DBSCAN matlab代码实现

以下是DBSCAN算法的Matlab实现代码：

function [clusterIdx, nClusters] = dbscan(X, epsilon, minPoints)
% DBSCAN clustering algorithm
% X: data matrix (nSamples x nFeatures)
% epsilon: neighborhood radius
% minPoints: minimum number of points in a cluster
% clusterIdx: cluster index for each point (-1 for noise)
% nClusters: total number of clusters

% initialize cluster index
nSamples = size(X, 1);
visited = false(nSamples, 1);
clusterIdx = zeros(nSamples, 1);
clusterId = 0;

% calculate pairwise distance matrix
D = pdist2(X, X);

for i = 1:nSamples
    if ~visited(i)
        visited(i) = true;
        
        % find neighbors
        neighbors = find(D(i, :) <= epsilon);
        
        % check if core point
        if length(neighbors) >= minPoints
            clusterId = clusterId + 1;
            clusterIdx(i) = clusterId;
            
            % expand cluster to density-reachable points
            j = 1;
            while j <= length(neighbors)
                k = neighbors(j);
                if ~visited(k)
                    visited(k) = true;
                    
                    % find k's neighbors
                    kNeighbors = find(D(k, :) <= epsilon);
                    if length(kNeighbors) >= minPoints
                        neighbors = [neighbors, kNeighbors];
                    end
                end
                
                % assign cluster index
                if clusterIdx(k) == 0
                    clusterIdx(k) = clusterId;
                end
                
                j = j + 1;
            end
        else
            % mark as noise
            clusterIdx(i) = -1;
        end
    end
end

% count number of clusters
nClusters = max(clusterIdx);
end

其中，输入参数X为数据矩阵，epsilon为邻域半径，minPoints为最小点数。输出参数clusterIdx为每个点的聚类标签，-1表示噪声点；nClusters为总聚类数。