International Journal of Advances in Computer Science and Its Applications
Author(s) : CHUN-CHE HUANG, HORNG-FU CHUANG, YING-LING HSIEH, ZHI-XING CHEN
In a dynamic database (DB), data deletion operations are a frequent feature of database management activities. Unfortunately, most existing DM algorithms assume that the database is static and that updating a database requires re-computation of all the patterns to enable rule extraction. Of the DM techniques available, the rough set (RS) approach is a knowledge discovery tool that can be used to help identify logical patterns hidden in massive data. It is also useful for knowledge discovery, pattern recognition and decision analysis. However, traditional RS approaches cannot produce rules with preferential order and often lack focus. They generate too many rules and cannot guarantee that the decision table is credible. This study proposes a DAREA (Decremental Alternative Rule-Extraction Algorithm) to address the issue of data deleted from the database and to generate preference-based rules, according to a strength index (SI), specifically for the case wherein the desired reducts are not necessarily unique. The algorithm does not need to recomputed rule sets that can quickly generate and complete rules, from the very beginning. Experiments are presented to validate that the proposed approach is superior to the traditional RS approach.