Research On Data Mining Methods Based On QRRECL

Knowledge discovery in databases (KDD), a new generation of tools and techniques for automatic and intelligent data analysis, is currently an active research area. Concept lattice (Galois lattice) is a formalized tool for concept discovery from data, and has been widely used in many scientific applications. This paper studies the data mining methods based on Quantized relative reduced extended concept lattices (QRRECL). The main topics include:1) Firstly we summarize the basic principles of KDD. Then we describe some notions of Concept lattices model and Extended Concept lattices model. Several algorithms for building a Galois lattice and their applications are also introduced. Lastly we give a theoretical comparison of Concept lattices and Decision Tree.2) In this paper, we present a novel QRRECL model. Because it denotes the extension and intension of concept in the compact form and shows the relations among the concepts more clearly, it is a more efficient tool for knowledge discovery especially in large databases. An incremental algorithm for building the QRRECL and Hasse diagram has been proposed and analyzed. We also study the maintenance of QRRECL when the database is updated, and present relevant algorithms for inserting an object into the QRRECL and deleting an object from the QRRECL respectively.3) Mining association rules is an important task in KDD. In this paper, we discuss the problem of mining association rules based on the QRRECL, and describ a formal framework for rapid generation of association rules. We study the methods to construct Frequent concept sub-lattice (FCSL) using dynamic pruning technique and static pruning technique respectively, and develop the algorithms for building FCSL. In addition, we show how to generate all high confidence rules from an example database of transactions based on FCSL, and propose the relevant algorithm for rule mining and pruning.