Algorithms for minimizing systemic risks in complex systems with guided network topology

Abstract

The topological characteristics of the underlying network of a system are very important to study and analyze the system. The mode of interaction among the components of a system may introduce many kinds of systemic risks. By controlling the formation and growth of the underlying network of a system, certain desirable properties may be attained and certain systemic risks may be minimized. Hence, various systemic risks of many real-world systems can be efficiently minimized by modeling the risk minimization problems as graph algorithmic problems. In this thesis work, we are mainly concerned with complex systems. A stock market and a humanitarian aid work are two such complex systems. Here, we have developed algorithms to proactively control the formation and growth of underlying networks of stock markets and humanitarian aid works to minimize certain systemic risks.

In this work, we have developed an algorithm to proactively generate a stock return correlation network of a stock market from the previous stock trading dataset to minimize the tendency of fatal random market crash, volatility and portfolio risk. We have conducted experiments on the Dhaka Stock Exchange trading dataset of the year 2015, 2016 and 2017 in this aspect. The resulting stock return correlation network is a scale-free network with the minimum structural entropy. We have used the topology of this network to recommend stocks for short term trading and portfolio scenario of a particular investment. We have also developed an algorithm to proactively generate the aid actors’ network of a humanitarian aid workflow from the aid dataset to minimize random failure, irrelevant interlinks, redundancy and communication delay. In this case, we have conducted experiments on the SIDR- 2007 dataset. The resulting aid actors’ network is a scale-free and small-world network with the minimum structural entropy.