What drives matching in online medical consultation services? A data-driven approach to learn two-sided preferences
Online medical consultation is rapidly expanding worldwide. This paper investigates doctor-patient matching in an online setting, a typical example of two-sided matching. We first evaluate distance metrics to ensure that attribute differences between doctors and patients are accurately captured. We then apply a multi-attribute value function to aggregate these differences into a comprehensive satisfaction level. Finally, we propose a data-driven preference estimation framework that jointly optimizes efficiency, stability, and fairness of matching. Using historical matching data obtained from the Haodf.com platform, together with preferred attribute values constructed under explicitly specified preference generation assumptions, we estimate the attribute weights of both doctors and patients that are consistent with the observed matching results. Across four assumed scenarios, the estimated weight patterns remain relatively stable and provide better predictive performance than an equal-weight benchmark. Our method improves average prediction accuracy from 83.25% to 92.17% across the four scenarios, corresponding to an increase of 8.92 percentage points. These findings offer actionable insights for designing more efficient and sustainable online healthcare services.
Authors
- Zhe George Zhang
- Xi Chen
- Wenbo Zhang
- Ilhyung Kim
Publication Details
- Journal
- Figshare
- Published
- 2026-11-15
- DOI
- https://doi.org/10.6084/m9.figshare.35013929
- Primary Topic
- Game Theory and Voting Systems
- Type
- article
- Field-Weighted Citation Impact
- 0.00