Mining And Application Of Characteristic Information In Fuzzy Complex Network Based On Factor Space Theory

This article researches on feature information mining problem in fuzzy complex networks and forms corresponding models and algorithm, based on factor space theory.This article starts from axiomatized definition of factor space theory, analyzes the application of background relationship, gives out concept tree generating algorithm and reasoning tree generating algorithm, and ascertains that the switching factor of factor space cane means a division of root node universe of discussion.This article gives out the definition of fuzzy complex networks, divide fuzzy complex network into point set fuzzy network and edge set fuzzy network as well as indicating fuzziness of vertex is more essential than edge set fuzziness and defining general fuzzy network through edge set fuzziness. In this article, we study the key objects of fuzzy network point - edge relation: weighted matrix, clustering distance matrix, fuzzy adjacency matrix and fuzzy weighted matrix. On this basis, 16 key indexes of fuzzy network are defined from three angles of vertex distance, local properties of network and vertex degree distribution, which constitute the key index system of fuzzy network.By introducing the membership degree of fuzzy edge and the definition of weighted well order, the basic properties of the fuzzy network are obtained by the standard synthesis function: the elements of N1-n times of the fuzzy adjacency matrix to the power of n (N is the number of vertices) represent the membership degree of standard additive synthesis function of the path whose even edge number between the two vertices is n; the shortest path formed by a fuzzy edge corresponding to a maximum degree membership path on the network. The key object on the network can be represented by the factor space cane from universe of discussion, so that the basic model and the generation algorithm of the fuzzy network are given.After the explicit feature information is formed by the binary switching factor, this article proposes the algorithm of generating hyper-vertex concept tree and kernel concept tree based on dynamic clustering and system clustering model of vertex set for static fuzzy network, produces the representation of internal and external universe of discussion kernel element of the fuzzy object concept. Based on the vertex clustering class, the hyper-vertex link prediction algorithm is proposed. On this basis, the class is merged into the community, and the reasoning tree is generated in the community and the kernel community. All the constantly true reasoning statements in the background relationship on the fuzzy network are obtained. The article indicates that the kernel element on the universe of discussion is equivalent to all the information kernels of the factor space cane that are composed of binary switch factors. For time sequence fuzzy network, this article proposes a method of vertex clustering on the online updating network and a community detection method which considers the hyper-vertex at all times, so as to unify the analysis method of time sequence fuzzy network and static fuzzy network.For the static fuzzy network, this article takes the "medium power nation" as the target concept of feature information mining analysis case; for the time sequence fuzzy network, this paper gives a stock selecting method based on historical returns. This article expands the application of fuzzy network: constructs a multi-drive investment decision model, based on fuzzy network vertex clustering method and discriminant analysis as well as conditional probability model and multi-objective genetic algorithm.This model will be applied to investment decision making of education funds; In signal analysis, take signal sample that is sampled according to cycle as the distance between classes, extend the RI index of hyper - vertex similarity to the contribution coefficient of signal fluctuation and produce a signal cycle discriminant model with an average correct rate of 85%.