Visual Characterization of Gathered Data for Digital Phenotyping

Digital phenotyping is the collection of data from mobile device sensors to use in the medical diagnosis of mental disorders. Data is collected from thousands of people to obtain meaningful patterns, requiring the opinion and validation of specialists before having information that allows the prognosis of new patients. This paper is part of a digital phenotyping project to obtain patterns for use in diagnosis. During data collection, some problems have been detected in the sensors, battery consumption, and people’s disinterest, among others. The goal of this work is to detect problems during the data collection phase to maximize usable data. The study intends the collection of accurate and usable digital information to support the data mining process. The programs that have been implemented make it possible to detect devices that produce erroneous or incomplete data so that a person can talk to the user to correct the problem or remove them from the data collection group. Some graphs are shown to exhibit the difference in data collected.