DIANA聚类代码

以下是一个简单的DIANA聚类的Python代码示例：

import numpy as np
from scipy.spatial.distance import pdist, squareform

def diana(X, k):
    # 计算初始距离矩阵
    D = squareform(pdist(X))
    n = len(X)

    # 开始聚类
    clusters = [[i] for i in range(n)]
    while len(clusters) < k:
        max_d = 0
        to_merge = None
        # 找到距离最远的簇对
        for i in range(len(clusters)):
            for j in range(i+1, len(clusters)):
                d = 0
                for m in clusters[i]:
                    for n in clusters[j]:
                        d += D[m][n]
                d /= len(clusters[i]) * len(clusters[j])
                if d > max_d:
                    max_d = d
                    to_merge = (i, j)
        # 合并簇
        i, j = to_merge
        clusters[i] = clusters[i] + clusters[j]
        del clusters[j]

    # 返回簇的索引
    labels = np.zeros(n)
    for i, c in enumerate(clusters):
        for j in c:
            labels[j] = i

    return labels

输入参数为数据集 `X` 和需要的簇的数量 `k`，输出为每个样本所属的簇的索引。这个代码使用了最短距离法，即每次合并距离最远的两个簇。