Facility Relocation Search For Good: When Facility Exposure Meets User Convenience Hui Luo Zhifeng Bao J. Shane Culpepper RMIT University, Australia RMIT University, Australia RMIT University, Australia hui.luo@rmit.edu.au zhifeng.bao@rmit.edu.au shane.culpepper@rmit.edu.au Mingzhao Li Yanchang Zhao RMIT University, Australia CSIRO, Australia mingzhao.li@rmit.edu.au Yanchang.Zhao@data61.csiro.au ABSTRACT In this paper, we propose a novel facility relocation problem where facilities (and their services) are portable, which is a combinatorial search problem with many practical applications. Given a set of users, a set of existing facilities, and a set of potential sites, we decide which of the existing facilities to relocate to potential sites, such that two factors are satisfed: (1) facility exposure: facilities after relocation have balanced exposure, namely serving equivalent numbers of users; (2) user convenience: it is convenient for users to access the nearest facility, which provides services with shorter travel distance. This problem is motivated by applications such as dynamically redistributing vaccine resources to align supply with demand for diferent vaccination centers, and relocating the bike sharing sites daily to improve the transportation efciency. We frst prove that this problem is NP-hard, and then we propose two algorithms: a non-learning best response algorithm (BR) and a rein- forcement learning algorithm (RL). In particular, the best response algorithm fnds a Nash equilibrium to balance the facility-related and the user-related goals. To avoid being confned to only one Nash equilibrium, as found in the BR method, we also propose the reinforcement learning algorithm for long-term benefts, where each facility is an agent and we determine whether a facility needs to be relocated or not. To verify the efectiveness of our methods, we adopt multiple metrics to evaluate not only our objective, but also several other facility exposure equity and user convenience metrics to understand the benefts after facility relocation. Finally, com- prehensive experiments using real-world datasets provide insights into the efectiveness of the two algorithms in practice. CCS CONCEPTS Information systems Information retrieval;• Computing methodologies Artifcial intelligence. KEYWORDS facility relocation, facility exposure, user convenience Corresponding author. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for proft or commercial advantage and that copies bear this notice and the full citation on the frst page. Copyrights for components of this work owned by others than the author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specifc permission and/or a fee. Request permissions from permissions@acm.org. WWW ’23, April 30–May 04, 2023, Austin, TX, USA © 2023 Copyright held by the owner/author(s). Publication rights licensed to ACM. ACM ISBN 978-1-4503-9416-1/23/04. . . $15.00 https://doi.org/10.1145/3543507.3583859 ACM Reference Format: Hui Luo, Zhifeng Bao, J. Shane Culpepper, Mingzhao Li, and Yanchang Zhao. 2023. Facility Relocation Search For Good: When Facility Exposure Meets User Convenience. In Proceedings of the ACM Web Conference 2023 (WWW ’23), April 30–May 04, 2023, Austin, TX, USA. ACM, New York, NY, USA, 11 pages. https://doi.org/10.1145/3543507.3583859 1 INTRODUCTION In this paper, we study the problem of facility relocation search, where facilities (and their services) are portable to provide higher service quality for social good [33]. The facility relocation prob- lem can provide support for a lot of domains, such as testing and vaccination for COVID-19 or similar pandemics that may occur in the future for medical health [38], and resource redistribution in vulnerable areas [37]. For example, COVID-19 checkpoints often need to be dynamically moved since new infection clusters can occur at diferent exposure sites. Since checkpoints have limited hu- man resources and cannot test a large number of people in a timely manner, it is important to let users travel fast to get services and meantime balance the number of people served by each checkpoint, which also refers to the concept of fairness among facilities [19]. A related application is the redistribution of medical resources (e.g., vaccines) in the vaccination sites, when the supply cannot meet the demand [12]. Since virus transmission changes rapidly and is often unpredictable, it is often urgent to relocate checkpoints or medical resources quickly. Another practical application example of this problem is to redis- tribute bike or scooter sites by bike sharing services, where bikes or scooters can be rented anywhere [16, 27, 46]. A company must fnd sites to distribute bikes and scooters daily, often based on de- mand [28, 45]. Providing a reasonable facility relocation strategy is important for the wellbeing of both service providers (i.e., facil- ities) and service recipents (i.e., users) [32]. We refer to this as a facility relocation problem, that is, a subset of existing facilities can be relocated to new locations. The benefts provided by the facility relocation are illustrated in the following example. Example 1. Fig. 1 illustrates an example before and after facility relocation. Initially, in Fig. 1a, there are three users ( 1 , 2 , and 3 ), two existing facilities ( 1 and 2 ), and two potential sites ( 1 and 2 ). The potential sites are idle locations for existing facilities to relocate to. Without loss of generality, users follow the closest proximity principle, which is widely adopted in the literature [8, 25, 36, 49]. It is clear that 1 is the nearest facility for all three users while 2 is unused. In Fig. 1b, when we move 2 to 1 , 1 (i.e., the raw 2 ) becomes the nearest facility of 1 and 2 . Based on Example 1, we can observe two immediate benefts for facility relocation: (1) Balanced exposure of each facility. We use 3937