1. 程式人生 > >聚類演算法 sklearn k_means (返回一維資料的最優聚類)

聚類演算法 sklearn k_means (返回一維資料的最優聚類)

from sklearn.cluster import KMeans
import numpy
import collections
import pandas
from sklearn import metrics

def k_means(pp1,clus):


    pv=list(pp1)
    if len(set(pv))>clus:
        gf=numpy.array([pv]).T
        estimator = KMeans(n_clusters=clus)#構造聚類器

        estimator.fit(gf)#聚類
        label_pred = estimator.labels_ #獲取聚類標籤
#print(label_pred) aa=collections.Counter(label_pred) print('aa=',aa) v=pandas.Series(aa) gg=list(v) index_max=gg.index(max(gg)) print('index_max=',index_max) centroids = estimator.cluster_centers_ #獲取聚類中心 print('centroids='
,centroids) #inertia = estimator.inertia_ # 獲取聚類準則的總和 center=centroids[index_max][0] return ((center)) else: return (pp1.mean()) def k_means_label(a): def km_index(k): pv=list(a) gf=numpy.array([pv]).T #from sklearn.cluster import KMeans
y_pred = KMeans(n_clusters=k, random_state=9).fit_predict(gf) index=metrics.silhouette_score(gf, y_pred, metric='euclidean') print('index',index) return index cs=list(range(2,6)) df=list(map(km_index,cs)) df1=pandas.Series(df,index=cs) df2=df1.sort_values(ascending=False) df3=list(df2.index)[0] return df3 a=numpy.random.randint(0,1000,10) cc=k_means_label(a) b=k_means(a,cc) print('b=',b)
index 0.804055967401
index 0.805649685362
index 0.65899543985
index 0.517110170591
aa= Counter({0: 5, 1: 3, 2: 2})
index_max= 0
centroids= [[ 160.8]
 [ 610. ]
 [ 824.5]]
b= 160.8