Research On Knowledge Discovery In Databases Based On Distributed Concept Lattices

As the result of Artificial Intelligence, Machine Learning, Databases and Statistics and other subjects, the aim of Data Mining and KDD is to find some meaningful patterns from databases with AI technology. So it has great application value. With the situation that the quick development of information technology leads a rocket in all kinds of data, fetching meaningful patterns effectively from large databases have become a challenge and an effective way to solve the problem is parallel\distributed technology. There are also many problems about how to organize distributed data storage and parallel processing whether on technology or on theory to research. Due to a stability theory base, self-contained structure and parallel character, concept lattice is a tool to solve the problems aforesaid. In this thesis the model of distributed concept lattice and data mining research on which is given. This thesis including the following content:1. The basic notion and the background of the data mining and KDD is introduced.2. The mathematical foundation, traditional concept lattice research, extend concept lattice model and the building algorithm of the concept lattice are introduced. The advantage and disadvantage of batch algorithms and incremental algorithms are analyzed, based on which a new kind of parallel algorithm which combines the advantages of both batch algorithms and incremental algorithms are put forward.3. A new kind of distributed concept lattice model is given out, advance the organic split way of data, which is different to the split way in traditional distributed databases that split the data in horizontal direction, vertical direction or mixed ways. The parallel concept lattice building algorithm SEA based on the organic split concept is put forward. The experiment show that the SEA algorithm is superior to the algorithm that gets the lattice from context directly in time-consuming.4. On the basis of wrong classification cost technology a classification-technology which use the class characters to classify the new instance is given on the basis of distributed concept lattice model.