HACKER: human and computer knowledge discovered event rules for telecommunications fault management

Visualization integrated with data mining can offer \´human-assisted computer discovery\´ and \´computer-assisted human discovery\´. Such a visual environment; reduces the time to understand complex data, thus enabling practical solutions to many real world problem to be developed far more rapidly than either humans or computers operating independently. In doing so the remarkable perceptual abilities that humans possess can be utilized, such as the capacity to recognize lanes quickly, and detect the subtlest changes in size, color, shape, movement or texture. One such complex real world problem is fault management in global telecommunication systems. These system have a large amount of built in redundancy to ensure robustness and quality of service. Unfortunately, this means that when a fault does occur, it can trigger a cascade of alarm events as individual parts of the system discover and report fallen making it difficult to locate the origin of the fault. This alarm behavior has been described as appearing to an operator as non-deterministic, yet it does result in a large data mountain that is ideal for data mining. The paper presents a visualization data mining prototype that incorporates the principles of human and computer discovery, the combination of computer-assisted human discovery with human-assisted computer discovery through a three-tier framework. The prototype is specifically designed to assist in the semi-automatic discovery of previously unknown alarm rules that can then be utilized in commercial role based component solutions, "business rules", which are at the heart of many of todays fault management systems.