k-means without numpy, 无第三方库的实现

k-means.py

from doctest import testmod
from math import dist as 距离
from pathlib import Path
from random import sample
from statistics import fmean
from typing import List, Tuple

Point = Tuple[float, ...]


def 计算中间点(点集: List[Point]):
    """
    >>> 计算中间点([(6,8),(-5,8)])
    (0.5, 8.0)
    """
    return tuple(map(fmean, zip(*点集)))


def kmeans(点集: List[Point], k: int):
    初始点集合 = sample(点集, k)
    结果字典 = {i: [x] for i, x in enumerate(初始点集合)}
    均值字典 = dict(enumerate(初始点集合))
    样本点的集合 = set(点集) - set(初始点集合)
    for 点 in 样本点的集合:
        最近的均值编号 = min(range(k), key=lambda i: 距离(均值字典[i], 点))
        结果字典[最近的均值编号].append(点)
        更新均值 = 计算中间点(结果字典[最近的均值编号])
        均值字典[最近的均值编号] = 更新均值
    return 结果字典


if __name__ == "__main__":
    testmod(verbose=True)

    def 读取文件():
        with open(Path(__file__).parent.absolute() / "data.txt") as 文件:
            yield from (tuple(map(float, 每一行.split())) for 每一行 in 文件)

    print(kmeans(list(读取文件()), 5))