Link Analysis in Complex Networks by Structural Method and Machine Learning

Document Type

Dissertation

Abstract

Systems biology and concept processing have been the center of attention for researchers to investigate the complexities of protein interactions and semantic associations. In this context, the link prediction problem has become crucial to reestablish lost connections and identifying patterns in complex networks. Although structural solutions to the link prediction problem were initially proposed for social networks, their applicability to other networks, such as biological and semantic networks, is still being determined. This study compares the link prediction performance of structural approaches and representation learning techniques in the formal classification task for different biological and semantic networks. The research findings reveal that representation learning techniques are more resilient to information leakage in these networks and can significantly outperform traditional structural indices for link prediction. Our study also highlights the impact of association strength on recovering the network structure in semantic networks. Specifically, we show that link prediction accuracy improves when considering stronger associations between words. This finding can affect various natural language processing tasks that rely on semantic association networks. Overall, this research contributes to the growing body of literature on complex network analysis by highlighting the effectiveness of representation learning techniques in predicting links in different types of networks. The results also suggest the importance of considering association strength in predicting links in semantic networks, which can have implications for various applications in natural language processing, information retrieval, and knowledge management.

Publication Date

6-2023

Comments

Thesis submitted to the Deanship of Graduate and Postdoctoral Studies

In partial fulfillment of the requirements for the M.Sc degree in Machine Learning

Advisors: Dr. Shangsong Liang, Dr. Bin Gu

Online access available for MBZUAI patrons

Share

COinS