scikit-learn估计器-KNN均值聚类

# -*- coding: utf-8 -*-
"""
Created on Thu Mar  2 21:43:06 2017
@author: zzpp220
"""
from pandas import DataFrame,Series
import pandas as pd
import numpy as np
import os,csv
from sklearn.cross_validation import train_test_split
from sklearn.cross_validation import cross_val_score#交叉验证
from matplotlib import pyplot as plt
data_folder=‘/media/zzpp220/Data/Linux_Documents/DOWNLOAD/python-DataAnalysis/xianku_book‘
data=os.path.join(data_folder,‘ionosphere.data‘)#csv --comma-seperated-values
X=np.zeros((351,34),dtype=‘float‘)
y=np.zeros((351,),dtype=‘bool‘)
with open(data,‘r‘) as input:
    reader=csv.reader(input)
#    X=np.array([])
    for i ,row in enumerate(reader):
        data=[float(datum) for datum in row[:-1]]
        X[i]=data#读数据至数组Ｘ
        y[i]=row[-1]==‘g‘#读类别值数组y
    ## set train_set and test_set
    X_train,X_test,y_train,y_test=train_test_split(X,y,random_state=14)
    ##导入ｋ近邻分类器，并为其初始化一个实例，参数用默认的，以后再讲调参，该算法默认选择５个近邻作为分类一句
    from sklearn.neighbors import  KNeighborsClassifier 
    estimator=KNeighborsClassifier()# 建立实例
    #估计器创建号之后，就要用进行数据训练，
    #k近邻估计器分析训练集中　的数据，比较待分类的新数据点和训练集中的数据，找到新数据点的　近邻
    estimator.fit(X_train,y_train)
    #用测试集测试孙发，【评估在测试集上的表现
    y_predicted=estimator.predict(X_test)
    accuracy=np.mean(y_test==y_predicted)*100
    print ‘the accuracy is {0:.1f}%‘.format(accuracy)
    
    scores=cross_val_score(estimator,X,y,scoring=‘accuracy‘)
    average_accuracy=np.mean(scores)*100
    print ‘the accuracy is {0:.1f}%‘.format(average_accuracy)
    
    ##调参优化结果
   #ｋ近邻算法有多个参数，最终药店　是初始紧邻点的选择，想测试一系列的紧邻点的值，可以进行多次实验
    avg_Scores,all_score=[],[]
    parameter_values=range(1,21)
    for num_neighbor in parameter_values:
        multi_test_estimator=KNeighborsClassifier(n_neighbors=num_neighbor)
        multi_testscore=cross_val_score(multi_test_estimator,X,y,scoring=‘accuracy‘)
        all_score.append(multi_testscore)
        avg_Scores.append(np.mean(multi_testscore))
    plt.plot(parameter_values,avg_Scores,‘-o‘)#画图，横坐标，纵坐标，点的标志
    plt.savefig(‘knn-multi-neighbor-test.pdf‘,dpi=400,bbox_inchs=‘tight‘)#保存成pdf格式，后面是像素和白边的设置

附件列表