The Research On Association Rules And Its Distributed Processing Based On Quantitative Rule Lattice

The formal concept analysis has been greatly applied to many fields,such as software engineering,scientific discovery and information retrieval in recent years since it was presented by professor Wille.Concept lattice,which can uncover the relationship between concepts through Hasse Diagram and extract the association rules directly,is the core of formal concept analysis.Due to the time and space complexity of mining association rules from concept lattice,this paper is decreasing the time of mining rules.The general method in the precedent studies as follows,the first step is to construct concept lattice and another is to work out association rules from concept lattice,but the critical step is the that scope of concept lattice is increasing sharply by exponent,in accordance with formal context, which can reduce to the low effectiveness of calculation.This paper put forward a new method which could integrate the former two steps into one step and is easily to be executed. The main body of this thesis includes:(1) Firstly, this paper presents quantitative rule lattice, an extended modal based on the concept lattice. The process of increment of construct lattice could bring out SLITs respectively which can directly calculate the exact rule and approximate rule without repeatedly scanning the whole lattice structure and take advantages on the normal method in terms of the calculating speed and complex.(2) Secondly, this paper presents a new algorithm in calculating lattice which can, directly and precisely, extract the exact rules according to the given SLIT and also can extract approximate rule as well according to the sub-concept of SLIT.(3) Thirdly, this paper presents the access to a calculation algorithm based on the approximate rule of quantity lattice. The parallel calculation is an effective way to speed up the effectiveness through analyzing the calculation steps and grading analysis, which could mix up the partial association rules into the whole association rules ultimately.(4) Finally, this paper enlarged the platform of P2P-MPI and achieved the new algorithms by utilizing JAVAMPI on ShengTeng 1800 cluster system. Some specific analysises are also discussed in this paper.