PerSE: Visual analytics for identifying periodicity in spatio-temporal event data
ISBN 978-85-88783-11-9
Authors
1Swedberg, B.
1THE PENNSYLVANIA STATE UNIVERSITY Email: bws180@psu.edu
Abstract
Although time has many linear qualities, much of our everyday behavior is governed by periodic events. Periodic events are events that recur at regular intervals, such as religious holidays, the workweek, rush hour, rise and fall of the tide, and seasonal changes. These events are embedded in every society, and are of paramount importance because of their correlative societal behavior. This includes indirect reactions that may not be obvious or well known. For instance, the holy-day of Ashura causes tension, and sometimes violence, between Shia and Sunni Muslims because it highlights their religious differences. During the Chinese New Year, thieves target Asian New Yorkers because cash-gifts play a role in traditional Chinese New Year celebrations. The success of military operations may hinge on the nighttime illumination (i.e. the lunar cycle) given a certain terrain and technological assets. In all three cases there are spatial, temporal, and attribute components related to the event. For the purpose of the current context I refer to the knowledge of another society’s periodic events as cross-cultural knowledge. As it is imperative for intelligence analysts to establish and maintain a deep cultural knowledge of their analytical domain, it is crucial to understand the spatial, temporal, and attribute makeup of periodic events. If properly understood, analysts can identify normal behavior, given a time and location, and subsequently predict future events. Knowledge of normal behavior also allows an analyst to uncover abnormal or unusual behavior. However, incomplete or inaccurate cross-cultural knowledge hinders the ability to determine normal behavior regarding periodic events. Some of these hindrances include disparate calendars (e.g. Islamic vs. Gregorian), localized idiosyncrasies, and varying cultural context all of which may take years of experience and training to fully understand. Current research provides tools that address spatio-temporal periodicity, but these tools are limited by the presumption that events are structured via the Gregorian calendar. Thus, a significant amount work would be required to analyze data through the lens of multiple calendar systems at changing temporal or spatial scales. To address this, I am building PerSE (Periodicity in Spatio-temporal Events): a web based visual analytics application supporting the analysis of periodic events in categorical spatio-temporal event data. PerSE contains four interactive coordinated views: (1) a seasonal subseries plot, with built-in support for multiple calendar systems to identify periodicity (2) a similarity matrix to provide an overview on periodicity strength, (3) a map to provide spatial context and navigation, and (4) a selection/filter layer management system. The ultimate goal of PerSE is to provide a flexible user interface that will support the identification of and improve understanding on periodic events. Further work includes a usability study. The goals of this study are to examine the tool’s learnability, the user’s ability to identify a periodic event, the user’s ability to understand a given periodic event, and to gauge the perceived utility of the tool. The usability study will consist of three phases: teach, test, and gauge. First the user will be taught how to use the tool through a video and written instruction. Later, the user will be tested on their ability to use the tool to answer a set of realistic questions. Measures of success will be based on time and correctness. Finally user perception of the tool will be gauged via a series of Likert scale and open-ended questions. Periodic events exist in all societies, and understanding how these occur allows prediction of other, related events. Nevertheless, challenges related to space, time, and culture make a gaining of such understanding difficult. The interrelated visualizations of PerSE are specifically designed to help gain that understanding.
Keywords
GEOINT; Geovisualization; Periodicity