Research On Payment Modeling In China National Automatic Payment System

The goal of this paper is to design and implement the modeling system of payment information in China National Automatic Payment System.So the key technology for modeling the main topic of this paper.According to the fact of payment information, this paper discusses background knowledge representation of payment information, concept mapping,optimization of traditional clustering algorithm,clustering algorithm based on concept network,estimation of rules in association rule results etc.On background knowledge representation, this paper at first talks about the common ontology concept, and then analysis the relationship between the ontology in payment modeling system and that in payment process system. To discover more significant model, the concept network is put forward to present background knowledge of payment information. On concept mapping, the usual methods of concept mapping are summarized firstly. Then two new ways are debated.One is based on conceptual normalization, the other is based on statistics for the same one concept.Subsequently, the algorithms of the two new means are discussed.On optimization of traditional clustering algorithm, the clustering results are firstly compared through experiments between from k-means algorithm which is a classic algorithm based on distance and from EM algorithm that is based on possibility computation.The difference of comparison indicates that the EM algorithm is better than algorithm based on distance.Later this paper brings forword an improved EM algorithm which only scans database only one time to get the clustering results.This notably increases the processing efficience of algorithm on large database.And the qualities of advance EM algorithm is better than satandard EM algorithm naturally, which is demostrated by followed experiments.At the same time, this paper discusses the two steps of clustering based on concept network.The first step is to cluster the attribution of nodes in concept network and the second is to cluster the relationship among network nodes.This kind of cluster way is to avoid to lose important business information while this is the shortcoming of traditional clustering methods based on database.When research on modeling of capital flow, current main relationship discovered algorithms are at first compared.And at last the method of association rule is the best choice to modeling capital flow in payment.Additional,a new way is introduced into association rule method to computer the importance of rules to gain more accurate results.In the end of the paper,the structure of system for payment modeling and the structure of modeling algorithm system are discussed.And on the bases of above discussed technology,a prototype system is...