A classifying ubiquitous clinic recommendation approach for forming patient groups and recommending suitable clinics

Abstract

Existing ubiquitous clinic recommendation systems assume that patients' preferences are alike and thus apply the same recommendation mechanism to all mobile patients. However, this assumption may be unreasonable and must be relaxed. Accordingly, this study proposes a classifying ubiquitous recommendation approach. The classifying ubiquitous recommendation approach divides patients into multiple groups by mining their unknown preferences before recommending them suitable clinics. To tune the recommendation mechanism of each patient group, an integer nonlinear programming problem is solved in a rolling manner. When a new patient accesses the ubiquitous clinic recommendation system, the recommendation mechanisms of all groups are applied to recommend the suitable clinic to the new patient. The results of a regional experiment indicated that the classifying ubiquitous recommendation approach improved the successful recommendation rate by up to 52%. Therefore, patients' unknown preferences are different and affect their behaviors in choosing clinics, which should be considered in grouping patients and in tailoring the clinic recommendation mechanism.