CAREER: Harnessing the Positive Power of Negative Links for Network Analytics

Much of today's big data can be best represented as networks, which has led to the rise of network analytics for harnessing this inherent structure to best capture the data. However, most traditional network analytical methods have primarily been developed without consideration of modeling negative relations within the data. For example, online social media users can not only create positive friendships with one another, but they can also unfollow, block, and develop distrust or even animosity towards each other. The goal of this project is to provide dedicated advancements in network analysis methods built to fully harness not only the positive but also these negative relations in data, which are both desired and essential to better analyze the past and make predictions about the future. The results of this project will advance state-of-the-art network analysis methods having both a strong and immediate impact in multiple domains including in industry and academia, online social media and e-commerce. This project will play an integral role in introducing existing network analytics to students through developed undergraduate and graduate courses, while also providing first-hand research experiences through semester projects. Additionally, this project will support the mentorship of students through various university programs to leverage their unique strengths in our next generation of scientists and engineers. This project is focused on the advancement of methods, theories, models, and measurements, for networks with negative links, known as, signed networks. The project will provide dedicated efforts towards deeper analysis and better predictions on networks with negative links while spanning the four main pillars of network analytics. Specifically, novel network theories are proposed and built upon the complex positive/negative link interactions, dedicated signed graph neural networks will be developed for overcoming existing challenges in prior models, improved signed network modeling towards better synthetic generation and evaluation, and finally robust signed network measurements. The project is motivated by unprecedented opportunities to harness the additive power of negative links to improve network analytics, but also in solving the research challenges associated with mining and learning on signed networks. Each of the proposed research directions is positioned to advance network analytics through investigating original unexplored problems and developing novel theories and methodologies. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.