基于Sklearn原生Kmeans算法的自动化聚类

from sklearn.cluster import KMeans
import numpy as np
import matplotlib.pyplot as plt
from pandas import Series
import random
from sklearn.metrics import silhouette_samples, silhouette_score
import time

def pressure_cluster(n_clusters,pressure_array):
    '''
    n_clusters是要聚类的类数
    pressure_array是要进行聚类的压力数据数组
    '''
    data_set = np.array(pressure_array).reshape(-1,1)
    clf1 = KMeans(n_clusters=n_clusters) #调用KMeans算法
    y_pred = clf1.fit_predict(data_set) #加载数据集合
    silhouette_avg = silhouette_score(data_set,clf1.labels_ )
    sample_silhouette_values = silhouette_samples(data_set,clf1.labels_)
    # print(silhouette_avg)
    cluster_res_label = clf1.labels_
    data_set_df=Series(pressure_array,index=cluster_res_label)
    pressure_ctr=[]
    pressure_out=[]
    #算出每一个类别的均值和方差，取mean+2*std，作为该类别的值
    for i in range(n_clusters):
        pressure_ctr.append(data_set_df[i].mean()+2*data_set_df[i].std())
    #将原始压力数据替换成聚类结果，确定压力控制时间段
    for i in range(len(pressure_array)):
        pressure_out.append(pressure_ctr[clf1.labels_[i]])
    return pressure_out,silhouette_avg

def kmeans_auto(dataset,n):
    start=time.time()
    score_list=[]
    for k in range(2,n,1):
        pressure_out,score=pressure_cluster(k,dataset)
        score_list.append(score)
    best_k=score_list.index(max(score_list))+2
    pressure_out,score=pressure_cluster(best_k,dataset)
    end=time.time()
    print((end-start))
    return pressure_out


if __name__=="__main__":

    num_point = 72*2
    signal1=[]
    for i in range(num_point):
        if i<30:
            signal1.append(15+1.5*random.random())
        elif i<100:
            signal1.append(20+2*random.random())
        elif i<120:
            signal1.append(25+2*random.random())
        else:
            signal1.append(20+3*random.random())

    pressure_cluster_res=kmeans_auto(signal1,10)
    plt.plot(signal1)
    plt.plot(pressure_cluster_res)
    plt.show()

结果图片