RWTH Aachen University
Mining association rules from databases has attracted great interest because of its potentially very useful application. A generalization of algorithm is proposed for implementing the well-known effective inductive method of constructing sets of cardinality (q+1) ((q+1)-sets) from their subsets of cardinality q ((q)-sets). A new neural net-work-like combinatorial data-knowledge structure supporting this algorithm is advanced. This structure can drastically increase the efficiency of inferring functional and implicative dependencies as like as association rules from a given dataset.