Case-based Reasoning for Complex Telecommunication Systems Alfons Schuster, Roy Sterritt, Ken Adamson, Mary Shapcott, Edwin P. Curran University of Ulster, Faculty of Informatics, School of Information and Software Engineering Shore Road, Newtownabbey, Co. Antrim BT37 0QB, Northern Ireland email: {a.schuster, r.sterritt, k.adamson, cm.shapcott, ep.curran}@ulst.ac.uk ABSTRACT: This paper aims to identify the potential of case-based reasoning (CBR) for problem-solving in complex telecommunication systems. The system under investigation is an environment on which automated test procedures are carried out frequently. The data gathered in these tests is used by engineers to assess the performance and the quality of the system. However, in some situations the complexity of the system together with the large size of the recorded data can lead to situations in which an engineer may find it difficult to come up with a quick and valid assessment. Based on research carried out in previous projects this paper emphasises the potential of CBR for data analysis in this area. The document therefore has different goals. Initially it provides an overview of the automated testing environment and the problems in this domain. The paper then highlights the advantages CBR may provide for intelligent data analysis. KEYWORDS: telecommunication systems, automated testing, data analysis, case-based reasoning 1 INTRODUCTION There is a clear correlation between developments in telecommunications modelling and those within the fields of inferential statistics, computing science and technology. Exploring this correlation the authors in collaboration with NITEC - (Northern Ireland Telecommunications Engineering Centre) - NORTEL NETWORKS (from now on simply referred to as NORTEL) have undertaken research projects in the telecommunications domain. The projects were developed with the aim of understanding further the behavioural issues of large scale Synchronous Digital Hierarchy (SDH) networks Tanenbaum (1996), Sexten & Reid (1992). For example, a knowledge discovery approach using live data from NORTEL’s test SDH network was utilised to produce an abstracted cause and effect behavioural model of the network Sterritt (1998a). Currently the authors are collaborating with NORTEL to automate their shop floor test facility. This will involve the development of real-time knowledge acquisition tools and aims to extract abstract behavioural system models. This work has raised new issues and spawned new ideas to improve the efficiency and robustness of the testing process. A number of problems have been identified in relation to current techniques including, for example, deficiencies in the knowledge elicitation process where parameters on a specific event have not been assimilated in the model, the lack of robustness, the fast complexity of the domain, the rigidity and inability to work within the constraints of limited data, and finally the inflexibility to accommodate new developments can be identified as shortcomings. This paper aims to highlight the potential of CBR as a possible solution to some of these shortcomings. For example, CBR is able to operate in domains that are not well defined, or where problems arise due to missing data and/or data shortage. CBR also provides the facility to reduce the complexity of many systems through simple, robust and easy maintainable entities, so-called cases. And finally, CBR further provides advantages in terms of learning and explainability. Although the major intention of the paper therefore is to more clearly identify the value of CBR in the automated testing field the paper also provides an overview of future research directions. These directions propose a hybrid system, not only for (intelligent) data analysis, but also for data storage (database, data warehouse) to enable the achievement of continuous improvement in the testing process by a rigorous and competent sustained learning approach to automated testing. The remainder of the paper is organised as follows. Section 2 describes the automated testing environment. Section 3 provides an overview on the data processing involved in the project. CBR and its advantages to some of the problems in this domain are summarised in Section 4. Finally, Section 5 ends the paper with conclusions and future work. 2 TESTING OF TELECOMMUNICATION EQUIPMENT NORTEL’s main business activity is the design and manufacture of telecommunication equipment. They are the world leader in supplying information transmission systems. For example, their 10 gigabit per second transport products