Integrating Network and Community Discovery for Mining Constant and Outlier Nodes

Information Retrieval and Extraction Major Project 2016

Integrating Network and Community Discovery for Mining Constant and Outlier Nodes

In real world, we often don't have access to the entire network of nodes and edges. This motivated researchers to propose various link prediction and community detection algorithms. A lot of work has been done on network discovery(link prediction) and community detection individually. However, recently, there has been an attempt [1] to integrate these two problems hoping that each of these would help the other.

During the network discovery and community detection process, it is possible that certain nodes would rapidly change the communities as we discover the network while certain nodes would remain constant. The goal is to find the characteristics of these nodes and understand why are they constant (or rapidly changing).

The following are the different stages of the project:

• Understanding [1] end to end, demonstrated through a viva or presentation .
• Implementation of the algorithm and reproduce the results mentioned in [1]
• Proposing methods to detect constant nodes/rapidly changing nodes while detecting network and communities.
• Studying the characteristics of these nodes.

[1] http://hanj.cs.illinois.edu/pdf/wsdm15_jliu.pdf