基于自适应差分演化算法的动态社区发现问题研究  

Research on dynamic community detection based on self-adapted differential evolution algorithm

作　　者：柴丹炜 刘建生[1] 

机构地区：[1]江西理工大学理学院,江西赣州341000

出　　处：《江西理工大学学报》2016年第5期79-86,共8页Journal of Jiangxi University of Science and Technology

基　　金：国家自然科学基金资助项目(61364015)

摘　　要：动态社区结构的发现问题已经成为复杂网络中重要的研究方向,其发现算法是一个十分关键的核心问题.为了提高对社区结构进行发现的准确度,提出一种基于差分演化思想的自适应调整差分演化动态社区发现算法.该算法通过最大化当前时间快照上聚类质量和最小化相邻快照间社区演化开销,得到以相邻时刻间网络结构差异度最小化的优化目标.使用自适应调整差分演化算法对人工动态网络实施有效划分,实验结果表明,该算法不仅收敛速度快,而且能够降低复杂网络中的社区结构发现的错误率.The dynamic community detection has been an important research direction of complex network and the detection algorithm is a crucial core issue. To improve the accuracy of the community detection result, a self-adapted differential evolution algorithm for dynamic community detection based on the theory of differential evolution is presented. The difference between the networks from one timeslot to successive one is the optimal objective of the algorithm, which has been realized by both maximizing the snapshot cost of the current timeslot and minimizing the temporal cost between neighboring timeslot. synthetic data set and real-life data set demonstrate that the efficiency and accuracy complex networks by using SDEDCD algorithm are higher than those obtained by the Experimental results on of community detection in other two algorithms.

关 键 词：动态网络 社区发现 差分演化 聚类质量 演化评价 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]