원문정보
초록
영어
Graph mining is a dynamic and active research area. In recent years, there is a remarkable boost in graph-structured data resulting graph mining a serious topic in research community. Graph clustering is the process of identifying similar structures in a large set of graphs. Graph clustering is also known as graph partitioning or grouping. This problem plays an important role in various data mining applications. Traditional approaches are centric towards optimization of graph clustering objectives such as ratio association or normalized cut. Spectral methods are also introduced which required Eigen-Vector computation. However these techniques are slow. We have presented a novel algorithm for detecting closely related groups of graph structures in KEGG metabolic pathways. The technique is based on structural similarity of connected fragments in graph-structured data. The technique is scalable to directed as well as undirected graphs. Preliminary experiments with synthesized data collected from KEGG were performed and their results are reported. The second contribution of this study is the modeling and analysis of combined metabolic reaction networks and relation network and showing their behavior towards scale free network.
목차
1. Introduction
2. Graph Theory Preliminary
3. Graph Modeling for Metabolic Pathways
4. Clustering graph-structured Data
4.1. Step 1: Data Assignment
4.2. Step 2: Relocation of “Means”
5. Literature Review
6. gMean: Proposed Framework
6.1. Step-1 Training Graphs Data
6.2. Step 2: Graph Modeling
6.3. Step-3: Clustering
6.4. Step-4: Pruning
6.5. Step-5: Results
7. Results and Discussion
8. Conclusion
References