An artificial immune algorithm for association rule mining among concepts with uncertainty

During a design procedure of association rule mining approach, there are two common issues: the transformation method from continuous quantitative attributes to qualitative concepts, and efficiency of data mining. In order to acquire association rules in a database with different types of attributes, the cloud transformation which is included in the cloud model theoretical framework is applied as an uncertain concept extraction tool in this paper. By the feature analysis of association rule mining in uncertain concept space, the frequent item-set generation is converted to a combination optimization problem. A modified object function and artificial immune algorithm for association rule mining are designed accordingly. A novel method of non-frequent item hyper set detection is introduced to reduce the number of database scanning and improve the efficiency. The numerical experiments show that the proposed algorithm can accomplish the association rule mining by global random search, with the robustness that the computational cost is insensitive with the variation of threshold parameters.