kmeans

Posted on

玩一下kmeans，调戏以下国足，顺便预测一下18世界杯冠军，18-7-15 23:00世界杯

物以类聚，人以群分，选亚洲15支球队

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
data = {"zhongguo":[50,50,50,40],
            "riben":[28,9,29,12],
            "hanguo":[17,15,27,26],
            "yilang":[25,40,28,18],
            "shate":[28,40,50,25],
            "yilake":[50,50,40,40],
            "kataer":[50,40,40,40],
            "alianqiu":[50,40,50,40],
            "wuzibiekesitan":[40,40,40,40],
            "taiguo":[50,50,50,40],
            "yuenan":[50,50,50,50],
            "aman":[50,50,40,50],
            "balin":[40,40,50,50],
            "chaoxian":[40,32,50,50],
            "yinni":[50,50,50,50]}

依次选2006年，2010年，2014年，2018年世界杯的数据作为聚类样本，打进世界杯的得分用排名衡量，预选赛小组未出线的给50，预选赛十强的给40，澳大利亚没统计，18年的排名是估计的，虽然11点是冠亚军决赛，理论上其他队伍排名已经定了，但是我不会，这样算，得分越多的越low。

k选3，初始中心选中国，日本，沙特，先计算每一条数据到三个中心点的欧氏距离，并将其归为最近点那一类，处理完所有数据后，计算每个类的中心点，更新聚类中心，重新以上步骤，知道聚类中心不再变化，代码：

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
import numpy as np


def name2indexf(names):
    name2index = {}
    for index, name in enumerate(names):
        name2index[name] = index
    return name2index
def cacul_eudist(vec1, vec2):
    assert len(vec1) == len(vec2)
    dist = np.linalg.norm(vec1 - vec2)
    return dist

if __name__ == "__main__":
    data = {"zhongguo":[50,50,50,40],
            "riben":[28,9,29,12],
            "hanguo":[17,15,27,26],
            "yilang":[25,40,28,18],
            "shate":[28,40,50,25],
            "yilake":[50,50,40,40],
            "kataer":[50,40,40,40],
            "alianqiu":[50,40,50,40],
            "wuzibiekesitan":[40,40,40,40],
            "taiguo":[50,50,50,40],
            "yuenan":[50,50,50,50],
            "aman":[50,50,40,50],
            "balin":[40,40,50,50],
            "chaoxian":[40,32,50,50],
            "yinni":[50,50,50,50]}
    data_array = np.zeros(shape=(len(data.keys()),4))
    name2index = name2indexf(data.keys())
    for name in data.keys():
        index = name2index[name]
        data_array[index] = np.array(data[name])
    k_center = np.array([data_array[0],data_array[1],data_array[4]])
    k_center_with_near = {}

    # ------cacul center and it near-------
    epoch = 0
    while True:

        for index, item in enumerate(k_center):
            k_center_with_near[index] = []
        # ----------choice nearest for each data---------
        for index in range(len(data.keys())):
            data_item = data_array[index]
            near = 0
            dist_min = 100000
            for i in range(len(k_center)):
                dist = cacul_eudist(data_item, k_center[i])
                if dist_min > dist:
                    dist_min = dist
                    near = i
            k_center_with_near[near].append(index)

        # ------recacul center---------------
        end_tag = True
        for center_near_index in k_center_with_near.keys():
            contry_index = k_center_with_near[center_near_index]
            center_near_data = []
            for item in contry_index:
                center_near_data.append(data_array[item])
            center_near_data = np.array(center_near_data)
            new_center = np.mean(center_near_data, axis=0)
            if not (k_center[center_near_index] == new_center).all():
                end_tag = False
            k_center[center_near_index] = new_center
        print("epoch:",epoch, "center:", k_center)
        epoch += 1
        test = data.keys()
        if end_tag:
            for item in k_center_with_near.keys():
                print("classv{} include".format(item), [list(data.keys())[n] for n in (k_center_with_near[item])])
            break

结果：

1
2
3
classv0 include ['zhongguo', 'yilake', 'kataer', 'alianqiu', 'wuzibiekesitan', 'taiguo', 'yuenan', 'aman', 'balin', 'chaoxian', 'yinni']
classv1 include ['riben', 'hanguo']
classv2 include ['yilang', 'shate']

这样算，中国队在亚洲只能算3流球队

预测克罗地亚冠军，虽然实力比法国弱一些，但是不要低估对冠军渴望的心