Outlier Detection Techniques On Uncertain Moving Objects

With the development of wireless network communication technology and localization technology, moving objects are widely applied to real life, such as traffic surveillance and control, locaton-based service and so on. As an important part in data mning field, outlier detection technology can discover abnormal data in data sets and can be used to remove noise data or discover new knowledge. Applying outlier detection to moving objects can find out the objects lying in rare region and has the guide role to deploy or redeploy objects in some related applications.Due to inherent and manmade factors, moving objects have unavoidable uncertainty, which is represented by their data uncertainty, existence uncertainty and position uncertainty and so on. Among these, both the existence uncertainty and position uncertainty can make the distribution of them uncertain, while traditional outlier detection is implemented according to their distribution. In this thesis, we concern the existence uncertainty and position uncertainty of moving objects, present the definition of distance-based outlier for uncertain moving objects, study the distance-based outlier detection on uncertain moving objects from these two aspects respectively, and propose corresponding detection technique. The uncertainty brings new challenges to distanced-based outlier detection on moving objects, and the influence they make on the distribution of objects should be fully concerned. Thus, the main goal of this thesis is to explore the uncertainty of moving objects completely and efficeiently and accurately do distance-based outlier detection on uncertain moving objects.In this thesis, for outlier detection on existence uncertain moving objects, we first study the influence of existence uncertainty on the distribution of objects, then analyse the relationship between the likelihood of being outlier and the distribution of nearby objects, and finally propose a dynamic programming based outlier detection technique acoording to these analysis. For outlier detection on position uncertain moving objects, we first study its difference from existence uncertainty, then analyse the influence of position uncertainty on the distribution of objects and the new relationship between the likelihood of being outlier and the distribution of nearby objects in position uncertainty model, and finally propose a sampling based approximating technique, which can be applied to distance-based outlier detection on position uncertain moving objects.