Supervisor: Dr. Katarzyna Wac

Project rationale

Emerging high-speed wireless networks (e.g. UMTS, HSxPA) providing Internet access give rise to new demanding mobile services in various application domains, including healthcare. An example is the tele-monitoring service, allowing for the continuous monitoring of a patient’s vital signs, and generating a notification event in a healthcare centre in case of an emergency. These types of applications pose strict constraints on the delivered minimum Quality of Service (QoS) expressed in terms of, e.g. end-to-end delay or data loss performance metrics.

The usability of the tele-monitoring service depends largely on the QoS provided by the underlying mobile environment, which consists of various wireless and wired networks with diverse QoS characteristics that influence the overall system QoS. To make sure that the mobile environment meets the required QoS along the tele-monitoring service delivery, this QoS can be measured on a continuous basis along the user’s movements, i.e., location and time change. It would be of high benefit to a mobile service to know the QoS offered by the underlying mobile environments in particular location and time, however, in today’s practice online analysis of the QoS measurements data is hardly feasible in operational environments.

Objective

The objective of this project is to develop a generic data-mining based methodology for QoS evaluation. Collected QoS-related data will be mined with the aim of discovering patterns that reveal dependencies between observed features of mobile environments and the resulting quality of service.

The prospective predictive features, based on which the QoS offered by the underlying mobile environments changes, are: mobile user’s location and time, wireless access network provider (e.g. Swisscom) used and wireless network technology used (e.g. UMTS or WLAN), network signal strength and mobile device battery level. However, many more features may be relevant and the objective of this thesis is to point them as well.

The developed methodology should take into account quality of the collected QoS data itself, and should be incremental, i.e., learn upon a new data. Appliance of this methodology should facilitate adaptivity of mobile applications and services.

Approach

• Relevant background research related to
  - methodologies and tools used to measure the QoS in mobile environments, a structure of the collected data and its quality measures
  - suitable data mining techniques for incremental pattern recognition
• Development of the methodology, with emphasis on objectives such as data quality control, feature engineering and selection, incremental learning
• Validation of the methodology by pattern recognition in pre-collected QoS data
• Writing thesis

Requirements

The candidate should have an experience of the statistics, the Matlab/R and preferably – also data mining techniques. The candidate should also have Java/C# programming skills.