On selection rules for database compression using KDD

Despite the drop in disk prices in recent years, the costs associated with disks still represent the primary costs for large-scale databases such as t hose used in data warehouses. Compression and storage of such databases is seen as an effective means to reduce these costs. The authors' research gro up has already proposed a compression method that exchanges the priority fo r rules and data after extracting rules latent in data by using knowledge d iscovery as a database compression method which can access a database in it s compressed state. However, this proposed method has the problem of produc ing differences in the compression ratio due to the priority with which the extracted rules are used for compression. Simply finding all the combinati ons for the priority for the use of such rules is not practical, and so in this paper the authors propose a rule selection method which provides compa ratively good compression ratios without excessive computational requiremen ts, then demonstrate the validity of their method using experimental result s. (C) 2001 Scripta Technica.