A Classification Model for Rheumatoid Arthritis Patients Based on Rough Sets and Three-Way Multi-Attribute Decision Making

In clinical diagnosis, mainstream machine learning models often face common issues of incomplete data and multi-source heterogeneity when classifying patients. To address this, the paper defines distance measures for different attribute types in incomplete heterogeneous information systems, enabling effective handling of such information. First, based on the mixed distance between two objects, a binary relation in incomplete heterogeneous information systems was derived, and a multi-granularity rough set model for incomplete heterogeneous information was constructed. Second, on the basis of the multi-granularity rough set model, three-way multi-attribute decision-making was introduced, building a three-way multi-attribute decision-making classification model that considers risk costs and learning costs in the medical classification process. Finally, the proposed model was validated using real clinical data in the experimental section, demonstrating its effectiveness in classifying rheumatoid arthritis patients and improving classification accuracy.