Principal Investigator: Dr Masato Abe
RIKEN, Chuo-ku, JapanTags: 46409, activity data, cognitive function, machine learning method, time-series analysis, wellbeing
Human behaviors emerge from interactions between brains and the environment. Thus, human daily activity can be associated with cognitive brain functions and subjective well-being. However, the relationship between them remains unknown. In this project, we will analyze accelerometer data, and reveal how the characteristics of activity data in daily life reflect cognitive functions and subjective well-being. To analyze time-series accelerometer data we will use statistical mechanics and machine learning method. The former method can give us a distribution of consecutive activity or inactivity in time-series. Previous studies revealed that the distribution followed a power law distribution and the exponents were different between healthy subjects and those with mental illness (Nakamura et al. 2007, Physical Review Letter; Sano et al. 2011, Plos one). The latter method is useful for predicting time-series data and outcomes. By predicting the future values from past data, we can characterize the predictability of human daily life and how complex the daily activity is. Finally, we will conduct statistical analysis for revealing the relationship between the parameters extracted from time-series data and other characteristics of subjects. The project duration will be 18 months. Quantifying how accelerometer data correlates with the subject’s daily life patterns and well-being can provide novel insights into human behavior research field and health care. The possible application includes finding biomarkers of human well-being only from the time-series of accelerometer data.