Algorithms for Building Concept Sets and Concept Lattice

Abstract

In our days there is an increasing interest on the application of concept lattices for data mining, especially for generating association rules. The building of concept lattice consists of two, usually distinct phases. In the first phase the set of concepts is generated. The lattice is built in the second phase from the generated set. The paper gives an overview of the available methods and presents a proposed method for contexts of large size where the full context can not be stored in the main memory and some objects may be repeated in the context several times. The proposed algorithm for concept set generation is a fine-tuned version of the incremental concept set building method. At the end of the paper, the test results for comparing the new method with some known methods are given. The proposed method yields in a significantly better cost value than the other methods under the assumed conditions.