The Diclique Representation and Decomposition of Binary Relations
Journal of the ACM (JACM)
Principles of data mining
Algorithm MONSA for all closed sets finding: basic concepts and new pruning techniques
WSEAS Transactions on Information Science and Applications
Some ideas for determinacy analysis realisation
ASC '07 Proceedings of The Eleventh IASTED International Conference on Artificial Intelligence and Soft Computing
| Hi-index | 0.00 |
Determinacy Analysis (DA) is a method that solves tasks of data mining (it enables to describe by the rules the subset of objects determined by the user). There are different treatments in DA: step by step and DAS-like in algoritmic view, one solution and multiple solutions as a result, additive and non-additive sets of rules in systematic view. Thereat the essence of the method itself does not change, only the rules change. There is a number of lacks in DA base algorithms: they are too labour-intensive (step by step approach) or they find only a limited set of rules (i.e. only one system of rules of many possible systems, in case of DAS). In this paper we show that DA can be reduced to the diclique finding task that is well-known from the graph theory, we present the prerequisites to take into account and explain how it influences the rules. The diclique-based DA enables to set up the DA tasks of new type: to find a system with minimal number of rules, to find a system with minimal number of shortest rules (for example). Reducing DA to a diclique finding task, the basis for the new generation of DA algorithms is founded.