Study Of Several Problems In Rough Set Theory And Its Applications

With dramatic advance and wide application of the Internet and information systems, we can easily attain large quantities of data that is also in rapid increment daily. Thereby it becomes impractical to handle these data manually. We wish that computers can automatically process these data and extract potentially useful knowledge from them to help us arrange managements and make decisions. This requires a broader and more in-depth study on Machine Learning, particularly Knowledge Discovery in Databases.Rough Set Theory is an important tool to extract knowledge in databases. Due to its capability of dealing with incomplete and inaccurate data, and the high readability of the rules extracted, Rough Set Theory has been considered the apple of researchers' eyes recently. Therefore to make an intensive research on Rough Set Theory will be of great benefit to extract valuable and easily understood knowledge from large amounts of data more effectively. It will also be of great benefit to popularize and apply Data Mining to commercial systems.This paper shows an improved algorithm on the basis of the revised Discernibility Matrix, an important basis in the Rough Set Theory, to increase the efficiency of calculating the core. Since probably all Data Mining algorithms using Rough Set approach are based on the core, this revision will increase the efficiency of knowledge extraction. This paper also presents a further study on the data-mining approach with combination of Rough Set Theory and Expansion Matrix Theory, and uncovers the advantages and disadvantages of both. This survey will be of great help to find a more ideal data-mining algorithm. In addition, we present a rough set-based algorithm to mine a novel pattern in databases under the guidance of the philosophy that there is an exception to every rule. These algorithms are all implemented in our KDD software KnowledgeInside.