机器学习基础（四）之KNN算法简析

Jupyter Notebook

k近邻算法

import numpy as np
from matplotlib import pyplot as plt
import math

#自己弄的训练集和结果
X_train = np.array([[1,1.2],[3,3.3],[5,2.2],[7,5.7],[9,8.4]])
X_train
array([[1. , 1.2],
       [3. , 3.3],
       [5. , 2.2],
       [7. , 5.7],
       [9. , 8.4]])
       
Y_train = np.array([0,0,0,1,1])

Y_train
array([0, 0, 0, 1, 1])

plt.scatter(x_train[y_train == 0,0],x_train[y_train == 0,1],color="r",label="normal")
plt.scatter(x_train[y_train == 1,0],x_train[y_train == 1,1],color="g",label="bad")
plt.title("search")

机器学习基础（四）之KNN算法简析


`#x是我们要测试的值在下图用蓝色表示
x = np.array([6,3.7])

b
plt.scatter(x_train[y_train == 0,0],x_train[y_train == 0,1],color="r",label="normal")
plt.scatter(x_train[y_train == 1,0],x_train[y_train == 1,1],color="g",label="bad")
plt.scatter(x[0],x[1],color="b")
plt.title("search")
Text(0.5, 1.0, 'search')

` 机器学习基础（四）之KNN算法简析
KNN，求x与其他点的距离，并保存进distant中去

distant = []
for x_train in X_train:
    d = math.sqrt(np.sum((x - x_train)**2))
    distant.append(d)

distant
[5.5901699437494745,
 3.026549190084311,
 1.8027756377319946,
 2.23606797749979,
 5.57584074378026]
 
对所获取到的距离进行下标排序
np.argsort(distant)
np.argsort(distant)
array([2, 3, 1, 4, 0], dtype=int64)

nearst = np.argsort(distant)
获取到排序号的点坐标，进行相应Y_train中坐标的值，才是我们想要的

k =3 说明与三个数进行比较，看离谁近，则进行投票

#从Y_train结果训练集中获取最近三个点的结果，通过索引来获取
topk_y = [Y_train[i] for i in nearst[:k]]

topk_y
[0, 1, 0]
此时我们得出了0，1，0说明 0胜出，预测到的结果应该是0

from collections import Counter
Counter(topk_y)
Counter({0: 2, 1: 1})
计数完进行投票

votes = Counter(topk_y)

votes.most_common(1)
[(0, 2)]

votes.most_common(1)[0][0]
0
把值存在predict中来
predict = votes.most_common(1)[0][0]

predict
0
至此，简单的K临近算法已经差不多了呀

机器学习基础（四）之KNN算法简析

相关推荐