SEMINAR: A Study of Geovisual Analytics for Exploring Event Anomalies Over Multiple Geospatial Data

Md. Monjur Ul Hasan
M.Sc. Candidate
Co-supervisors: Dr. Orland Hoeber and Dr. Wolfgang Banzhaf

A Study of Geovisual Analytics for Exploring Event Anomalies Over
Multiple Geospatial Data Sets

Department of Computer Science
Wednesday, July 9, 2014, 12:00 p.m., Room EN 2022


 

Abstract

Geospatial data sets are often used to describe events performed by moving entities, such as shoppers purchasing products from stores, tourists visiting historical places, or fishing vessels trawling. When such events are independently collected in multiple data sets, comparing common events for positional discrepancies with their spatial and temporal contexts may reveal important insights, such as data entry, instrumental, intentional, and/or processing errors. The analysis of these event anomalies is particularly challenging when the data sets are collected using different temporal granularities and cover large spatial and temporal ranges. In this work, independently collected geospatial point data and movement data are considered, describing event locations and movement activities of the entities that performed these events, respectively. Detecting event anomalies may be a trivial task if these data sets have the same temporal granularity and are temporally synchronized. However, such conditions cannot be guaranteed for independently collected data sets. Manual analysis of these event anomalies requires significant cognitive effort and focus in order to compare the data for the same entities to one another. At the same time, automatic methods are difficult to tune in order to both find important anomalies and avoid an overload of false-positives. To address this problem, a geovisual analytics approach is introduced as a combination of human-centred analysis and automatic processing methods. Candidate event anomalies are extracted from the data, using analyst-adjustable thresholds based on their domain specific knowledge, experience, and the type of anomalies being sought. The detected event anomalies are visualized on a map, showing clear links between the movement paths of the entities and the event locations, with the goal of reducing the cognitive effort for matching the data to one another. This approach makes extensive use of spatial and temporal filtering, geovisualization, colour encoding, and anomaly threshold filtering. It is highly interactive, supporting knowledge discovery through visual exploration and analysis of the data sets. In order to illustrate the benefit of this approach, a prototype system has been developed for the purposes of analyzing event anomalies within two real world data sets related to fisheries enforcement. Field trials were performed with five expert fisheries analysts to evaluate the system. Results from this study confirm the value of the approach and its potential for supporting geospatial anomaly analysis of commercial fishing events.

Contact

Department of Computer Science

230 Elizabeth Ave

St. John's, NL A1B 3X9 CANADA

Tel: (709) 864-2530

Fax: (709) 864-2552

becomestudent@mun.ca