Assessing Data Traffic Classification to Priority Access for Wireless Healthcare Application

Abstract

New healthcare applications rely on wearable devices to collect and send patient's physiological and biomedical information to cloud servers, which executes most of the data processing and analysis. However, healthcare applications have strict requirements of packet delivery and traffic latency to ensure accurate information for medical/healthcare purposes. In this paper, we introduce a device management system, called MAESTRO, to improve the Quality of Service (QoS) for healthcare applications. The MAESTRO system combines a machine-learning traffic classification with a prioritization algorithm to provide a required transmission priority for physiological data. We set up the machine learning module in the R language, using the algorithms in the caret package. We implemented and simulated the prioritization algorithm in NS-3, in a scenario where wearable medical devices share network access with generic stations. Results confirmed the machine learning module achieved 91.5% of accuracy when identifying the physiological data and assigning the expected priority. Further, MAESTRO reached 60% of improvement in the packet delivery ratio for physiological data, in a scenario with a variable number of devices and stations.