Date: Wednesday March 18th 2020

 

Time:  9:30 am to 12:30 pm

 

Location: Lambton Tower 3105

 

 

Identifying successful teams for completing given tasks under some constraints is a challenging process and proves to be an NP-Hard problem. In this dissertation, we model a computational framework to discover teams of experts in various applications and predict the potential for collaboration in the future from a given Social Network. Inspired by an evolutionary search technique using a higher-order cultural evolution, a framework is proposed using Knowledge-Based Cultural Algorithms to identify teams from co-authorship, industrial settings, and healthcare settings. We also propose a unified framework by incorporating several required cost functions as a multi objectives optimization framework. 

 

To predict future interaction with the teams, we analyze the temporal behaviors of the vertices in a dynamic social network and employ Multi-Layer Perceptron as a classifier to predict the link formation. Additionally, to reduce the complexity of the prediction process, we examine the benefits of extracting subgraph of a target link and its feature for future interaction. We use Convolutional Neural Networks to extract higher-level features of subgraph efficiently and Long Short-Term Memory neural networks to learn long-range dependencies of sequential data and capture the evolving patterns of the subgraph in the dynamic networks.

 

Keywords: Social Network Analysis, Team formation, Link Prediction, Cultural Algorithm, Deep learnings

 

 

Internal Reader: Dr. Saeed Samet

 

Internal Reader: Dr. Sherif Saad Ahmed

 

External Reader: Dr. Kathryn Pfaff

 

External Examiner: Dr. Hamid Usefi

 

Advisor(s): Dr. Ziad Kobti and Dr. Mehdi Kargar

 

Chair:    TBA

 

5113 Lambton Tower 401 Sunset Ave. Windsor ON, N9B 3P4 (519) 253-3000 Ext. 3716 csgradinfo@uwindsor.ca