• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

计算机工程与科学

• 论文 • 上一篇    下一篇

求解最大团问题的混合修复遗传算法及其在社会网络中的应用

张素琪1,顾军华2,尹君2,郭京津2   

  1. (1.天津商业大学信息工程学院,天津 300134;2.河北工业大学计算机科学与软件学院,天津 300401)
  • 收稿日期:2015-09-18 修回日期:2015-12-02 出版日期:2016-12-25 发布日期:2016-12-25
  • 基金资助:

    天津市自然科学基金(15JCQNJC00600);天津商业大学青年科研培育基金(150115)

A genetic algorithm based on mixed repair for solving the
maximum clique problem and its application in social networks

ZHANG Suqi1,GU Junhua2,YIN Jun2,GUO Jingjin2   

  1. (1.School of Information Engineering,Tianjin University of Commerce,Tianjin 300134;
    2.School of Computer Science and Software,Hebei University of Technology,Tianjin 300401,China)
  • Received:2015-09-18 Revised:2015-12-02 Online:2016-12-25 Published:2016-12-25

摘要:

社会网络分析是数据挖掘中与社会生活联系最紧密的热点之一,凝聚子群分析是一种典型的社会网络子结构分析方法,其中最大团结构是关系最紧密的凝聚子群,最大团问题的研究在社会网络分析中有重要意义。针对遗传算法在求解最大团问题中运行时间长、部分基准图例求解精度不高等问题,提出了一种基于混合修复策略的遗传算法MGA。MGA算法融合度修复和随机染色体修复方法并结合随机配对的精英选择、均匀块交叉和倒位变异算子,可以有效避免算法陷入局部最优,在加快收敛速度和丰富种群多样性方面有明显效果。算法在DIMACS基准图例和典型的社会网络实例上进行了测试,实验结果表明MGA算法具有较好的求解精度和较快的收敛速度。

关键词: 社会网络分析, 最大团, 遗传算法, 混合修复策略, 精英选择, 均匀块交叉, 倒位变异

Abstract:

Social network analysis is a hot spot in data mining and it also has a very close contact with us in social life. The structure of the maximum clique, a typical social network substructure analysis method, is the most compact cohesive subgroup, so research on the maximum clique problem (MCP) has great significance in social network analysis. Aiming at the defects of the genetic algorithm (GA) for the MCP, such as longrunning, poor generality and low precision in some benchmark graphs, we propose a new genetic algorithm based on mixed repair, called MGA. A new mixed method based on degree and random chromosome repair, combing with the elitist selection based on randomly paire, uniform block crossover and inversion mutation operator, is adopted in the MGA, which can accelerate search speed, increase population diversity, and effectively prevent the algorithm from trapping into local optimum. We test the proposed algorithm on DIMACS benchmark graphs and typical social network instances. Experimental results show that the MGA has better performance and high generality, can more effectively resolve the MCP in social network analysis.
 

Key words: