EURO 2024 Copenhagen
Abstract Submission

EURO-Online login

4258. Attaining Approximate Stability in Large Two-Sided Matching Markets

Invited abstract in session MD-43: Simulation in economics II, stream Agent-based Models in Management, Economic and Organisation Sciences.

Monday, 14:30-16:00
Room: 99 (building: 306)

Authors (first author is the speaker)

1. Chun Wang
Concordia university
2. Jie Gao
Transport & Planning, Delft University of Technology (TU Delft)
3. Shixuan Hou
Concordia University

Abstract

A two-sided market is an economic platform that enables direct interactions between two distinct customer groups, providing each other with network benefits. Matching is the fundamental function that ensures the viability and service quality of such markets. While the Gale-Shapley algorithm and its variants can be used to compute stable matchings for two-sided markets effectively, these methods typically require participants to rank all potential partners, posing a significant cognitive burden to them and rendering it impractical for large-scale markets involving hundreds or thousands of options on each side. This paper presents an efficient framework for computing approximately stable matchings in large-scale and dynamic market settings. Instead of explicitly eliciting preferences, we employ probabilistic choice behavior models trained on participants’ historical interaction data with the platform. These models serve as incremental preference elicitation tools, which adapt themselves through continuous machine learning techniques. Additionally, we devise procedures for robust matching in the presence of probabilistic and evolving preference information provided by the behavior models. We conduct a simulation study to experimentally quantify the impact of enforcing approximate stability within a dynamic crowd-sourced delivery services setting, utilizing preference survey data collected from crowd-sourced drivers.

Keywords

Status: accepted

Back to the list of papers