In this paper we shall introduce an approach that forms a basis for temporal data mining. A relation algebra is applied for the purpose of representing simultaneously dependencies among instants, dependencies between instants and intervals, and dependencies between intervals. This enables one to specify and recognise complex interrelationships among point events in data streams. An example is set for the case of thrown alarms in communication networks. This domain, however, suffers from uncertainties in temporal patterns. Therefore a fuzzification is applied to the proposed relational system.
展开▼
