python实现基本的机器学习算法系列（4）：KNN

# sum(y_pred == y_test)
# [max(y_pred_k[i], key=list(y_pred_k[i]).count)， 以key的函数判断max是怎么运算，这里是求出现最多次的值
# y_pred_k[i]=[1,2,3,3]  结果为3

import numpy as np
from sklearn.datasets import load_digits
from sklearn.model_selection import train_test_split

mnist = load_digits()   # (1797,64)和(1797,)
X = mnist['data']
y = mnist['target']

x_train, x_test, y_train, y_test = train_test_split(X, y)
print(x_train.shape)
print(y_train.shape)
print(x_test.shape)
print(y_test.shape)


class knn:

    # 每一个测试样本要和所有训练样本比较距离
    def __init__(self):
        pass

    def fit(self, x_train, y_train):
        self.x = x_train
        self.y = y_train

    def euclidean_distance(self, x_):

        if x_.ndim == 1:   # only one sample
            Ed = np.sqrt(np.sum((self.x-x_)**2, axis=1))

        if x_.ndim == 2:   # samples * dimensions
            samples, dimensions = x_.shape
            Ed = [np.sqrt(np.sum((self.x-x_[i])**2, axis=1)) for i in range(samples)]   # test_sample * train_sample

        return np.array(Ed)

    def prediction(self, x_test, y_test, k):

        Ed = self.euclidean_distance(x_test)
        if k==1:
            if x_test.ndim == 1:
                idx = np.argmin(Ed)
                # y_pred = self.y[idx]
            else:
                idx = np.argmin(Ed, axis=1)
            y_pred = self.y[idx]

        else:
            if x_test.ndim == 1:
                Ed_order = np.argsort(Ed)   # 从小到大进行排序，小的值索引在前，大的值索引在后
                idx = Ed_order[:k]
                y_pred_k = self.y[idx]
                y_pred = max(y_pred_k, key=list(y_pred_k).count)
            else:
                Ed_order = np.argsort(Ed, axis=1)
                idx = Ed_order[:, :k]
                y_pred_k = self.y[idx]
                y_pred = [max(y_pred_k[i], key=list(y_pred_k[i]).count) for i in range(y_pred_k.shape[0])]

        # score = len(np.where(y_pred == y_test)[0])/len(y_pred)*100
        score = sum(y_pred == y_test)/len(y_pred)*100
        return y_pred, score

if __name__ == '__main__':
    knn = knn()
    knn.fit(x_train, y_train)
    y_pred, score = knn.prediction(x_test, y_test, k=3)
    print(f"Score is {score}.")
    print(y_pred[:10], y_test[:10])