Association rules are a form of knowledge representation used in decision making systems due to their simple structure and high information storage potential. This feature can be obtained through association rule mining algorithms, such as Apriori, which takes a dataset as an input parameter and returns a set of association rules. However, the existing algorithms return a large number of rules, which makes the use of association rules costly for computer systems and very hard to interpret for domain experts. In order to overcome this difficulty and facilitate the application of association rules in solving decision making problems, many researches have been searching for a computational solution to reduce the amount of association rules in such a way that there is no significant loss of information. This paper presents two computational procedures for minimizing the number of association rules that fully represent a dataset. Then, the authors present the tests performed and a comparative study with other methods in the literature. In view of the success achieved, the authors make their considerations about the results and point out the new direction of the project.
|
