Burstiness in human behavior has been observed at various levels, from individual behavior to collective social phenomena. Identifying the non-trivial mechanisms that cause burstiness is a crucial issue for understanding the universality of human behavior. Many studies on human behavior use web-based data and show burstiness in various contexts. In such cases, the state and attributes (e.g., demographic information) of the individuals are often unknown. On the other hand, there are still few studies on human behavior using physical activity data, where the individual’s state and attributes are clear. Previous research on human physical behavior has shown that the shape of the distribution of inter-event times (IETs) changes with mental illness. The burstiness of the activity may depend on the various states of the individual. Thus, investigating how burstiness is affected by an individual’s state and attributes will help to clarify the mechanism of the universality of human behavior and has possible medical implications.
Therefore, in this study, we measured human physical activities using accelerometers, along with recording the subjects' states (e.g. active or resting) and investigated those burstiness and temporal correlation. Here, subjects are infant (2-month-old), children (3-year-old), and adults.
We find that children showed the lower scaling exponent in cumulative distribution of IETs. Also, some of the physical activity data show 1/f noise, which has also been identified in the brain activities. Finally, we also discuss the point processes that may generate our observed features.