Python机器学习k-近邻算法(K Nearest Neighbor)实例详解

# -*- coding:utf-8 -*-
#! python2
import numpy as np
import operator
# 训练集
data_set = np.array([[1., 1.1],
           [1.0, 1.0],
           [0., 0.],
           [0, 0.1]])
labels = ['A', 'A', 'B', 'B']
def classify_knn(in_vector, training_data, training_label, k):
  """
  :param in_vector: 待分类向量
  :param training_data: 训练集向量
  :param training_label: 训练集标签
  :param k: 选择最近邻居的数目
  :return: 分类器对 in_vector 分类的类别
  """
  data_size = training_data.shape[0] # .shape[0] 返回二维数组的行数
  diff_mat = np.tile(in_vector, (data_size, 1)) - data_set # np.tile(array, (3, 2)) 对 array 进行 3×2 扩展为二维数组
  sq_diff_mat = diff_mat ** 2
  sq_distances = sq_diff_mat.sum(axis=1) # .sum(axis=1) 矩阵以列求和
  # distances = sq_distances ** 0.5 # 主要是通过比较求最近点,所以没有必要求平方根
  distances_sorted_index = sq_distances.argsort() # .argsort() 对array进行排序 返回排序后对应的索引
  class_count_dict = {} # 用于统计类别的个数
  for i in range(k):
    label = training_label[distances_sorted_index[i]]
    try:
      class_count_dict[label] += 1
    except KeyError:
      class_count_dict[label] = 1
  class_count_dict = sorted(class_count_dict.iteritems(), key=operator.itemgetter(1), reverse=True) # 根据字典的value值对字典进行逆序排序
  return class_count_dict[0][0]
if __name__ == '__main__':
  vector = [0, 0] # 待分类数据集
  print classify_knn(in_vector=vector, training_data=data_set, training_label=labels, k=3)