S-Space College of Engineering/Engineering Practice School (공과대학/대학원) Dept. of Electrical and Computer Engineering (전기·정보공학부) Theses (Master's Degree_전기·정보공학부)
Node Splitting Based Overlapping Community Detection Framework in MapReduce : 맵리듀스에서의 노드 복제 기반 중복 커뮤니티 구조 검출 프레임워크
- Namyoon Kim
- 공과대학 전기·컴퓨터공학부
- Issue Date
- 서울대학교 대학원
- 학위논문 (석사)-- 서울대학교 대학원 : 전기·컴퓨터공학부, 2015. 2. 김형주.
- This paper proposes implementations of community structure detection algorithms in MapReduce, a parallel programming framework. One of the community detection algorithms we have implemented in MapReduce is the Peacock algorithm, which pre-filters nodes likely to be a member of more than one community. These nodes are copied by the number of communities they might belong to. The actual communities are found in the next step, and Peacock algorithms modular nature allows for its use with other community detection algorithms, specifically those that cannot detect overlapping communities. We show performance differences of two disjoint community detection algorithms used in tandem with Peacock: the Girvan-Newman and fast unfolding algorithms. We demonstrate on a real world social network dataset that MapReduce increases scalability of the algorithms.