A1365
Title: Covariate assisted graph matching
Authors: Jesus Arroyo - Texas A&M University (United States) [presenting]
Trisha Dawn - Texas AM University (United States)
Abstract: Data integration is essential across diverse domains. A crucial initial step in this process involves merging multiple data sources based on matching individual records. When the datasets are network data, this problem is typically addressed through graph matching methodologies. For such cases, auxiliary features or covariates associated with nodes or edges can be instrumental in achieving improved accuracy. However, most existing graph matching techniques do not incorporate this information. To overcome these limitations, the aim is to propose two novel covariate-assisted seeded graph matching methods. The first one utilizes the quadratic assignment problem (QAP), while the second one leverages the local neighborhood structure of non-seed nodes to guide the matching process. Theoretical guarantees are established for model estimation error and exact recovery of the solution of the QAP, demonstrating perfect alignment accuracy with high probability under sufficient signal strength. The effectiveness of the methods is demonstrated through numerical experiments. Finally, the proposed approach is applied to match two real-world networks. By leveraging additional covariates such as institution, country, and graduation year, improved alignment accuracy is achieved. The power of integrating covariate information is highlighted in the classical graph matching setup, offering a practical and improved framework for combining network data with wide-ranging applications.
