Abstract
Algorithms to discovering association rules are still the most popular and important areas. Various algorithms have being proposed. Examples are Apriori [2], DIC [3], partition-based [4], Samples [5], DHP [7], etc. The common property for these methods is that examined data within the underlying databases are static. However, in some application such as pages in web environment, the examined data are subject to frequently updated, deleted, or inserted. Therefore the discovered association rules or knowledge by these methods may not be up-to-date. To discover association rules for such a dynamical environment, we proposed a new method based on hash and clustering technique, and it is called H-clustering method. To evaluate the efficiency of the new methods, its performance is evaluated with respect to that of the partition-based method and DIC method. The reason of selecting these two methods for comparison is that both methods take fewer iterations to scan the transactions. The measurement metrics are the total number of transaction been scanned and the total execution time to discover the required association rules.