58 results about "Community mining" patented technology

The invention relates to an evaluating system and method for community influence in a social network. The method comprises the steps that a social network chart with social network users as nodes and user relationships as sides is built; according to the social network chart, the community structure of the social network is obtained by carrying out community division through the label propagation algorithm; according to the social network chart and matrixes which communities belong to, the parameter of the community influence is calculated, and the initial influence of each community is generated; according to the transmission probability model of the influence, an influence transmission probability matrix is generated; according to the influence transmission probability matrix and the community influence iterative computation model, the community influence is iterated and upgraded until the iteration end condition is met, the influence value of each community is obtained, and the sequence of the community influence, namely, the influence estimation result of each community in the social network is obtained after normalization. The system and method can effectively analyze the distribution of the community influence in the social network and can be used for high-influence community mining, thereby being capable of being applied to the fields of network marketing and the like.

In community mining based on core objects and affiliated objects, a set of core objects for a community of objects are identified from a plurality of objects. The community is expanded, based on the set of core objects, to include a set of affiliated objects. According to one aspect, a model of a community of objects is obtained by grouping a first collection of a plurality of objects into a center portion, and grouping a second collection of the plurality of objects into one or more concentric portions around the center portion. The groupings of the first and second collections of the objects are identified as the community of objects.

The invention provides a complex network community mining method based on local minimum edges. The complex network community mining method comprises the following steps: obtaining an adjacent matrix A of a complex network; calculating a similarity matrix R; carrying out community detection C to a complex network diagram G; looking up one group of local minimum edges; detecting local topological structures on two end points of each local minimum edge, and confirming and removing the local minimum edges which enable a current community structure to more conform to a community definition; detecting whether a new connected subgraph appears in the network, recalculating a weight of each edge if new connected subgraph does not appear in the network, and judging whether division is reasonable if new connected subgraph appears in the network; and if the division is unreasonable, outputting a result, recalculating the weight of each edge if the division is reasonable, and carrying out a next iteration process. The invention has the characteristics of being high in precision, high in speed and good in universality.

The invention discloses a complex network community mining method based on an improved genetic algorithm and belongs to the technical field of complex network community mining method research. The complex network community mining method based on the improved genetic algorithm uses the improved genetic algorithm based on clustering and double population thought fusion to mine communities in a complex network. The complex network community mining method based on the improved genetic algorithm uses a normalization common information similarity standard as the standard for measuring the similarity between individuals in the population and fuses the clustering and double population thought. The complex network community mining method based on the improved genetic algorithm includes that introducing the clustering thought, using a minimum spanning tree clustering method to classify the population, introducing the double population thought, and determining the main type and auxiliary type for the clustering. The main type maintains the population evolution direction to get close to the optimal solution of an objective function; the auxiliary type is mainly used for duly providing diversity for the main type so as to enable the main type to be capable of coming out to search the other solution space to realize the complex network community mining when the main type is located at the local optimum.

The invention discloses an abnormal account detection method and device. The method comprises the steps of obtaining a to-be-analyzed sample data set; obtaining associated account information and equipment information from the sample data set; obtaining identification information of equipment from the equipment information, and establishing a complex network between the identification informationof the equipment and accounts associated with the identification information according to the equipment information; carrying out community mining on the complex network according to a preset community mining algorithm, thereby obtaining a plurality of community clusters, wherein each community cluster comprises a plurality of associated account nodes; and extracting target community clusters of which number of account nodes is greater than a preset threshold from all community clusters, and determining accounts contained in the target community clusters as abnormal accounts. The problem thatin the prior art, a conventional abnormal account identification method is liable to lose efficacy is solved, and the purpose of relatively comprehensively and deeply identifying the abnormal accountsare realized.

The invention discloses a network community mining method and device, a computer device and a storage medium. The method comprises: acquiring nodes corresponding to claim settlement data, and the nodes corresponding to the claim settlement data being divided into a plurality of sub-graphs in parallel through spectral clustering; respectively carrying out networking on the plurality of subgraphs toobtain an initial claim settlement social network topological graph; and clustering the initial claim settlement social network topological graph through community detection to obtain a network community. According to the method, efficient real-time networking and network cutting are performed on the total data of the claim settlement data, and the network community is quickly obtained based on the community discovery algorithm, so that suspected fraudulent gangs can be searched based on the community aggregation.

