KNN代码整理

原创已于 2022-09-08 15:55:13 修改 · 357 阅读

0 ·

CC 4.0 BY-SA版权

文章标签：

#python #机器学习 #numpy

于 2020-06-30 20:23:09 首次发布

笔记同时被 2 个专栏收录

42 篇文章

订阅专栏

机器学习

12 篇文章

订阅专栏

SKNN+train_test_split实例

import numpy as np
import matplotlib.pyplot as plt
from sklearn import datasets

1. 导入鸢尾花的数据集

iris=datasets.load_iris()
x=iris.data
y=iris.target
x.shape

(150, 4)

y.shape

(150,)

array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
       0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
       1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
       1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
       2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
       2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2])

2.train_test_split过程

2.1 将x训练集中的元素进行乱序处理，返回索引np.random.permutation(len(x))

shuffle_indexes=np.random.permutation(len(x))

shuffle_indexes

array([ 65,   9, 132,  90, 128,  40,  63,  28, 130,   3,  27, 108, 105,
        71,  26,  94,  37,  43,  73,  22,  31, 138,  15,  52, 131,  29,
        84,  93,  55, 149,  49,  91,   7, 116, 127,  36, 106, 137, 115,
        46, 124,  96,  77,   4,   8,  57, 136,  21, 113,  82, 134, 143,
       114,  42, 112,  88,  85, 118, 147,  50,  13,  14,  48,  69,  67,
        12,  16,  11, 141, 117, 142,   5, 126, 121,  19,  17, 122,  39,
        30,  38,  45,  75, 144, 123,  34,  51,  23, 109, 148, 110,  56,
        81,  54,  68,  61,  35,  41,  78, 103,  32,  99,   0, 145, 140,
        58,  10,  47,  72, 104,  87, 111,  64, 107, 102,  33,  80,  74,
        83,  59,  95, 135,  20,  89, 146,  18,  24,  86,  92,  66,  76,
        25,   2,  98, 101,  53,  79,  70,  60, 129, 133, 139,  62, 119,
         1,  97, 125, 120,   6,  44, 100])

2.2 测试数据集

test_ratio=0.2
test_size=int(len(x)*test_ratio)

test_size

test_indexes=shuffle_indexes[:test_size]
train_indexes=shuffle_indexes[test_size:]

x_test=x[test_indexes]
y_test=y[test_indexes]
x_train=x[train_indexes]
y_train=y[train_indexes]

print(x_train.shape)
print(y_train.shape)

(120, 4)
(120,)

print(x_test.shape)
print(y_test.shape)

(30, 4)
(30,)

3. sklearn 中的train_test_split

from sklearn.model_selection import train_test_split
x_train,x_test,y_train,y_test=train_test_split(x,y,test_size=0.2,random_state=666)
print(x_train.shape)
print(y_train.shape)

(120, 4)
(120,)

print(x_test.shape)
print(y_test.shape)

(30, 4)
(30,)

4.进行预测y_predict——SKNN算法

from sklearn.neighbors import KNeighborsClassifier
kNNClassifier=KNeighborsClassifier(n_neighbors=3)
kNNClassifier.fit(x_train,y_train)
y_predict=kNNClassifier.predict(x_test)
y_predict

array([1, 2, 1, 2, 0, 1, 1, 2, 1, 1, 1, 0, 0, 0, 2, 1, 0, 2, 2, 2, 1, 0,
       2, 0, 1, 1, 0, 1, 2, 2])

5. 算准确率用y_test与y_predict结果对比

y_test

array([1, 2, 1, 2, 0, 1, 1, 2, 1, 1, 1, 0, 0, 0, 2, 1, 0, 2, 2, 2, 1, 0,
       2, 0, 1, 1, 0, 1, 2, 2])

sum(y_predict==y_test)/len(y_test)

1.0

具体SKNN过程实现举例

import numpy as np
import matplotlib.pyplot as plt
raw_data_x=[[3.3,2.3],
            [3.1,1.7],
            [1.3,3.6],
            [3.5,4.6],
            [2.2,2.8],
            [7.4,4.6],
            [5.7,3.5],
            [9.1,2.5],
            [7.7,3.4],
            [7.9,0.7]
           ]
raw_data_y=[0,0,0,0,0,1,1,1,1,1]
x_train=np.array(raw_data_x)
y_train=np.array(raw_data_y)

x_train

array([[3.3, 2.3],
       [3.1, 1.7],
       [1.3, 3.6],
       [3.5, 4.6],
       [2.2, 2.8],
       [7.4, 4.6],
       [5.7, 3.5],
       [9.1, 2.5],
       [7.7, 3.4],
       [7.9, 0.7]])

y_train

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

x=np.array([8.0,3.3])

1. 计算测试集中y离x的距离distance

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

[4.805205510693586,
 5.154609587543949,
 6.706713054842886,
 4.684015371452148,
 5.821511831131154,
 1.431782106327635,
 2.308679276123039,
 1.360147050873544,
 0.31622776601683783,
 2.601922366251537]

distances= [sqrt( np.sum((x_train- x)** 2)) for x_train in x_train ]
distances

[4.805205510693586,
 5.154609587543949,
 6.706713054842886,
 4.684015371452148,
 5.821511831131154,
 1.431782106327635,
 2.308679276123039,
 1.360147050873544,
 0.31622776601683783,
 2.601922366251537]

2. 求离x最近的k个数据对应的结果

np.argsort(distances)

array([8, 7, 5, 6, 9, 3, 0, 1, 4, 2], dtype=int64)

nearest=np.argsort(distances)
k=6
topK_y=[y_train[i] for i in nearest[:k]]
topK_y

[1, 1, 1, 1, 1, 0]

3. 结果中个数最多的值为预测值

from collections import Counter
Counter(topK_y)

Counter({1: 5, 0: 1})

votes=Counter(topK_y)
votes.most_common(1)

[(1, 5)]

y_predict=votes.most_common(1)[0]

4. votes.most_common()求票数最多的几个，返回结果为一个二维数组

votes=Counter(topK_y)
votes.most_common(1)

[(1, 5)]

y_predict=votes.most_common(1)[0][0]
y_predict