- Affairs Relational Database-based Association Rule Mining Algorithm

Abstract Mining for association rules between items in a large database has been descried as an important database mining problem. Since the problem was proposed by Agrawal etc. in 1993, it has been paid attention to. Because of mining for association rules in a very large transaction database, one of the main challenges in database mining is developing fast and efficient algorithms. Previous algorithms perform directly in the transaction database. Although some methods of improving I/O overhead were given, those methods require to scan the whole transaction database many times. Therefore, the algorithms have much I/O overhead and can not achieve significant improvement of efficiency. In this paper, I present efficient algorithms for mining association rules that are fundamentally different from known algorithms. Firstly, the transaction database is changed to the item梩ransaction connection database, and then the algorithm for mining for association rules performs directly in the item? transaction connection database. Because item(itemset) is regarded as index key and every record includes a Tid梥et of a item(itemset), the support of frequent k梚teniset can be easily generated by joining the Tid梥et of the two frequent (k1) itemsets. In addition, when frequent k梚temsets are generated, the item梩ransaction connection database Dk is generated. Dk includes frequent k梚temsets and related Tid梥ets. If the algorithm will generate frequent (k-i-1)梚temsets, it needs to scan Dk much smaller than D? The step is prune step and helps to reduce I/O overhead. In this paper, my main results as follows: l.Gjving the algorithm for changing the transaction database to the item梩ransaction connection database. 2. Presenting the algorithm for mining single條evel association rules and the algorithm for mining multilevel association rules in the item梩ransaction connection database. 3.Discussing increasingly mining association rules and giving the algorithm. 4. Discussing the algorithm for mining association rules which have the new minimum support.