Abstract
A proliferation of wearable sensors that record physiological signals has resulted in an exponential growth of data on digital health. To select the appropriate repository for the increasing amount of collected data, intelligent procedures are becoming increasingly necessary. However, allocating storage space is a nuanced process. Generally, patients have some input in choosing which repository to use, although they are not always responsible for this decision. Patients are likely to have idiosyncratic storage preferences based on their unique circumstances. The purpose of the current study is to develop a new predictive model of health data storage to meet the needs of patients while ensuring rapid storage decisions, even when data is streaming from wearable devices. To create the machine learning classifier, we used a training set synthesized from small samples of experts who exhibited correlations between health data and storage features. The results confirm the validity of the machine learning methodology