dbscan算法 python代码

以下是 DBSCAN 聚类算法的 Python 代码实现：

from sklearn.neighbors import NearestNeighbors
import numpy as np

class DBSCAN:
    def __init__(self, eps=0.5, min_samples=5):
        self.eps = eps
        self.min_samples = min_samples
        self.core_indices_ = None
        self.labels_ = None
        
    def fit(self, X):
        n_samples = X.shape[0]
        self.core_indices_ = np.zeros(n_samples, dtype=bool)
        nbrs = NearestNeighbors(n_neighbors=self.min_samples, metric='euclidean').fit(X)
        distances, indices = nbrs.kneighbors(X)
        core_distances = np.zeros(n_samples)
        for i in range(n_samples):
            core_distances[i] = distances[i, self.min_samples - 1]
            if core_distances[i] <= self.eps:
                self.core_indices_[i] = True
        labels = np.zeros(n_samples)
        current_label = 1
        for i in range(n_samples):
            if not self.core_indices_[i]:
                continue
            if labels[i] != 0:
                continue
            labels[i] = current_label
            neighbors = indices[i, 1:].tolist()
            while len(neighbors) > 0:
                j = neighbors.pop(0)
                if labels[j] == -1:
                    labels[j] = current_label
                if labels[j] != 0:
                    continue
                labels[j] = current_label
                if self.core_indices_[j]:
                    neighbors += indices[j, 1:].tolist()
            current_label += 1
        self.labels_ = labels
        return self.labels_

该算法实现了 DBSCAN 算法，其中 `eps` 和 `min_samples` 是 DBSCAN 算法的两个超参数，分别表示邻域半径和最小样本数。在 `fit` 方法中，首先计算每个样本点的核心距离，然后确定核心点的索引，通过 BFS（广度优先搜索）遍历所有核心点的邻居，将它们分配到对应的聚类中，并标记为已访问。最终返回所有样本点的聚类标签。