Analysis of Mobile Service Usage Behaviour with Bayesian Belief Networks

Abstract

The purpose of this paper is to identify probabilistic relationships of mobile serviceusage behaviour, and especially to understand the probabilistic relationship between overallservice usage diversity and average daily service usage intensity. These are topical themes dueto the high number of services available in application stores which may or may not lead tohigh usage diversity of mobile services. Four analytical methods are used in the study, all arebased on Bayesian Networks; 1) Visual analysis of Bayesian Networks to find initiallyinteresting patterns, variables and their relationships, 2) user segmentation analysis, 3) nodeforce analysis and 4) a combination of expert-based service clustering and machine learning forusage diversity vs. intensity analysis. All the analyses were conducted with handset–based datacollected from university students and staff. The analysis indicates that services exist, whichmediate usage of other services. In other words, usage of these services increases theprobability of using also other services. A service called Installer is an example of this kind of aservice. In addition, probabilistic relationships can be found within certain service cluster pairsin their usage diversity and intensity values. Based on these relationships, similar mediationtype of behaviour can be found for service clusters as for individual services. This is mostvisible in the relation between System/Utilities and Business/Productivity service clusters. Theydo not have a direct relationship but usage diversity is a mediator between them. Furthermore,segmentation analysis shows that the user segment called “experimentalists” uses moremediator services than other user segments. Furthermore, “experimentalists” use a muchbroader set of services daily, than the other segments. This study demonstrates that a BayesianNetwork is a straightforward way to express model characteristics on high level. Moreover,Node Force, Direct and Total effect are useful metrics to measure the mediation effects. Theclustering implemented as a hybrid of machine learning and expert-based clustering process isalso a useful way to calculate relationships between clusters of more than a hundred individualservices.