A Methodology to Rank the Patients Prone to Chronic Diseases in Telehealth

Context: From last decade, telehealth research community have started to use machine learning techniques to automate the certain aspects of healthcare systems for prediction of patients prone to chronic diseases such as checking of insulin level, cancer, and monitoring blood-glucose levels. Though existing automated systems aid the telehealth practitioner, however high computing effort, time, and cost is required in terms of several classifiers, multi class problem, or large sample size to obtained results on the basis of ground reality. Goal: We work on this concern and propose a methodology which is employed via "Learning to Rank" (i.e., a supervised learner) to rank the patients on the basis of their data relevancy. We also propose a model to investigate the efficacy of the proposed method. We assess the efficacy of the proposed method with seven public datasets associated with certain diseases. Datasets are retrieved from UCI repository Machine Learning Repository. Results: We observed LambdaMART outperform than AdaRank and Coordinate Ascent with a minor difference in terms of NDCG (Normalized Discounted Cumulative Gain) and MAP (Mean Average Precision) to predict patients prone to chronic diseases. Conclusion: The results of study indicates the usage of the proposed method as a recommendation system to rank the patients on the basis of their severity level.