Crowdsensing: Assessment of Cognitive Fitness Using Machine Learning

Abstract

—The expanded use of smartphones and the Internet of Things have enabled the usage of mobile crowdsensing technologies to improve public health care in clinical sciences. Mobile crowdsensing enlightens a new sensing pattern that can reliably differentiate individuals based on their cognitive fitness. In previous studies on this domain, the visual correlation has not been illustrated between physiological functions and the mental fitness of human beings. Therefore, there exists potential gaps in providing mathematical evidence of correlation between physical activities & cognitive health. Moreover, empirical analysis of autonomous smartphone sensing to assess mental health is yet to be researched on a large scale, showing the correspondence between ubiquitous mobile sensors data and Patient Health Questionnaire-9 (PHQ-9) depression scales. This research systematically collects mobile sensors’ data along with standard PHQ-9 questionnaire data and utilizes traditional machine learning techniques (Supervised and Unsupervised) for performing necessary analysis. Moreover, we have conducted statistical t-tests to find similarities or to differentiate between people of distinct cognitive fitness levels. This research has successfully demonstrated the numerical evidence of correlations between physiological activities and the cognitive fitness of human beings. The Fine-tuned regression models built for the purpose of predicting users’ cognitive fitness score, perform accurately to a certain extent. In this analysis, crowdsensing is perceived to differentiate several people’s cognitive fitness levels comprehensively. Furthermore, our study has addressed a significant insights to assessing people’s mental fitness by relying upon their smartphone usage.