KIT | KIT-Bibliothek | Impressum | Datenschutz

Comparing Predictions of Object Movements

Taghizadeh, Saeed; Schäler, Martin; Böhm, Klemens

Abstract:

Estimating the future location of moving objects using different estimation models, such as linear or probabilistic models, has been investigated extensively. However, the location estimations of those models are generally not comparable. For instance, one model might return a position for some object, another one a Gaussian probability distribution, and a third one a uniform distribution. Similar issues arise for query answers. In this paper, we examine the question how estimations of different models can be compared. To do so, we propose a general model based on the central limit theorem. This allows handling different PDF-based approaches as well as models from the other groups (i.e., linear estimations) in a unified manner. Furthermore, we show how to inject privacy into the general model, a fundamental pre-requisite for user acceptance. Thus, we support well-known approaches like k-anonymity and spatial obfuscation. Based on our general model, we conduct a comprehensive experimental study considering a real-world road network; comparing models form different groups for the first time. Our results, for instance, reveal that estimation models based on individual velocity profiles are not necessarily better than models, which estimate the future location of objects only based on their direction. ... mehr


Volltext §
DOI: 10.5445/IR/1000081005
Veröffentlicht am 12.03.2018
Cover der Publikation
Zugehörige Institution(en) am KIT Institut für Programmstrukturen und Datenorganisation (IPD)
Publikationstyp Forschungsbericht/Preprint
Publikationsjahr 2018
Sprache Englisch
Identifikator ISSN: 2190-4782
urn:nbn:de:swb:90-810050
KITopen-ID: 1000081005
Verlag Karlsruher Institut für Technologie (KIT)
Umfang 13 S.
Serie Karlsruhe Reports in Informatics ; 2018,3
KIT – Die Forschungsuniversität in der Helmholtz-Gemeinschaft
KITopen Landing Page