Research

Scenario Clustering and Reformulation Techniques for Multi-Stage Network Interdiction

In the Human Trafficking Interdiction with Decision-Dependent Success paper, we showed the computational difficulties of solving network interdiction models over multiple stages. In this work, we investigate the properties of our problem to solve larger instances. We reformulate the model as a network flow problem on a binary scenario tree and then fathom the scenario tree to reduce the problem size. We show that this approach preserves optimality and reduces the problem size. Initial experiments show reductions in problem sizes of around ~60% for grid-like networks. Our approach can potentially be carried over to other decision-dependent network design problems including energy networks.

Preprint: Coming Soon

Code: Coming Soon

Network Interdiction Models for Disrupting Human Trafficking

The open border between Nepal and India and the NGO disruptions to human trafficking in this region motivated us to design a network interdiction model that considers traffickers adapting to law enforcement/NGO interdictions. We capture that identifying trafficking on an arc is probabilistic. We use a multi-stage model to model how traffickers and law enforcement/NGO decisions can alter these probabilities over time.

Our main finding is that the capability of adapting for a trafficker drastically changes interdiction decisions and strategies over time. So, decision-makers should invest in understanding how the traffickers can adapt to current strategies and adjust their strategy over time to combat adaptation. These findings are similar in the spirit of displacement theory from criminology and the balloon effect from drug policy studies.

Manuscript links:

Socio-Economic Planning Sciences - https://www.sciencedirect.com/science/article/pii/S0038012123000149?dgcid=coauthor
Preprint - https://engrxiv.org/preprint/view/1068/

A presentation for this work from INFORMS Annual Conference 2020. I have also presented this at the INFORMS Security Conference 2022.

Systems Models for Disrupting Sex Trafficking

Trafficking research shows a relationship between vulnerabilities, such as homelessness or addictions, and sex trafficking recruitment. In this study, we teamed up with trafficking researchers and human trafficking survivors to understand the mechanisms behind sex trafficking recruitment and vulnerabilities. We first propose a Markov Chain model to represent the vulnerability and sex trafficking recruitment dynamic. However, due to the inherent uncertainty in transition probabilities, we need to consider a set of Markov Chains. On top of these Markov Chains, we consider community-based resource allocation decisions to disrupt trafficking recruitment.

System's Perspective Video: Coming Soon

Preprint: Coming Soon

Below is a presentation from the initial stages of this project from INFORMS Annual Conference 2021. I have also presented parts of this work at IISE Annual Conference 2022 and INFORMS Annual Conference 2022, both in person.

Page updated

Google Sites

Report abuse