李航《统计学习方法》K近邻学习算法实现

>>> from sklearn.datasets import load_iris
>>> iris = load_iris()
>>> iris.target[[10, 25, 50]]
array([0, 0, 1])
>>> list(data.target_names)
['setosa', 'versicolor', 'virginica']

import numpy as np
import random
import heapq

#对数据进行预处理，分成训练集和测试集
def Preprocessing():
    from sklearn import datasets

    testset = []       #测试集
    test_labels = []   #测试集标签
    trainset = []      #训练集
    train_labels = []  #训练集标签
    
    #加载scikit-learn中自带的Iris数据集
    iris = datasets.load_iris()
    datas = iris.data[:, 0:3] #获取所有数据，不含标签
    label = iris.target       #获取所有标签
    
    #获取0到149中任意不重复的10的数
    test_list = np.sort([random.randint(0,149) for _ in range(10)])
    #对150个数据进行划分
    for i in range(150):
        if i in test_list:
            testset.append(datas[i])
            test_labels.append(label[i])
        else:
            trainset.append(datas[i])
            train_labels.append(label[i])

    testset = np.mat(testset) #转换为矩阵类型
    test_labels = np.mat(test_labels).transpose() #进行矩阵的转置
    trainset = np.mat(trainset)
    train_labels = np.mat(train_labels).transpose()
    return testset, test_labels, trainset, train_labels

def Predict(testset, trainset, train_labels, n):
    k = n
    predict = []
    #遍历测试数据集中的测试用例坐标
    for  test_vec in testset:
        dist_list = []      #当前测试数据与训练数据的距离
        knn_list = []       #当前k个最邻近点
        
        #遍历训练数据集中的训练用例坐标及标签
        for i in range(len(train_labels)):
            label = train_labels[i]
            train_vec = trainset[i]
            
            dist = np.linalg.norm(train_vec - test_vec)   #计算两个坐标的欧式距离
            dist_list.append((dist, label))
        
        #通过python中堆结构获取k个最近邻
        knn_list = heapq.nsmallest(k, dist_list, key=lambda x: x[0])
        
        # 统计选票
        class_total = 3
        class_count = [0 for i in range(class_total)]
        for dist, label in knn_list:
            label = label.tolist()  #将矩阵类型转换为列表类型
            class_count[label[0][0]] += 1
        
        # 找出最大选票
        mmax = max(class_count)

        # 找出最大选票标签
        for i in range(class_total):
            if mmax == class_count[i]:
                predict.append(i)
                break
    
    return predict

def main():
    k = 10
    testset, test_labels, trainset, train_labels = Preprocessing()
    predict = Predict(testset, trainset, train_labels, k)
    test_labels = test_labels.tolist()

    error_count = 0
    for i in range(10):
        print("testset:",test_labels[i][0], "predict:", predict[i])
        if test_labels[i][0] != predict[i]:
            error_count += 1
    print("error_count:",error_count)
   
if __name__ == '__main__':
    main()

testset: 1 predict: 1
testset: 1 predict: 1
testset: 1 predict: 1
testset: 1 predict: 1
testset: 2 predict: 2
testset: 2 predict: 2
testset: 2 predict: 2
testset: 2 predict: 1
testset: 2 predict: 2
testset: 2 predict: 2
error_count: 1