聚类一、K-Means1、代码fromsklearnimportdatasetsfromsklearn.clusterimportKMeansimportmatplotlib.pyplotaspltirisdatasets.load_iris()xiris.datayiris.targetkmeansKMeans(n_clusters3,random_state0).fit(x)y_prekmeans.predict(x)#模型聚类结果plt.scatter(x[:,0],x[:,1],cy)#绘制iris原本的类别plt.scatter(x[:,0],x[:,1],cy_pre)#绘制k-means聚类结果plt.show()2、模型评估fromsklearn.metricsimportsilhouette_scorefromsklearn.metricsimporthomogeneity_completeness_v_measure#轮廓系数silhouette_score(x,kmeans.labels_,metriceuclidean)#接近1则说明样本聚类合理#同质性、完整性与调和平均homogeneity_completeness_v_measure(y,y_pre)#分别返回同质性、完整性、调和平均二、层次聚类fromsklearn.clusterimportAgglomerativeClusteringclusing_wardAgglomerativeClustering(n_clusters3).fit(x)clusing_ward.labels_#簇类别标签#单链接聚类cw_ypreAgglomerativeClustering(n_clusters3).fit_predict(x)plt.scatter(x[:,0],x[:,1],ccw_ypre)plt.rcParams[font.sans-serif]SimHeiplt.rcParams[axes.unicode_minus]Falseplt.title(单链接聚类,size17)plt.show()#均链接聚类cw_ypreAgglomerativeClustering(linkageaverage,n_clusters3).fit_predict(x)plt.scatter(x[:,0],x[:,1],ccw_ypre)plt.rcParams[font.sans-serif]SimHeiplt.rcParams[axes.unicode_minus]Falseplt.title(单链接聚类,size17)plt.show()#全链接聚类cw_ypreAgglomerativeClustering(linkagecomplete,n_clusters3).fit_predict(x)plt.scatter(x[:,0],x[:,1],ccw_ypre)plt.rcParams[font.sans-serif]SimHeiplt.rcParams[axes.unicode_minus]Falseplt.title(单链接聚类,size17)plt.show()三、DBSCANfromsklearn.clusterimportDBSCANimportnumpyasnpx1,y2datasets.make_blobs(n_samples1000,n_features2,centers[[1.2,1.2]],cluster_std[[0.1]],random_state9)x2,y1datasets.make_circles(n_samples5000,factor0.6,noise0.05)xnp.concatenate((x1,x2))plt.scatter(x[:,0],x[:,1],markero)plt.show()dbsDBSCAN().fit(x)#生成DBSCAN模型dbs.labels_#DBSCAN模型的簇标签四、GMMfromsklearn.mixtureimportGaussianMixtureirisdatasets.load_iris()xiris.datayiris.targetgmmGaussianMixture(n_components3).fit(x)gmm.weights_#gmm模型的权重gmm.means_#gmm模型的均值gmm_pregmm.predict(x)plt.scatter(x[:,0],x[:,1],cgmm_pre)plt.title(GMM,size17)plt.show()
