Study On Constructing Algorithms Of Concept Lattice

Concept Lattice is a powerful tool for data analysis, which reveals the generalization/specification relationships between concepts in a vivid and concise way by a Hasse diagram. A concept lattice is built from data sets through the process of concept clustering. The can be used in many other machine learning tasks, such as classification or association rule mining. Recent studies of practical applications in data mining and software engineering have put the emphasis on the need for both efficient and flexible algorithms to construct the lattice , for which various algorithms have been developed.These algorithms can be divided into two main categories, batch construction and incremental construction. The batch algorithms be used building concept lattice of formal context which had been given., for example Bordat's, Chein's, Ganter's, Nourine's. After the concept lattice have been built, the way of building concept with incremental algorithms is more convenient than batch algorithms with various dataset updates.The efficient algorithm among batch construction has been suggested by Bordat which generates both the concept set and the Hasse diagram of the lattice. The obvious drawback of the method is that a concept is generated several times, once per each lower cover. Our paper presents a approach for lattice construction for remedy of the problem. The definition of hierarchical of elements be given by the length of chain. Hierarchical concept lattice properties are discussed, and two theorems are given .So according to the properties and theorems, a new algorithm is provided to construct the concept lattice , and one concept can't be generated several times at this process.The design of flexible algorithms was pioneered by Godin who designed an incremental method for constructing the concept lattice. On condition that dataset updating can't add attributes, we present an incremental algorithm for lattice construction. It be proved that the new algorithm can greatly reduce the searchspace of parents and children of newborn concept node.