Vis enkel innførsel

dc.contributor.advisorRamampiaro, Heri
dc.contributor.advisorNørvåg, Kjetil
dc.contributor.authorDuong, Huy Quang
dc.date.accessioned2021-02-15T13:00:55Z
dc.date.available2021-02-15T13:00:55Z
dc.date.issued2021
dc.identifier.isbn978-82-326-5874-9
dc.identifier.issn2703-8084
dc.identifier.urihttps://hdl.handle.net/11250/2728116
dc.description.abstractThe availability of modern technology and the recent proliferation of devices and sensors have resulted in a tremendous amount of data being generated, stored and handled in various applications that affect almost all aspects of daily life. The analysis and detection of interesting events from such massive amounts of data, which typically originate from multiple sources and have many different forms and characteristics, are important tasks. A major challenge is that the data generated in this way are inherently dynamic, and their underlying distribution may change and evolve over time. Despite the efficiency of existing state-of-the-art methods, many challenges remain to be solved. One limitation is that most existing methods are designed to work well with certain specific characteristics, or for certain predefined assumptions about the data, for example that they are stable and independent, and have identical underlying distributions. Moreover, although many existing approaches have been shown efficient and effective in practical applications, the proposed solutions often lack formal theoretical foundations. Hence, these results are mostly based on heuristics and empirical observations. An important aspect, particularly in dynamic environments, is that the data characteristics are normally unknown beforehand, and that such assumptions about data are rarely valid in real life. With this in mind, the main goal of this thesis is to address the aforementioned drawbacks and challenges, by developing novel approaches and techniques for detecting events, while taking into account the different characteristics and the dynamic nature of the underlying distribution of the data. An important contribution of this work is that in addition to demonstrating the efficiency and effectiveness of our methods in practice, via empirical studies and experiments, we also provide principles and theoretical foundations to prove that the efficiency of the proposed methods also holds in formal proofs. Several extensive experiments are carried out to thoroughly evaluate the performance of the proposed methods. In particular, we perform comprehensive evaluations using both synthetic datasets with ground truths and real-world datasets. The experimental results show that our proposed approaches outperform alternative state-of-the-art algorithms on event detection problems, and demonstrate their efficiency, effectiveness, and applicabilityen_US
dc.language.isoengen_US
dc.publisherNTNUen_US
dc.relation.ispartofseriesDoctoral theses at NTNU;2021:61
dc.titleEvent Detection in Changing and Evolving Environmentsen_US
dc.typeDoctoral thesisen_US
dc.subject.nsiVDP::Technology: 500::Information and communication technology: 550en_US


Tilhørende fil(er)

Thumbnail

Denne innførselen finnes i følgende samling(er)

Vis enkel innførsel