The invention relates to the technical field of financial insurance, in particular to a product data pushing method based on knowledge relation mining and related equipment. The method comprises the steps of accessing a user insurance policy database, acquiring multiple user IDs and a user address book, and marking the user address book to generate an address book ID; obtaining a user portrait corresponding to the user ID from a user portrait database, and generating a user knowledge map; defining each entity in the user knowledge graph as a node, calling a preset community mining algorithm, and carrying out community division on the plurality of nodes to obtain a plurality of optimal communities; and obtaining attributes of all nodes in the community, screening out one or more items withthe most common characteristics, and pushing the product data corresponding to the common characteristics. According to the invention, a community mining algorithm is introduced, users are divided into a plurality of communities, product marketing is carried out by using common attributes of the users based on the characteristic that personnel in the communities have homogenization, and the insurance marketing efficiency is improved.

The invention discloses a river runoff prediction method mainly for runoff prediction for PUBs (prediction ungauged basins). The basic principle of the method refers to: using a complex network to mine topological characteristics of a hydrological spatiotemporal sequence, and performs runoff prediction on ungaged basins on such basis; a runoff complex network model is established according to runoff data of a monitoring network station, FN (fast Newman) algorithm is used to perform community mining on such basis, and candidate nodes are selected based on community mining results. The method of the invention considers relevancy between basin division and PUBs, common nodes and characteristic nodes are selected as candidate nodes, and runoff of a station to be predicted is predicted by means of transplantation process. The ungauged basin runoff prediction method considering both relevancy of runoff data topological structures and the runoff data itself is provided herein.

The invention belongs to the technical field of evolutionary computation and complex network community mining and discloses a complex network community detecting method based on prior information and network inherent information. The method is mainly used for community division of complex networks. The method comprises the steps of establishing a network adjacent matrix, initializing the population by means of adjacent matrix information, conducing preprocessing according to the inherent information of the adjacent matrix to reduce invalid searching, optimizing a modularity function Q, conducting gene interlace operation and mutation operation, using the local search with mutation operator method (LSMM) based on variation and network inherent information, and testing a community division result by means of an evaluation function NMI. According to the method, the community network is detected by fully utilizing prior knowledge and inherent information contained in the network adjacent matrix, an optimal solution is obtained more effectively by means of the LSMM based on variation and network inherent information, and the community structures of a real world network and a synthetic network can be better found compared with an ordinary genetic algorithm.

The invention provides a reactive power partition method based on reactive power source-charge numbers and community mining and belongs to the technical field of power system automation. Basic parameters are input through programs by means of a computer, then a power system reactive equivalent lossless network is determined, then reactive power, reactive source-charge numbers and reactive coupling modularity of all power transmission routes in a power system are determined, the optimal reactive voltage partition of the power system is determined by means of the traditional genetic algorithm, and finally the partition reactive coupling degree and the reactive balance degree of the optimal partition scheme are inspected. The method has the advantages that in the reactive voltage partition process, influences of reactive power output by a reactive power source on reactive loads through the transmission routes between reactive source-load node pairs are taken into consideration, reactive coupling performance of reactive source-load nodes in partitions can be strong, reactive coupling performance of the reactive source-load nodes between the partitions can be weak, and on the basis that reactive balance in the partitions is guaranteed, global optimality of the partitions is ensured from the perspective of reactive transmission. The method can be widely used in the reactive voltage partitions of the power system and provides scientific bases for reactive voltage control of the power system.

The invention discloses a complex network community mining method based on a cellular automatic learning machine. The method comprises the six steps of initializing the cellular automatic learning machine, generating state vectors of the cellular automatic learning machine, decoding the state vectors of the cellular automatic learning machine to obtain the corresponding communities, calculating response signals, updating the cellular automatic learning machine, and comparing community structures. According to the complex network community mining method, the whole network is modeled into the cellular automatic learning machine, the communities in the network are mapped into the state vectors of the cellular automatic learning machine, and the network is searched for the optimal community structure through iterative updating of the cellular automatic learning machine. The complex network community mining method is quite low in time complexity and applicable to a large-scale complex network. In addition, the complex network community mining method has the good performance for searching for the globally optimal solution, and can ensure compactness of local communities, so that community detection precision is quite high.

The invention belongs to the technical field of social network and computer application, and discloses a relation combination optimization and seed expansion-based multi-relation community discovery method. Multiple relation networks are fused into a low-noise single relation network capable of effectively integrating information of relation communities by optimizing a weight ratio of various relations in the networks; then community division information of the relations in the multiple relation networks is integrated for finding out people of the same communities in the relations; and small communities consisting of the people are taken as seed communities, and the multiple social relation networks are subjected to community mining by adopting a seed expansion policy to obtain community structure division with higher accuracy. An experiment shows that compared with a conventional method, the method provided by the invention has the advantages of high result accuracy and good anti-noise capability.

