Two-sided online bipartite matching in spatial data: experiments and analysis

Abstract

With the rapid development of sharing economy and mobile Internet in recent years, a wide range of applications of the T wo-sided O nline B ipartite M atching (TOBM) problem in spatial data are gaining increasing popularity. To be specific, given a group of workers and tasks that dynamically appear in a 2D space, the TOBM problem aims to find a matching with the maximum cardinality between workers and tasks satisfying the spatiotemporal constraints. Many works have studied this problem, but the settings of their problems are different from each other. Moreover, no prior works have compared the performances of the algorithms tailored for different settings under a unified definition. As a result, there lacks a guideline for practitioners to adopt appropriate algorithms for various scenarios. To fill the blank in this field, we present a comprehensive evaluation and analysis of the representative algorithms for the TOBM problem in this paper. We first give our unified definition and then provide uniform implementations for all the algorithms. Finally, based on the experimental results on both synthetic and real datasets, we discuss the strengths and weaknesses of the algorithms in terms of short-term effect and long-term effect, which can be guidance on selecting appropriate solutions or designing new methods.