基于无偏Q值反馈的社区划分算法

Community division algorithm based on feedback of unbiased Q value

在分析现有社区划分算法的基础上,针对当前算法Q值有偏及权值未体现等缺陷,提出了一种基于无偏Q值反馈的社区划分算法.该算法首先利用传递权值计算出节点间的相似度;然后,采用随机游走策略确定最优社区数,以解决现有划分算法中Q值有偏的问题;最后,在最优社区数确定的情况下,利用划分结果评价Q值反馈更新信息素矩阵以驱动后续的划分,从而达到快速收敛的目的.针对计算机构造的数据集以及实际网络的实验分析结果表明,与现有算法相比,该算法在社区划分方面具有更高的准确率及更快的收敛速度,能够达到社区划分以及核心节点发现的目的,可被推广应用至移动社会网络模型的建立中.

Based on the analysis of the existing community division algorithms,a community division algorithm based on the feedback of unbiased Q value is proposed to avoid the bias of Q value and the ignorance of link weight ignored.Firstly,the similarity of nodes is calculated through the transitive weights.Then,the optimal community number is determined by means of random walk strategy in order to make the Q value unbiased.Finally,with the optimal community number determined,the pheromone matrix is updated by using the feedback of the Q value to drive up the follow-up division,which makes the algorithm converge to the optimal solution quickly and accurately.The experimental analysis results of data set constructed by computer analysis and the actual network show that this algorithm has higher accuracy and faster constringency than the existing community division algorithms in respect of community division and core node detection.It can be used to construct the social mobile network.