Research And Implementation On Frequent Closed Pattern Mining Algorithm Based On Element Increasing Searching Strategy

The amount of data is increasing rapidly with the fast development of database technique and the diversification of methods of acquiring data. How to extract useful information and knowledge from huger and huger database has already been a hot research field of computer science technique, which is the data mining technique. Associate rules mining is an important branch of data mining filed. Associate rules mining, which has great research value and application value, is to mine rules reflecting some deep relationship of items in a database.Mining frequent closed patterns is a pivotal step of associate rules mining, which determine the entire performance of associate rules mining. Previous research work mainly focused on algorithms based on depth first searching strategy, and very limited researches were done on algorithms based on other searching strategies and the storage structure of frequent patterns. According to the reasarch status mentioned above, this thesis first proposes two novel data structures for frequent patterns'storage:Suffix Tree and Ordered Graph, and proposes algorithms of query a set, add a set and delete a set on Suffix Tree and Ordered Graph. Experimental results show that the two storage structures both perform better than previous storage structures. This thesis also proposes a novel algorithm of frequent closed pattern mining:EISAM algorithm. The algorithm uses a novel searching strategy:element increasing searching strategy instead of using the traditional depth first or breadth first searching strategies, which makes EISAM algorithm has the property of increment maintenance. Efficient pruning rules and pre-process techniques are also proposed in order to improve the performance of the algorithm, which make the algorithm useful under more situations. Experimental results show that the performance of EISAM algotithm is better than previous algorithms.This thesis first briefly introduces the background knowledge and previous work, and then proposes the structures of Suffix Tree and Ordered Graph, and operations related. The EISAM algorithm based on element increasing searching strategy is proposed next, and at last experiments were done in order to test the performances of the storage structures and the algorithm.