Algorithm Identification and Integrated with Push Service for Telemedicine System

Abstract

Telemedicine systems, while overcoming physical space constraints, often lack personalized interactions. By incorporating a push service and leveraging prediction-oriented algorithms, these systems can offer an improved user experience. Such enhancements enable timely treatment options and reduce unnecessary resource usage in on-site outpatient clinics. This research work starts by creating a robust algorithm using data mining techniques. Next, it establishes the foundation for a telemedicine push service. The service includes essential modules for disease differentiation, doctor recommendations, and diagnosis predictions. To optimize these modules, a merged algorithm combining k-nearest neighbor classification, nearest neighbor recommendation, and FP-growth is needed. This work aims to enhance treatment options for patients and streamline resource usage in on-site outpatient clinics. Moreover, this work has carried out empirical research for identification of algorithm by using available data at a public Chinese telemedicine system. The results of data analysis show the follows: 1. For disease diagnosis, the KNN model (k=1) is more accurate but less efficient, SVM and LibSVM are more efficient but less accurate than the KNN model; 2. In terms of doctor recommendation, nearest neighbor recommendation performs better but is not as efficient as matrix factorization; 3. in diagnostic prediction, the combination of introducing association mining and data segmentation can play a better role. The developed algorithm and its conclusions from this study could make easier and more efficient to provide treatment options for undecided-condition patients.