Detecting local communities within a large scale social network using MapReduce

作者：Yin Hongjun; Li Jing; Niu Yue

来源：International Journal of Intelligent Information Technologies, 2014, 10(1): 57-76.

DOI：10.4018/ijiit.2014010104

• Community detection
• Design and implements
• Interested community
• Map-reduce programming
• Network partitioning
• Partition algorithms
• Personalized PageRank
• Stationary distribution

Social network partitioning has become a very important function. One objective for partitioning is to identify interested communities to target for marketing and advertising activities. The bottleneck to detection of these communities is the large scalability of the social network. Previous methods did not effectively address the problem because they considered the overall network. Social networks have strong locality, so designing a local algorithm to find an interested community to address this objective is necessary. In this paper, we develop a local partition algorithm, named, Personalized PageRank Partitioning, to identify the community. We compute the conductance of the social network with a Personalized PageRank and Markov chain stationary distribution of the social network, and then sweep the conductance to find the smallest cut. The efficiency of the cut can reach O(√αlog|S|). In order to detect a larger scale social network, we design and implement the algorithm on a MapReduce-programming framework. Finally, we execute our experiment on several actual social network data sets and compare our method to others. The experimental results show that our algorithm is feasible and very effective.

• 出版日期2014
• 单位中国科学技术大学