A Novel Method of Mining Association Rule with Multilevel Concept Hierarchy

In data mining, there are several works proposed for mining the association rules which are frequent. Researchers argue that mining the infrequent item sets are also important in certain applications. Discovering association rules are based on the preset minimum support threshold given by domain experts. The accuracy in setting up this threshold directly influences the number and the quality of association rules discovered. Even though the number of association rules is large, some interesting rules will be missing and the rules quality requires further analysis. As a result, decision making using these rules could lead to risky actions. Here the focus is mainly on mining both the frequent and infrequent association rules which are more interesting and does not have redundant rules. This is based on predefined rules formed using propositional logic and then the predefined rules are processed by comparing with elements in the actual dataset. The association rules which are obtained will not have redundancies and they will be logically correct. Generalized association rules will be obtained if single level mining is performed. These rules can only help in very high level decision making. In order to allow for in-depth decision making, more specific association rules are obtained. Therefore multiple level mining processes is employed here.