The embodiment of the invention provides a fraud information detection method and system based on community mining. The fraud information detection method comprises the following steps: acquiring a plurality of subject cluster communities and a plurality of behavior information cluster communities; then, according to a main body object behavior label corresponding to a target main body object, obtaining a reference main body object corresponding to the target main body object, and according to an information category label corresponding to the target behavior information, obtaining reference behavior information corresponding to the target behavior information; and finally, respectively calculating an association parameter between the target main body object and each piece of behavior information in the behavior information sequence, and detecting whether suspected fraud information exists in the behavior information corresponding to the target main body object in the behavior information sequence according to the association parameter. Thus, the suspected fraud information is detected based on the data mining mode of the main body cluster community and the information cluster community, and whether the suspected fraud information exists in the behavior information corresponding to the target main body object or not can be effectively detected.

The invention provides a subject community mining method for an online social networking service.The subject community mining method is based on nonnegative matrix factorization (NMF).User node link information and content information can be integrated in a unified mode through an NMF model, an affiliation matrix between user nodes and communities and a correlation intension matrix between communities and subject feature words are obtained with the matrix approximative decomposition method, and then subject community mining can be directly conducted by means of matrix decomposition information.By the adoption of the method, user node link information and content information can be processed with a unified model, mining is easier and more efficient, mining quality is higher, and therefore the method is more suitable for being actually applied to mining of subject communities in the online social networking service.

The invention belongs to the technical field of networks, and provides a community mining method and device based on statistic models. The method comprises the steps of reading an adjacent matrix A of a signed network N, setting a variation range of a community number to be [Kmin, Kmax], and initializing the community number K to be equal to Kmin, wherein the total number of nodes of the signed network is n, and Kmin and Kmax are integers within the range of n; initializing a statistic model NMK corresponding to each community number K, performing fitting on the statistic models NMK and the signed network N, and calculating a selection standard HK of each statistic model NMK; comparing the selection standards HK of all the statistic models NMK, and selecting the statistic model NMK with the maximum selection standard HK as the optical model NMoptim; according to the optical model NMoptim, determining the community to which each node i belongs in the signed network N, wherein 0<i<=n. According to the community mining method and device based on the statistic models, community mining of the signed network based on the statistic models is achieved, and the signed network community mining accuracy is effectively improved.

The invention provides a user community mining method and system based on differential privacy, which are used for solving the problem of low data availability caused by a privacy protection technology in an actual environment. The method comprises the following steps: firstly, segmenting a generalization movement track sequence into generalization track segments according to an original movement track of a user, and constructing a distance function of the generalization track segments; secondly, respectively quantifying similarity weights between the generalization trajectory segments from geographic space and semantic space, and constructing an optimal generalization trajectory segment selection model; optimizing the optimal generalization track section selection model by utilizing a track sequence function generation algorithm to obtain a generalized track sequence; and finally, publishing the generalized track sequence to a central server, and mining potential user communities by the central server according to the semantic distance and the geographic distance between the tracks. According to the method, the user community discovery of privacy protection is realized, the privacy protection intensity is quantified by using the differential privacy protection method, and the reliability and controllability of the system are improved.

The invention discloses an Overlapping community division method based on COPRA, relates to the field of computers, and aims to improve the instability and randomness of a COPRA algorithm in allusionto the problem of low accuracy of an algorithm result caused by an unstable division result and high randomness of a social network division algorithm in the prior art. Firstly, an information entropyconcept is introduced, direct node and indirect node concepts are provided, then node label updating is sorted according to the sum of node entropies by obtaining the node entropies of nodes, and therandomness of label updating is reduced. Then, a label value concept is provided, the subordinate coefficient of the label, the entropy sum of the nodes with the label and the degree of the nodes arecomprehensively considered mainly from the nodes, a label value calculation formula is provided, and the execution process of the algorithm is given. The overlapping community structure in the complex social network can be discovered, the community mining result is good, and the accuracy is high.

