A User Behavior Clustering Based Fault Pre-detection Mechanism for QoE Assurance on Mobile Devices ⋆

Pre-detection mechanisms are often used to ensure the users’ QoE on mobile devices before potential service/application malfunctions really occur. In order to improve the efficiency of pre-detection, an optimized detection mechanism is proposed. In the mechanism, a priority set of services is selected by considering user’s dependency degree on services, service’s priority and network load. Firstly, a model based on fuzzy theories is proposed to weigh the dependency degree of user to each service. Then, a clustering model by analyzing user behavior is introduced to divide mobile users to several clusters. After that, an optimized algorithm for selecting priority set of services to be pre-detected is proposed by considering the user-service dependency degree generated before, combined with failure rate of services and network load status. Based on that, the service/application with higher usage rate and higher failure rate could have the priority of being diagnosed. Meanwhile, the proposed fault pre-detection mechanism is implemented on a prototype system, and the algorithm is validated by computing time and services availability comparisons under various network environments.