knn算法原理

import pandas as pd
import numpy as np


def distance(v1, v2):
    """
    自实现距离计算
    :param v1: 点v1
    :param v2: 点v2
    :return: 距离
    """
    # 法一
    # ndim = len(v1)
    # summary = 0
    # for i in range(ndim):
    #     summary += (v1[i]-v2[i])**2
    # dist = np.sqrt(summary)

    # 法二
    # 借助power进行距离计算
    dist = np.sqrt(np.sum(np.power(v1-v2,2)))
    return dist


def knn(train, k, test):
    """
    进行knn自实现
    :param train: 训练集
    :param k: 取前几个
    :param test: 测试集
    :return: 电影类型预测的类别
    """
    # 进行相似度计算 ---》距离计算
    index_num = train.shape[0]
    for i in range(index_num):
        dist = distance(train.iloc[i, 1:4], test[1:])
        # print(dist)
        train.ix[i, '欧氏距离'] = dist
    train = train.sort_values(by='欧氏距离').iloc[:k, :]
    # 对前k行进行众数统计分析
    # print(train['电影类型'].mode()[0])
    # print(train)
    return train['电影类型'].mode()[0]


# 加载数据
mov = pd.read_excel('电影分类数据.xlsx')
# print(mov)
train = mov.iloc[:, 1:6]
# print(train)
test = mov.columns[-4:]
test = np.array(test)
# print(test)

# 取前5个样本
k = 5
# 进行knn自实现
y_predict = knn(train, k, test)
print('knn算法预测值：', y_predict)