【摘要】 启发式聚类算法具有收敛速度快、易实现等优点,但初始解敏感,严重影响了聚类算法的质量。针对这一问题,提出了一种烟花搜索导向的多路启发式聚类算法。该算法通过多次调用经典启发式聚类算法,产生多个局部最优解;在搜索空间中以多个局部最优解为搜索起点,采用烟花搜索进行多路搜索;基于信息熵浓度设计烟花选择算子确定搜索方向;再经过变异、映射、偏移算子变换局部最优中心点,以发现质量更好的搜索起点;直至算法收敛获得新的搜索起点;最终以新的搜索起点调用经典启发式聚类算法获得高质量聚类结果。实验结果表明,烟花搜索导向的多路启发式聚类算法在不同数据集上的聚类质量明显高于对比其他聚类算法的聚类质量。更多还原

【Abstract】 Heuristic clustering algorithm has the advantages of fast convergence and easy implementation, but the initial solution is sensitive, which seriously affects the quality of clustering algorithm. In order to solve this problem, this paper proposes a multi-channel heuristic clustering algorithm for fireworks search guidance. The classical heuristic clustering algorithm is called many times to generate several local optimal solutions. In the search space, multiple local optimal solutions are used as the starting point, and fireworks search is used for multi-path search. Design fireworks selection operator based on information entropy concentration to determine the search direction; Then the local optimal center point is transformed by mutation, mapping and migration operator to find a better search starting point. Until the algorithm converges to obtain a new search starting point; Finally, the classical heuristic clustering algorithm is used to obtain high quality clustering results. Experimental results show that the clustering quality of FSG_MHC algorithm on different data sets is obviously higher than that of the contrast clustering algorithm.更多还原