Hierarchical association rule mining in large and dense databases using genetic network programming

In this paper we propose a new hierarchical method to extract association rules from large and dense datasets using genetic network programming (GNP) considering a real world database with a huge number of attributes. It uses three ideas. First, the large database is divided into many small datasets. Second, these small datasets are independently processed by the conventional GNP-based mining method (CGNP) in parallel. This level of processing is called local level. Finally, new genetic operations are carried out for small datasets considered as individuals in order to improve the number of rules extracted and their quality as well. This level of processing is called global level. The amount of small datasets is also important especially for avoiding the overload and improving the general performance; we find the minimum amount of files needed to extract important association rules. The proposed method shows its effectiveness in simulations using a real world large and dense database.