地,函数k_means会依次返回质心,每个样本对应的簇的标签,inertia以及最佳迭代次数。 
     
    from sklearn.cluster import k_means
    k_means(X,4,return_n_iter=True)
  • 相关阅读:
    TFS应用层服务器获取F5用户的真实IP地址(高可用性)
    安装TFS(2015)工作组模式代理服务器(Agent)
    Team Foundation Server 15 功能初探
    TFS 2013 生成(构建)历史记录保持策略(Retention Policy)
    TFS代码变更和工作项关联,为系统变更提供完美的跟踪轨迹
    修改TFS客户端的工作区类型
    比较TFS与SVN,你必须知道的10点区别
    数据字典
    查看源码 类图结构图(Eclipse + Idea)
    Mybatis对应的java和数据库的数据类型
  • 原文地址:https://www.cnblogs.com/qiu-hua/p/14946393.html
Copyright © 2020-2023  润新知