Authors: Diep Dao*, University of Colorado - Colorado Springs, Craig Ravesloot, Rural, The Rural Institute, University of Montana, Lillie Greiman, The Rural Institute, University of Montana, Tannis Hargrove, The Rural Institute, University of Montana
Topics: Geographic Information Science and Systems
Keywords: spatial-temporal data mining, sequential patterns, daily mobility, daily activities, health, GEMA, GPS
Session Type: Paper
Presentation File: No File Uploaded
Geographical explicit ecological momentary assessment (GEMA) data collection platform provides extremely rich geospatial datasets capturing near-real-time interactions between human participants and the environment. This helps to significantly gain behavior insights linking daily mobility, activities, and health. However, the task of analyzing these datasets effectively is not straightforward because of large multivariable dimensions and ill data formats. This work presents challenges and an effective solution toward handling these GEMA datasets. Our data on activities, mobility, and health responses (both physical health like pain and psychological health like happiness) was collected on 91 human individuals over 14 days at 5-minute and 2-hour frequencies, respectively, for location data and qualitative EMA responses. Our work utilizes a data mining approach to extract frequent spatio-temporal sequential patterns linking daily activities, daily mobility, and health. A frequent spatio-temporal sequential pattern in this context is defined to be a list of time-sampled locations, activities, and health responses that frequently occurs within the dataset. This inductive mining approach works robustly with large and complex – both in dimension and format – datasets. Challenges and a potential solution to visualizing the extracted sequences will also be presented.