Virtual data mining of an ATM network

ATM networks provide customers with a wide variety of services from video links to audio telephones. In order to remain competitive a company will have to guarantee provision of a given quality of service which the customer may wish to independently verify. However this may prove to be very difficult since vast quantities of data may be logged, typically 90 MB in 15 minutes. Therefore it is important that we find a means by which we may mine the key performance parameters, which determine the health of the system, from the vast amount of data which is collected and present them to the customer in a manner which is easy to understand. We describe our initial work regarding an approach to this problem which combines the emerging technologies of data mining and virtual reality (VR). The information regarding network performance is made available to the user by data mining techniques which underpin the VR representation of the network. In practice functionality is assigned to each virtual world object using the software´s scripting language so that data mining of the network performance data source may occur. A prototype was developed using the Superscape virtual reality software. This software uses an object orientated approach to world construction (a world in this case consisting of all network elements) which is particularly suited to the hierarchical nature of networks. Simulated ATM network data in file format, was mined by the virtual reality data mining tool (VDMT) in order to simulate information transfer between the real world network and the virtual world network