The embodiment of the invention provides a real-time fraud detection method and device based on a complex network, and electronic equipment. A large-scale image storage database is created based on user data; based on a large-scale image storage database and a preset scoring model, a multi-hop query mechanism is adopted and information is associated with fraud characteristics, user fraud detectionis carried out, the problem that data is too much, the model operation is slow and detection efficiency is low is solved, real-time anti-fraud detection is realized; real-time defense is realized, the fraud detection efficiency is improved; a single-entity knowledge graph network and a multi-entity knowledge graph network are constructed based on a large-scale graph storage database; based on a single-entity knowledge graph network and a multi-entity knowledge graph network, a graph propagation algorithm is adopted, and community mining is performed according to information with fraud characteristics, so that the fraud gang identification range is quickly expanded, fraud gangs are deeply mined, and the fraud detection efficiency is improved.

The invention relates to a malicious domain name community mining method based on a graph structure, which comprises the following steps: carrying out cleaning and rule filtering on an input DNS (Domain Name Server) log, and constructing a dependency graph among domain names according to data of three basic fields (a client, a timestamp and a server); on the basis of the constructed graph structure, discovering domain names with high dependency from the graph by using sequence similarity and jump probability of nodes, and mining domain name communities; and selecting a specific feature to carry out training learning on whether the community is malicious or not to obtain a classifier which is finally used for judging and screening out the malicious domain name community. Through adoption of the method, expansion detection of malicious domain name detection in an Internet security system can be met, the effect of malicious domain name cluster mining is achieved, the cost of network criminals is improved, and the method and the device are more suitable for increasingly complex Internet scenes in the future.

The invention discloses an overlapping community mining method based on multi-tag propagation, belongs to that technical field of the complex network, and can mine the overlapping community structurefrom the network. The method comprises the following steps: firstly, the importance of the nodes in the network is sorted by using the LeaderRank method so as to determine the update order of the nodes and assign a unique tag to each node; secondly, the similarity of all the nodes is calculated to obtain a similarity matrix, the tag update strategy is designed based on the similarity between the nodes, and tag propagation is carried out through tag update strategy; and finally community partitioning is completed, the nodes with the same tag belong to the same community and the nodes with multiple tags belong to overlapping nodes. Compared with the traditional multi-tag propagation overlapping community mining method, the overlapping community structure can be more quickly and accurately marked out.

The embodiment of the invention provides a pushing method and system supporting fragmentation learning of an open source project. The method and system specifically comprise the steps: constructing aknowledge graph of the open source project; obtaining a learning entrance node of the knowledge graph; taking the learning entrance node as a path starting point, taking a target node with a learningrequirement of a user as a path terminal point, respectively obtaining a plurality of reachable paths, and selecting a path with the widest coverage as a target learning path; cutting a peripheral sub-graph involved in the target learning path to obtain a target sub-graph; performing community mining on the target sub-graph to obtain a plurality of graph network communities, and respectively generating target learning contents; and pushing the target learning content to the user in sequence in a fragmentation manner. According to the method and the system provided by the embodiment of the invention, the knowledge graph of the open source project is constructed, the learning entrance is analyzed, the learning path is recommended according to the user demand, the fragmented learning contentis generated and pushed to the user, and the learning efficiency of a developer for the strange open source project is improved.

The invention discloses a network graph segmentation method, which comprises the following steps of: randomly taking one node in an obtained network graph node set as a first node, traversing all neighbor nodes of the first node to serve as a first neighbor node set, traversing all neighbor nodes of the first neighbor node set to serve as a second neighbor node set; deleting a first node, a firstneighbor node set and a second neighbor node set in the node set, and adding the first node into an isolated set; repeating the above steps until the node set becomes an empty set, obtaining an isolated set after segmentation, continuing to obtain all network node sets in the graph, and deleting network nodes contained in the isolated set to obtain a remaining node set; and repeating the steps until the residual node set is zero. By means of the method, when the community mining algorithm is used for social network analysis, the operation speed can be increased, the analysis efficiency can beimproved, and the mining and analysis capacity for large-scale social networks can be improved.

The invention discloses a community discovery algorithm based on an improved association rule, and the algorithm comprises the steps: firstly carrying out the self-adaption of a support degree, and calculating the minimum support degree through a mathematic method; secondly, introducing a Boolean matrix and a transaction weight thought to improve an Apriori algorithm, and reducing the database scanning frequency; and finally, combining with a Spark platform to realize association rule improved community discovery algorithm parallelization. According to the community discovery algorithm based on the improved association rule, the community members are mined by using the MAC address. The Apriori algorithm is improved by introducing the idea of support degree self-adaption and adding a transaction weight to generate a Boolean matrix, the improved algorithm is combined with Spark to realize parallelization of the algorithm, and the relationship between community members is mined by mininga frequent item set. Experimental results show that the ARCD algorithm solves the problems of subjectivity of manual setting of support degree and redundancy of community mining results, has good expandability, and improves the mining speed of community discovery.

