Abstract: In order to identify frequent co-occurrence communities in non-relational login log generated by Wireless City. Propose an undirected weighted graph-based community detection method called “Clique Search”. The algorithm maps the log data to graph, mining potential social communities by searching for the cliques in the graph. Compared to the traditional algorithm, this algorithm significantly reduces the computation time complexity and space complexity. The experimental results show that with the increase in the size of the input data set, the algorithm is still able to accomplish the computation in a constant level of time. Algorithm parameter [δ] influences the community tightness of the results obviously, by using different [δ] can meet the needs of different applications.

Key words: wireless city, community detection, Mac address, Apriori, clique search

摘要： 为了在缺失社交关系的无线城市接入日志中挖掘频繁共现的社团结构。提出了一种基于无向有权图的社团发现方法：团搜索（Clique Search）。该算法将日志数据映射到图空间，通过挖掘其中的团来对潜在的社团关系进行挖掘。相较于传统算法，该算法显著减小了运算时间复杂度与空间复杂度。实验结果表明，随着输入数据集规模的增长，该算法仍然能够在常数级别时间内完成计算。算法中的参数[δ]对结果社团中成员的联系紧密度影响比较明显，使用不同的[δ]值可以满足不同应用的需求。

关键词: 无线城市, 社团发现, Mac地址, Apriori算法, 团搜索