Matching Problems in Refugee Resettlement

该项目将促进科学的进步，并通过推进分析决策工具来应对美国难民重新安置的业务挑战，为国家繁荣和福利做出贡献。重新安置的目标是使难民逐步融入东道国社会，同时兼顾社区的局限性和难民的需要。这项研究将通过使用包括机器学习和数学优化在内的分析方法来增强目前的人工难民重新安置决策。这些技术将改善人道主义决策，并有可能改变国内和全球重新安置决策的方式。收容社区将受益于难民的融合，为匹配的社区带来新的技能、青年和多样性。与美国难民组织建立的合作将指导研究，并使开发的模型在现实世界中得到验证。该项目将做出两个主要的方法学贡献。首先，它将探索积分幺半群的代数和几何性质，以更好地理解和利用它们的结构。整幺半群在容量灵活、目标多样的情况下表现出优越性。这项研究将开发适当的算法和数据结构，以有效地和明智地编码和检索幺半群信息，并将包括算法分析，以确保计算的易处理性。如果成功，这项研究将有助于优化方法的新进展，特别是积分幺半群的算法使用解决其他硬线性和非线性匹配，背包，广义分配和包装问题。其次，这项研究将通过利用现有难民安置和融合（结果）数据的监督机器学习技术来构建新的目标函数。这些新的目标函数将指导寻求更成功的重新安置结果。预测模型还将揭示人口和区域因素如何促进难民融合的以前未被发现的见解。该奖项反映了NSF的法定使命，并已被认为是值得通过使用基金会的智力价值和更广泛的影响审查标准进行评估的支持。

This project will promote the progress of science and contribute to the national prosperity and welfare by advancing analytical decision tools tackling the operational challenges of refugee resettlement in the United States. The goal of resettlement is to progressively integrate refugees into host societies, while balancing limitations of communities with the needs of refugees. This research will augment current manual refugee resettlement decision-making by using analytical methods that include machine learning and mathematical optimization. These technologies will improve humanitarian decision-making and have the potential to transform how domestic and worldwide resettlement decisions are made. Host communities will benefit by integrating refugees that bring new skills, youth and diversity to the matched communities. An established collaboration with a US-based refugee organization will guide the research and enable validation of the developed models in a real-world setting.This project will make two main methodological contributions. First, it will explore the algebraic and geometric properties of integral monoids to better understand and capitalize on their structure. It is believed that integral monoids can excel in contexts with flexible capacity and multiple objectives. This research will develop appropriate algorithms and data structures to efficiently and judiciously encode and retrieve monoid information, and will include algorithmic analyses to ensure the computational tractability. If successful, this research will contribute to new advances in optimization methodology, specifically the algorithmic use of integral monoids to solve other hard linear and nonlinear matching, knapsack, generalized assignment, and packing problems. Second, this research will construct novel objective functions by leveraging supervised machine learning techniques on existing refugee placement and integration (outcome) data. These new objective functions will guide the search toward more successful resettlement outcomes. The predictive modeling will also reveal previously undiscovered insights into how demographic and regional factors contribute to refugee integration.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.