Egocentric Information Abstraction and Visualization for Heterogeneous Social Networks

計畫名稱:Egocentric Information Abstraction and Visualization for Heterogeneous Social Networks

所屬單位:資訊系

研究團隊:機器發明與社群網路探勘實驗室

計畫主持人:林守德

研究人員:黃安達

資源需求:C, C++, Java, Python, Perl, Matlab, MPI

使用期間:2009/03~

研究主題:
Egocentric Information Abstraction and Visualization for Heterogeneous Social Networks

研究內容概述:
Social Network is a powerful representation and visualization schema that allows the depiction of the relationships information between entities. However, for real-world tasks, the constructed heterogeneous networks are usually too complex for users to perform advanced investigations. In this paper, an unsupervised mechanism is proposed for egocentric information abstraction and visualization in heterogeneous social networks. Our abstraction consists of two levels. The first level of abstraction is a summarization process that maps the egocentric heterogeneous network onto a vector-space domain by identifying linear combination of link types as features and computing several statistical dependencies as feature values. The second level of abstraction focuses on using four diverse abstraction criteria to distill representative and/or informative messages, and use them to reconstruct the abstracted networks for visualization. The evaluations were conducted on a real world movie dataset and an artificial crime dataset. The experimental results not only demonstrate the abstracted networks but also show that such abstraction and visualization can facilitate more accurate and efficient crime investigation for human subjects.

詳細計畫內容 回到上一頁