Research On Data Mining Based On Distributed Concept Lattice Model

As the result of Artifical Intelligence,databases and statistics,the aim of data mining is to find some meaningful patterns from a lot of data. Data mining has great application value. But with the unprecedented growth-rate at which data is being collected and stored,the efficient mining of useful information from the data available is becoming an increasing scientific challenge. Now the most efficient way to solve this problem is Parallel and Distributed data mining by using parallel calculation and distributed storage. But to the parallel calculation and distributed storage themselves,how to organize the distributed storage of data in reason and how to perform the parallel processing effectively are a task to be solved urgently at present. Due to its qualities of solid mathematical basis and the adaptability to distributed process,distributed concept lattice is regarded as a very perfect tool to solve the issues. The aim of this thesis- is to build the mathematical model for distributed concept lattice,study its mathematical properties,and provide the mathematical foundation for its distributed storage and parallel processing.This the-sis includes the following content:1) The background of data mining is introduced. To solve the problems that follow the growth of data,we propose to build the mathematical model of distributed concept lattice by taking advantage of its qualities of solid mathematical basis and the adaptability to distributed process2) The mathematical basis of concept lattice is introduced,which includes the definitions of order theory and lattice theory related to concept lattice. We give two kinds of lattice construction algorithms:batch algorithms and incremental algorithms. The details of classical batch algorithms,such as Bordat and Chein algorithms,and classical incremental algorithms such as Gordin algorithms are introduced.3) We build a mathematical model of distributed concept lattice and discuss the union operation to combine two extention-independent same field lattices and the union operation to combine two intention-independent same field lattices. The experiments show that our union operation algorithm is superior to the algorithm that gets the lattice from context directly in time-consuming. For the completeness in mathematics,we also discuss the intersect operation to remove everything except the overlapping areas of two same field lattices.4) We introduce a classical algorithm which can extract implication rules based on concept lattice. The algorithm builds lattice incrementally and updates the set of rules simultaneously. We modify the structure of lattic to meet our requirement of finding frequent itemsets. Then we propose a union algorithm to combine two sets of implication rules for the requirement of distributed applications. The experiments also show that our union operation algorithm is superior to the algorithm that extracts the rules from context directly in time-consuming.