Structuring big data using principal components analysis and decision rules

Caregivers for people with dementia (PWD) face personal challenges every day. Caregivers must take care of themselves and provide support to the PWD. Agitation is one of the most common reasons caregivers experience increasing amounts of burden, stress, and depression. The Behavioral and Environmental Sensing and Intervention (BESI) study focuses on observing the influences of environmental factors on agitation in the home of the caregiver and person with dementia (aka, dyad). In this study, we use a remote ethnographic method to collect environmental and daily activity data from the caregiver over a thirty-day period. A large data set resulted from the ethnographic method requiring significant structuring and organization to reduce the data set to key parameters for use in modeling. In this paper, we present a framework for data structuring and reduction based on principal components and a decision rule. The process and framework will be described. Additionally, the contribution of each factor to the overall principal components will be discussed as well as the decision rule used to determine which components to retain for further modeling for early detection of agitation in dementia.