1、底层算法

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

#先随机设置十个样本点表示十杯酒
rowdata = {'颜色深度':[14.23,13.2,13.16,14.37,13.24,12.07,12.43,11.79,12.37,12.04],
          '酒精浓度':[5.64,4.38,5.68,4.80,4.32,2.76,3.94,3.  ,2.12,2.6 ],
          '品种':[0,0,0,0,0,1,1,1,1,1]}
# 0代表黑皮诺 ， 1代表赤霞珠
wine_data = pd.DataFrame(rowdata)
wine_data

X = np.array(wine_data.iloc[:,0:2])
X
y = np.array(wine_data.iloc[:,-1])
y
new_data = np.array([12.8,4.1])   #要判断的数据

# 1 算距离
from math import sqrt
distance = [sqrt(np.sum((x-new_data)**2)) for x in X]
distance

# 2 找邻居
sort_dist = np.argsort(distance)  #排序 返回数组的索引
sort_dist

k = 3   #选取前3个样本
topK = [y[i] for i in sort_dist[:k]]
topK

# 3 做分类
from collections import Counter   #对已有的数据类别进行计数，返回字典

votes = Counter(topK)
votes

predict = votes.most_common(1)[0][0]   #排序，索引 次数出现最多的
predict

#打包成函数
def KNN(inx,dataset,k):
    import numpy as np
    import pandas as pd
    from math import sqrt
    from collections import Counter
    
    result=[]
    distance = [sqrt(np.sum((x-inx)**2)) for x in np.array(dataset.iloc[:,0:2])]
    sort_dist = np.argsort(distance)
    topK = [dataset.iloc[:,-1][i] for i in sort_dist[:k]]
    result.append(Counter(topK).most_common(1)[0][0])
    return result
    
KNN(new_data,wine_data,3)

2、sklearn实现

from sklearn.neighbors import KNeighborsClassifier  #  调包  k近邻分类型

# 实例化（赋值的过程：将算法本身的模型赋值给一个变量）
clf = KNeighborsClassifier(n_neighbors=3)
# 训练模型
clf = clf.fit(X,y)
# 预测输出,返回预测的标签
result = clf.predict([[12.8,4.1]])
result

# 模型的评估，接口score返回预测的准确率
score = clf.score([[12.8,4.1]],[0])
score
# 返回预测的概率
clf.predict_proba([[12.8,4.1]])