自定义封装KNN算法

一：首先定义一个类，定义构造函数，对训练集进行赋值，fit和predict过程。

#自己手写KNN算法进行封装

import numpy as np
from math import sqrt
from collections import Counter

class KNN_classifier:
    #构造函数，指定key
    def __init__(self,k):
        self.k = k
        self._X_train = None
        self._Y_train = None
    #赋值
    def fit(self,x_train,y_train):
        self._X_train = x_train
        self._Y_train = y_train
        return self
    #predict
    def predict(self,x):
        assert self._X_train is not  None
        assert self._Y_train is not None
        assert self._X_train.shape[0]==self._Y_train.shape[0] ,'列数不一样'
        assert self._X_train.shape[1]==x.shape[0]

        arr = [i for i in x]
        pre = np.array(arr)

        prediction = self._prediction(pre)

        return prediction

    def _prediction(self,x):
        distice = [sqrt(np.sum(i-x)**2) for i in self._X_train]

        nearest = np.argsort(distice)

        topK = [self._Y_train[i] for i in nearest[:self.k]]

        final = Counter(topK)

        pre = final.most_common(1)[0][0]

        return pre

#测试
if __name__=='__main__':
    raw_data_X = [[1.232422,1.22324],
                  [2.324232,1.3224],
                 [2.3435353,2.3232342],
                 [3.434353,3.434353],
                 [4.54546,3.54544],
                 [7.42422,6.764353],
                 [6.42224534,7.533232],
                 [8.435353,8.5433],
                 [9.423534,9.422224],
                 [8.544444,9.4564454]]

    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 = np.array([7.5353343,8.53324232])

    knf = KNN_classifier(5)

    knf.fit(x_train,y_train)

    print(knf.predict(x))