构建K-Means++类

以下是一个基于 Python 的 K-Means++ 类的示例代码：

import numpy as np

class KMeansPP:
    def __init__(self, k):
        self.k = k
        
    def fit(self, X):
        n_samples, n_features = X.shape
        
        # Initialize centroids list and add the first random centroid
        centroids = [X[np.random.randint(n_samples)]]
        
        # Add remaining k-1 centroids
        for i in range(1, self.k):
            # Initialize distances list
            distances = []
            for j in range(n_samples):
                # Compute distance between each sample and the nearest centroid
                min_dist = np.inf
                for centroid in centroids:
                    dist = np.linalg.norm(X[j] - centroid)
                    min_dist = min(min_dist, dist)
                distances.append(min_dist)
            
            # Add a new centroid randomly based on the distance weights
            distances_sum = sum(distances)
            distances = [dist / distances_sum for dist in distances]
            new_centroid = X[np.random.choice(n_samples, p=distances)]
            centroids.append(new_centroid)
        
        self.centroids = centroids
        
        # Perform K-Means clustering
        clusters = [[] for _ in range(self.k)]
        for sample in X:
            distances = [np.linalg.norm(sample - centroid) for centroid in self.centroids]
            closest_centroid_idx = np.argmin(distances)
            clusters[closest_centroid_idx].append(sample)
        
        self.clusters = clusters

这个类的主要功能是进行 K-Means++ 聚类，其中 `k` 是簇的数量，`X` 是形状为 `(n_samples, n_features)` 的输入数据。在 `fit` 方法中，它首先从输入数据中选择一个随机初始中心点，然后根据每个样本到最近中心点的距离选择下一个中心点，并重复此过程直到选择 `k` 个中心点。然后，它使用这些中心点对输入数据进行 K-Means 聚类，并将每个样本分配到最近的中心点的簇中。 `centroids` 和 `clusters` 属性分别存储所选的中心点和生成的簇。