Research On Mining Constrained Association Rules

In association rule mining process, weusually take "support- confidence" mechanism to output the rule by the order, which can cause blindness mining and resulted in the lowalgorithm efficiency. And in the mining process the lack of the user's control, so the end result may produce a lot of useless rules and are not interest for the user. Also how to effectively use the results of other disciplines, especially intelligent algorithm artificial immune system, genetic algorithm promote association rule mining under complex constraints arealso worthy forthe further study. Therefore, this paper study on the constraint-based association rules mining algorithm.For the term constraint, this paper proposes an algorithm of constrained association rules mining based on indexed mechanism.The algorithm inherits Separate algorithms which divided into three parts according to the length ofthe satisfied different constraintsprinciple. And the algorithm introduces the filter ideas and indexing mechanism. The algorithm first uses the filter function to make items set conformities constraints number greatly reduce. Then thealgorithm introduces the index mechanism, which in the calculation the Support of the itemsets don't need to scan the data set over and over again and reduce the time. Experimental results show that the algorithm can effectively reduce the data size and the execution time in the case of the Support smaller and the fewer numbers of constraints.For the complex constraints, this paper proposes an algorithm to deal with the complex constraint association rules mining based on immune genetic algorithm.The algorithm is a metaphor from the immune response information transmission mechanism, by adjusting the competitiveness of the population between individuals, narrowing the search space. And the algorithm doesn'tgenerate large-scale candidate set by adaptively adjusting the direction and scope. The algorithm can overcome the shortcomings of prematurity in genetic algorithm and reach the globally optimal solution. By contrast with the artificial immune system, genetic algorithm, the experimental results show that the algorithm have strong convergence.