Developing Theory and Application of Community Generation

摘要

The primary focus of this project was to develop a theory on community generation. Specific techniques as the Block Value Decomposition (BVD) and soft correspondence ensemble clustering frameworks, spectral relational clustering algorithm, and the general relation summary network model were designed and built. The BVD framework is a general framework for co- clustering dyadic relational data, which is a typical type of relational data in many applications. The soft correspondence ensemble clustering framework is a general framework for combining different clustering results together to deliver the optimal clustering result, which has many applications in distributed data mining and privacy-preserving data mining. The spectral relational clustering algorithm we have developed is a powerful relational data clustering algorithm that can be used for any type of relational data clustering. Finally, the relation summary network is the most general model that incorporates all the previous work as well as many existing models and algorithms in the literature which may be considered as the special cases of this model.