Efficient mining of association rules in large dynamic databases

Abstract

Mining for association rules between items in a large database of sales transactions is an important database mining problem. However, the algorithms previously reported in the literature apply only to static databases. That is, when more transactions are added, the mining process must start all over again, without taking advantage of the previous execution and results of the mining algorithm. In this paper we present an efficient algorithm for mining association rules within the context of a dynamic database, (i.e., a database where transactions can be added). It is an extension of our Partition algorithm which was shown to reduce the I/O overhead significantly as well as to lower the CPU overhead for most cases when compared with the performance of one of the best existing association mining algorithms.