共有信息引导的启发式聚类算法

启发式聚类算法采用局部搜索策略发现使得目标函数取极小值的聚类结果，即局部最优聚类结果。算法虽然具有收敛速度快等优点，但是初始解敏感问题严重地影响了聚类结果的质量。利用多个局部最优聚类结果中的共有信息设计启发式聚类算法。首先给出共有信息的定义及其发现算法FCI_G；然后利用共有信息设计启发式聚类算法CIGC；最后在多组仿真和实际数据集上考察了CIGC算法的性能。实验结果表明，共有信息对提高聚类算法质量有着显著的作用。

Common information guide heuristic clustering algorithm

Heuristic clustering algorithm generates the local suboptimal clustering results which make the objective function converge to local minimum with local search method.Although, the convergences speed of heuristic clustering algorithm is fast, but the initialization sensitivity problem make it cannot guarantee the quality of clustering results.In this paper, the com- mon information derived from several local suboptimal clustering results is used to design heuristic clustering algorithm.The common information definition and its finding algorithm, FCI_G is given; the common information is used to design algo- rithm CIGC;the efficient of CIGC is tested on several synthetic and real world data sets.Experiment results show that common information has significant efforts on improving the clustering results.