Knowledge discovery with the associative memory modell neunet

Abstract

At the University of Linz a remarkable associative memory model has been developed. A neural network analogous self learning system with the capability of parallel and serial association. But, for data mining tasks it has one shortcoming. It can not reproduce how often it has seen a part of a pattern in its past - it is not able to compute frequencies. In this contribution we introduce an extension of the model with which frequencies, support and confidence are feasible. Besides, all advantages of the model could be retained. Short examples and a comparison with a common data mining tool complete the paper.