Bayesian Network Analysis of Mobile Service and Device Usage

Research output: ThesisDoctoral ThesisCollection of Articles

Abstract

The use of mobile services has changed significantly in the last twenty years. Initially, only mobile calls and short messages were the services used, whereas today the use of mobile services has evolved into multiple parts of a human's daily life. Due to this evolution, research into mobile service usage consists of many, often complex study themes. In parallel, statistical methods and computing systems have developed rapidly, providing sophisticated methods for researchers, for example, to study mobile service usage. The objective of this thesis is twofold. On one hand, the purpose is to describe the best practises related to a Bayesian networks-based statistical method. On the other hand, the evolution of mobile services in Finland and their usage patterns are explored empirically using the Bayesian networks approach. The thesis documents systematically Bayesian networks-based approaches for descriptive, predictive and explanatory data analysis. The documentation covers the whole process from the pre-processing of data, construction of the models to the inferencing based on the constructed models. Moreover, differences between predictive and causal models are clarified. The results from the empirical analysis deal with multiple aspects of mobile service usage. The following aspects were of greatest significance: structural breaks in the popularity of mobile device features in Finland were identified; these transition points over time acted as a basis for the explanation for the changes in the Finnish mobile market; mobile service users were segmented into main three groups based on service usage patterns; users' age as an important factor related to the user behaviour were indicated; and, finally, multiple risks associated with the use of mobile services were detected. Overall, the thesis provides practical information about Bayesian networks, which will hopefully encourage more researchers to use the Bayesian networks approach in the statistical analysis. Furthermore, the research provides several individual results, which are of interest to all mobile ecosystem players. These are, for instance, the discovered role of some services as mediators of other service usage, the central role of age as a reason for service intensity and diversity, and the creation of an expert knowledge elicitation process in cases when the experts from a knowledge and location point of view are dispersed.
Translated title of the contributionMobiilipalveluiden ja laitteiden käytön analyysi Bayes-verkoilla
Original languageEnglish
QualificationDoctor's degree
Awarding Institution
  • Aalto University
Supervisors/Advisors
  • Hämmäinen, Heikki, Supervising Professor
  • Hämmäinen, Heikki, Thesis Advisor
Publisher
Print ISBNs978-952-60-8794-8
Electronic ISBNs978-952-60-8795-5
Publication statusPublished - 2019
MoE publication typeG5 Doctoral dissertation (article)

Keywords

  • Bayesian networks
  • mobile service usage
  • mobile device
  • causal analysis
  • predictive analysis
  • clustering

Fingerprint Dive into the research topics of 'Bayesian Network Analysis of Mobile Service and Device Usage'. Together they form a unique fingerprint.

Cite this