信息检索——kNN分类

---

一、实验目的

  通过实验，掌握利用kNN分类算法

---

二、实验任务与要求

  计算k(1,3,5,8)取不同值时的kNN的分类结果，并总结差异

---

三、code

1.code

import matplotlib.pyplot as plt
import heapq
import math
import copy

def pic_show():
    black_x=[0.4,0.6,1,1,1.3,1.6,1.9,2,2,2,2.5,2.5,2.5,2.5,2.6]
    black_y=[4,6,5.5,7,3,4,6,2,6,7,1,3,4,5,6]
    black = list(zip(black_x, black_y))

    red_x=[4.4,5.5,5.8,6,6,6.5,7,7.5,7.5,7.5,8,8,9]
    red_y=[3,4,3.8,2,5,1,5,2,4,6,3,7,3]
    red = list(zip(red_x, red_y))

    test1_x = [4]
    test1_y = [5.5]
    test1 = list(zip(test1_x, test1_y))

    test2_x = [4.5]
    test2_y = [2]
    test2 = list(zip(test2_x, test2_y))


    plt.figure(figsize=(8, 4))#figsize:确定画布大小
    plt.scatter(black_x,black_y,c='black')
    plt.scatter(red_x,red_y,c='red')
    plt.scatter(test1_x,test1_y,c='green',label='test1')
    plt.scatter(test2_x,test2_y,c='blue',label='test2')
    plt.legend()
    plt.show()

    return black,red,test1,test2

def distance(xy,test):
    d = math.sqrt((xy[0] - test[0][0]) ** 2 + (xy[1] - test[0][1]) ** 2)
    return d


def kNN(nums, find_nums):
    if len(nums) == len(list(set(nums))):
        # 使用heapq
        min_number = heapq.nsmallest(find_nums, nums)
        min_num_index = list(map(nums.index, min_number))
    else:
        # 使用deepcopy
        nums_copy = copy.deepcopy(nums)
        max_num = max(nums) + 1
        min_num_index = []
        min_number = []
        for i in range(find_nums):
            num_min = min(nums_copy)
            num_index = nums_copy.index(num_min)
            min_number.append(num_min)
            min_num_index.append(num_index)
            nums_copy[num_index] = max_num

    return min_num_index, min_number

all_d=[]
black,red,test1,test2=pic_show()
black_l=len(black)
red_l=len(red)
#print(black,red,black_l,red_l)


all=black+red
print("全部数组为：",all)
test=test1


choice=eval(input("请输入你要选择的测试用例：1 or 2 :(默认为test1)"))
if choice == 1:
    test = test1
elif choice == 2:
    test =test2

for xy in all:
    d = distance(xy, test)
    all_d.append(d)
print("元组距离为:",all_d)

k=eval(input("请输入k值:"))
min_num_index, min_number  = kNN(all_d, k)
print('索引:', min_num_index)
print('元素值:', min_number)

black_num=0
red_num=0
for i in min_num_index:
    if i+1 <= black_l:
        black_num+=1
    else:
        red_num+=1

P_black=black_num/k
P_red=red_num/k

print("该文档属于black的概率为{}".format(P_black))
print("该文档属于red的概率为{}".format(P_red))

2.测试样例

这里测试文章取test1
k=1,3,5,8

3.实验结果

4.总结差异

  当取不同k值时，所选取的不同类别文章个数不同，进而使得测试文章的所属类别的概率发生变化，将每篇测试文档分到训练集中离它最近的k篇文档所属类别中最多的那个类别。
  需要注意的是：当训练集非常大的时候，kNN分类的精度很高；如果训练集很小，kNN可能效果很差。