Machine Learning in Action - KNN

#!/usr/bin/env python
# -*- Coding:UTF-8 -*-


import numpy as np
import operator
import os


def create_dataset():
    group = np.array([[1.0, 1.1], [1.0, 1.0], [0.0, 0.0], [0.0, 0.1]])
    labels = ['A', 'A', 'B', 'B']
    return group, labels





def class_dataset(test_data, train_data, train_label, k):
    '''
    function:knn分类
    :param test_data:
    :param train_data:
    :param train_label:
    :param k:
    :return:
    程序亮点：np.argsort, get, sorted, operator.itemgetter
    '''

    test_mat = np.tile(test_data, (len(train_data), 1))
    distance = ((test_mat - train_data) ** 2).sum(axis=1)
    distance = distance ** 0.5
    index_distance = distance.argsort()

    class_dic = {}
    for i in range(k):
        data_label = train_label[index_distance[i]]
        class_dic[data_label] = class_dic.get(data_label, 0) + 1

    class_dic = sorted(class_dic.items(), key=operator.itemgetter(1), reverse=True)

    return class_dic[0][0]






def read_file(filename):
    '''
    function:从txt文本数据---->arrary数据
    两行红色注释，可以分别与其下面代码替换
    :param filename:
    :return:
    '''
    file = open(filename, encoding='utf-8')
    file_list = file.readlines()
    num_file = len(file_list)
    num_feature = 3

    feature_mat = np.zeros((num_file, num_feature))
    label_mat = np.zeros(num_file)
    # label_mat = []

    index = 0
    for line in file_list:
        line = line.strip()
        line_list = line.split('\t')
        feature_mat[index, :] = line_list[0:num_feature]
        label_mat[index] = line_list[num_feature]
        # label_mat.append(int(line_list[num_feature]))
        index = index + 1
    return feature_mat, label_mat






def normal_data(data_mat):
    m = data_mat.shape[0]
    max_value = data_mat.max(axis=0)
    min_value = data_mat.min(axis=0)
    ranges = max_value - min_value

    data_mat = (data_mat - np.tile(min_value, (m, 1))) / np.tile(ranges, (m, 1))
    return data_mat, ranges, min_value




def class_test():
    ratio = 0.10
    training_data, train_label = read_file('datingTestSet1.txt')
    train_data, ranges, min_value = normal_data(training_data)
    num_train = train_data.shape[0]
    num_test = int(ratio * num_train)
    error_count = 0

    for i in range(num_test):
        class_result = class_dataset(train_data[i, :], train_data[num_test:, :],
                                     train_label[num_test:], 3)
        print('''The classifier came back with: {}, the real answer is :{}'''
              .format(class_result, train_label[i]))
        if class_result == train_label[i]:
            pass
        else:
            error_count = error_count + 1
    print('The total error rate is %2f' % (error_count / float(num_test)))








def class_person():
    '''
    误点：
    :return:
    '''

    class_dic = {1: 'not at all',
                 2: 'in small doses',
                 3: 'in large doses'}

    train_data, train_label = read_file('datingTestSet2.txt')
    train_data, ranges, min_vale = normal_data(train_data)


    fly = float(input('请输入每年获得的飞行常客里程数：'))
    game_time = float(input('请输入每天玩视频游戏所消耗时间百分比：'))
    ice_cream = float(input('请输入每周消费的冰淇淋公斤数：'))

    test_data = np.array([fly, game_time, ice_cream])
    test_data = (test_data - min_vale) / ranges
    ret = class_dataset(test_data, train_data, train_label, 3)
    return '系统提示：你对该名男子的感觉是：{}'.format(class_dic[ret])



def img_vector(filemane):
    file = open(filemane)
    num = 32
    vector = np.zeros((1, num*num))

    for i in range(num):
        line = file.readline()
        for j in range(num):
            vector[0, 32*i+j] = int(line[j])
    return vector







def hand_writing():
    train_label = []
    train_list = os.listdir('trainingDigits')
    num_train = len(train_list)
    train_data = np.zeros((num_train, 1024))



    for i in range(num_train):
        file = train_list[i]

        train_data[i, :] = img_vector('trainingDigits/%s' % file)
        file = file.strip().split('_')
        label = int(file[0])
        train_label.append(label)


    test_list = os.listdir('testDigits')
    error = 0.0

    num_test = len(test_list)

    for i in range(num_test):
        file = test_list[i]
        test_data = img_vector('testDigits/%s' % file)
        file = file.strip().split('_')
        test_label = int(file[0])
        ret = class_dataset(test_data, train_data, train_label, 3)
        if ret == test_label:
            pass
        else:
            error = error + 1
            print('''
                    The class result is: {}
                    The real answer is: {}'''.format(ret, test_label))
        # print('''
        # The class result is: {}
        # The real answer is: {}'''.format(ret, test_label))


    print('The total error rate is:{}'.format(error/float(num_test)))

代码调用：

#!/usr/bin/env python
# -*- Coding:UTF-8 -*-


import new_knn
import matplotlib.pyplot as plt
import matplotlib
import numpy as np



'''
group, labels = new_knn.create_dataset()
ret = new_knn.class_dataset([2,3], group, labels,3)
print(ret)
'''






'''
# 作图
mydata, labels = new_knn.read_file('datingTestSet2.txt')

fig = plt.figure()
ax = fig.add_subplot(1, 1, 1)

ax.scatter(mydata[:, 1], mydata[:, 2], 15.0*np.array(labels), 15.0*np.array(labels))
plt.show()


'''


'''
mydata, labels = new_knn.read_file('datingTestSet2.txt')
norm_mat, ranges, min_vals = new_knn.normal_data(mydata)
'''






'''
# 测试
new_knn.class_test()
'''




'''
# 构建完整可用的系统
ret = new_knn.class_person()
print(ret)

'''



'''
ret = new_knn.img_vector('testDigits/0_13.txt')
print(ret)

'''








new_knn.hand_writing()