The invention provides a community mining method and device, equipment and a storage medium, and relates to the technical field of data mining, and the method mainly comprises the steps: obtaining a plurality of pieces of transaction data, the transaction data being used for recording transaction behaviors between users, and the plurality of pieces of transaction data comprising annotation data; a network topological graph corresponding to the multiple pieces of transaction data is constructed, nodes of the network topological graph are users, and an edge connecting two nodes in the network topological graph indicates that transaction behaviors exist between the users; generating a feature vector of a feature corresponding to each piece of transaction data according to the network topological graph; according to the feature vector, a transaction data set to which the annotation data belongs is searched, the transaction data set comprises the annotation data and target data, and the target data comprises transaction data associated with the annotation data; and determining the transaction data set meeting a preset condition as a community mining result. According to the community mining method and device, the equipment and the storage medium provided by the invention, the efficiency and accuracy of community mining can be improved.

The invention provides an unstructured P2P botnet detection method and device based on SAW community discovery. The method comprises the following steps: step 1, converting original flow data in a pcap format into flow data in a netflow format; 2, converting the flow data in the netflow format into a quintuple cluster flow by using an F-link big data platform, and filtering from the quintuple cluster flow to obtain a P2P (Peer-to-Peer) cluster; 3, calculating weights of shared neighbor nodes among the P2P clusters by using a Jaccard coefficient, and constructing a shared neighbor graph; step 4, using SAW to access each vertex in the shared neighbor graph, generating a vertex matrix between the vertexes, using principal component analysis (PCA) to perform dimension reduction on the vertex matrix, calculating Bray-Curis dissimilarity, using hierarchical clustering to calculate node similarity, and performing community mining, thereby clustering the P2P nodes of the same kind; and 5, classifying clustering results by using community attributes, and filtering out the botnet.

The invention provides a community mining method. The community mining method comprises the following steps: acquiring a first data subset associated with an annotation data set in a first data set; constructing a first community based on the first data subset; compressing each community included in the first community into a node, acquiring a second data subset associated with the node in the first data set by taking the characteristics of the data in the first data set as granularity, and constructing a second community based on the second data subset, until the data in the first data set cannot be associated with the node, or a community result constructed based on the first data set does not change any more; and outputting the community result. The invention further provides a community mining device, electronic equipment and a storage medium, and community mining can be efficiently carried out through the community mining method and device, the electronic equipment and the storage medium.

The invention relates to an operation area site selection and layout method for shared vehicles in a free flow mode, and the method comprises the following steps: constructing a user travel network according to the historical travel data of a user based on a candidate map; then, adopting a community mining algorithm to mine closely related community areas from the user travel network, and specifically representing a single community area as a grid set; and finally, respectively calculating boundary fences corresponding to different community areas to obtain each operation area of the shared vehicle. Compared with the prior art, a distance attenuation method is adopted for the shared automobile in a free flow mode, the area of the community area can be effectively controlled, the meticulousdivision of the community area and the continuity in the community are ensured, and the area of the community area can be effectively controlled by calculating the electronic boundary fence of the grid set corresponding to the community area. The corresponding boundary range of the operation area can be rapidly and accurately obtained.

The invention provides a community mining method. The community mining method comprises the steps of obtaining node information of a newly added edge or a deleted edge in a relational network; performing local community division on the newly added edge or the deleted edge according to the node information of the newly added edge or the deleted edge to obtain a first community result in the relational network; and updating an original community result in the relational network according to the first community result to obtain a total community result. According to the method, community analysisis carried out on the relational network of the big data, more accurate results can be obtained while resources and time expenditure are saved, and the pushing accuracy is improved.

The invention discloses a user identity recognition method and device, electronic equipment and a storage medium, and the method comprises the steps: taking the information feature of each user as a network vertex, connecting two network vertexes with the same feature to form edges, and building a network diagram; dividing the network graph into a plurality of communities according to the aggregation degree of network vertexes, strong association users and weak association users in each community are determined, and identifying hidden merchants according to the types and occurrence frequenciesof release behaviors of the users. Therefore, according to the method, hidden merchants can be effectively identified and mined, and associated similar users are identified in a community mining mode, so that the identification method is prevented from being discovered and evaded, and the user identity identification efficiency is improved.