Multiobjective Optimal Control Methodology for the Analysis of Certain Sociodynamic Problems
Abstract
Social networks involve studying how relations form between individuals in a group based on their shared preferences and attributes. In this work, social force theory is used to model social interaction and long-term network dynamics while multiobjective optimal control theory provides a basis for predicting network structural formation. Using computer simulations, the author numerically analyzes the evolution and long-term behavior of optimal network structures based on the demographics of a small data set. She pays special attention to the effect that memory has on relationship choices, especially clique formation. After obtaining a snapshot of the network structure, she turns her attention to a very common task involving social networks: missing link prediction. The link prediction problem can be described as uncovering hidden or missing connections between nodes in an observed network. There are several link prediction methods in the literature that rely heavily on network topology to predict links and are primarily used on collaboration networks like e-mail or coauthorship networks. The problem with these networks is that they offer no qualitative information on nodal attributes. In this work, the author presents a new model for link prediction that is centered on nodal attributes and uses social force theory to provide behavioral rules for nodal interaction. The model for link prediction starts with the observed network structure and parameters derived from the multiobjective optimal control approach. She uses the known relationship patterns in the network to construct a new multiobjective optimal control problem constrained in a way that reproduces those existing relationships in the network. She then remove a random subset of links from the network, treating them as missing or hidden. Finally, she solves this new constrained multiobjective optimal control problem to reproduce existing links while uncovering the missing links in the process.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 2009
- Accession Number
- ADA496262
Entities
People
- Gloria L. Porter
Organizations
- North Carolina State University