Time Series Analysis in the Context of the Internet of Things: Classic Solutions vs New Approaches for Typical Problem Situations

von Schwerin, Clemens (2018) Time Series Analysis in the Context of the Internet of Things: Classic Solutions vs New Approaches for Typical Problem Situations. Masters thesis, Ulm University.

[thumbnail of MastersThesisVonSchwerin.pdf] PDF - Registered users only - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
Download (4MB)

Abstract

The Internet of Things (IoT) has been growing at a fast rate in size and significance over the past years. Data analytics for the IoT may enable various new applications, ranging over multiple domains and shaping the everyday life of the future. However, the IoT raises new challenges and requirements with respect to data processing and analytics methods. Varying infinite live data streams transport time series data of high volume, very high velocity and questionable veracity. Many IoT analytics applications require handling of those data streams in real time, being robust to intermittent connection loss, having limited knowledge about the data’s context or coping with bandwidth constraints. Finally, the landscape of protocols, formats, tools and frameworks used in the IoT context is still very heterogeneous and fragmented.

This thesis elaborates the challenges of and proposes new approaches for sophisticated IoT data analysis. For that purpose, following a design science approach, the gap between classic and IoT data analysis is systematically discussed from different perspectives as part of the rigor cycle. Also, key challenges and requirements for IoT data analytics are illustrated as part of the relevance cycle. Based on that, a suitable data processing architecture for the IoT using fog computing and a service-oriented architecture (SOA) is derived as a basis for sophisticated IoT data analysis. That architecture constitutes the first artifact in the design cycle. Real-time processing and a robust connection to IoT devices is ensured by bringing data processing capabilities spatially close to the IoT devices. At the same time, specifying standard service layers, identification methods for devices and resources as well as their description adds usability through metadata and counteracts fragmentation and heterogeneity in the IoT landscape. In this context, essential tools and methods for IoT data analytics are reviewed and standard solutions for common reoccurring patterns in workflow-based IoT data analytics are also developed as part of the design cycle. Lastly, two case studies are conducted in order to illustrate practical IoT data analytics using typical examples from important IoT application domains, thus completing the relevance cycle in a field test.

Item Type: Thesis (Masters)
Subjects: DBIS Research > Master and Phd-Thesis
Divisions: Faculty of Engineering, Electronics and Computer Science > Institute of Databases and Informations Systems > DBIS Research and Teaching > DBIS Research > Master and Phd-Thesis
Depositing User: Nicolas Mundbrod
Date Deposited: 14 Mar 2018 11:32
Last Modified: 14 Mar 2018 11:32
URI: http://dbis.eprints.uni-ulm.de/id/eprint/1603

Actions (login required)

View Item
View Item