| Human space-time activities are dynamic with many interacting dimensions. The analysis of human space-time activity patterns plays important roles for the understanding of many phenomena, such as urban sprawl and growth, transport planning and management, locating new facilities, disease outbreak control, etc. Understanding the dynamics of human space-time activities in human and environment systems, reasoning their relationships, modeling spatio-temporal behavior, and predicting future changes are essential for many people-based applications. In transportation research community, the analysis of space-time activity patterns has been extensively studied, but some issues have not been sufficiently addressed so far. First, space and time are two fundamental dimensions that are important for the understanding of spatiotemporal activity patterns. However, they are often analyzed separately, which makes the interpretation of space-time interaction difficult. Second, human activities take place in the real world, but few of the existing research can handle activity pattern with real geographic locations that are important to reason the space-time relations. Recently, GIS-based time-geographic analysis has been proposed to analyze human space-time activity patterns and it proves to be an efficient to understand complex patterns with real geographic locations. However, the number of individuals can be analyzed by this method is still limited, which limits broader applications. In this research, group pattern analysis based on GIS-based time-geographic analysis has been proposed based on activity pattern classification and geovisualization. This research aims to solve the following key problems in the classification of human space-time activity patterns. First, what kind of mathematic measures can be developed to compare the differences between space-time activity patterns with many interacting dimensions? Second, based on the time-geographic approach, how to analyze and classify the space-time trajectories when a large number of individuals are involved? Third, what kind of characteristics can help to explain the observed activity patterns? In this study, a multi-objective evolutionary algorithm has been developed to measure the dissimilarity between space-time trajectories and a k-medoids clustering algorithm has been utilized for activity pattern classification. Each pattern group and their representative patterns can be presented in a 3D GIS environment for explorative spatial data analysis. Statistical analysis shows that extracted pattern groups are correlated with socio-demographic and behavioral variables. |
