99 results about "Cluster coefficient" patented technology

The invention discloses a multimodal brain network feature fusion method based on multi-task learning, and the multimodal brain network feature fusion method based on the multi-task learning includes the steps of preprocessing the obtained functional magnetic resonance imaging (fMRI) images and diffusion tensor imaging (DTI) images, registrating the preprocessed fMRI image to the standard AAL template, carrying out a fiber tracking for preprocessed DTI images, calculating fiber anisotropy (FA) value, and constructing structure connection matrix through the AAL template. Clustering coefficient of each brain area in a function connection matrix and the structure connection matrix is calculated to be regarded as function features and structure features. As two different tasks, the function features and the structure features assess an optimal feature set by solving the problem of multi-task learning optimization. The method uses information with multiple modalities complementing each other to learn simultaneously and to classify, improves the classification accuracy, solves the problems that a single task feature does not consider the correlation between features, and the fact that only one modality feature is used for pattern classification can bring to insufficient amount of information.

The invention provides an electroencephalogram signal characteristic extracting method. Network average route lengths and clustering coefficients are calculated through wavelet reconstruction, windowing horizontal visibility map complex network conversion and complex network analysis. The average route lengths and clustering coefficients composed of electroencephalogram signals are calculated to achieve characteristic analysis of electroencephalogram signals and chaotic time sequence signals of the electroencephalogram signals of different rhythms. The electroencephalogram signal characteristic extracting method has the advantages that one-dimensional chaotic time sequences are converted into complex networks, according to analysis of network characteristic parameters, fractal characters of the electroencephalogram signals are revealed, the complex non-linearity signals of the electroencephalogram signals are depicted from a brand new angle. The electroencephalogram signal characteristics can be applied to automatic diagnosis of mental disease and a characteristic identifying module of a brain-machine port system. The electroencephalogram signal characteristic extracting method can effectively distinguish the electroencephalogram signals of an epilepsia attach stage and an epilepsia non-attach stage, the equation p<0.1 is met after Mann-Whitney detection, and the electroencephalogram signal characteristic extracting method can be applied to epilepsia electroencephalogram automatic identification.

The invention discloses a method for identifying key proteins in a protein-protein interaction network. According to the method, an undirected graph G is constructed according to the protein-protein interaction data, and the edge clustering coefficient of the graph is calculated. Compared with the prior art, the method provided by the invention has the advantages of combining the gene expression profile data and the gene function annotation information data on the basis of considering the topological structure characteristics of the protein-protein interaction network, and integrating three groups of data to predict the key proteins, so that the influence caused by the data noise of a single data source on the prediction correctness can be effectively decreased, and the key proteins in the network can be predicted through the key protein characteristics embodied by three types of data, such as the edge clustering coefficient in the protein-protein interaction network, the Pearson's correlation coefficient of the gene expression value and the gene function similarity index. According to the method, the identification correctness of the key proteins in the protein-protein interaction network can be remarkably improved, and abundant key proteins can be predicted once, so that the problem that the biological experiment method is high in cost and time-consuming is solved.

The invention relates to an edge clustering coefficient-based social network group division method. Specifically, the method comprises the following steps: reading the social network data; constructing a social network planning which takes social network users as nodes and takes user relationships as edges; randomly endowing each user with a unique label value; updating the labels of the user nodes by adopting an edge clustering coefficient-based label propagation algorithm; and after several iteration, owning, by the tightly connected nodes, same specific label value. By adopting the social network group division method, the user groups are divided through improving the label propagation algorithm according to the edge clustering coefficient attribute of the user relationship graph, and the division result has preferable application value for the monitoring of network public opinions and the searching of commercial customers.

The invention relates to a link prediction method and device based on node similarities; the method comprises the following steps: carrying out node and link expression for a to-be-parsed network; obtaining two nodes having no direct link connection in the network; respectively reading a neighbor node set of the two nodes having no direct link connection; obtaining the intersection of the neighbor node sets of the two nodes having no direct link connection, thus obtaining the common neighbor set; considering the common neighbor set as a subnet, and calculating the subnet global cluster coefficient and a random common neighbor node cluster coefficient in the subnet; calculating the node similarity of the two nodes having no direct link connection according to the subnet global cluster coefficient and the random common neighbor node cluster coefficient in the subnet; carrying out link prediction according to the calculated node similarity. The method can parse the mutual relation of the nodes in the complex network local structure from the cluster coefficient angle, thus defining the node similarity calculating new index based on the local cluster coefficient.

The invention relates to a priori knowledge based microblog user group division method. The method specifically comprises the steps of reading social network data; constructing a social network graph taking social network users as nodes and user relationships as edges; constructing a user similarity matrix; when labels of user nodes are initialized, giving the same labels to the nodes with high similarity, and updating the labels of the user nodes by adopting a label propagation algorithm; in a label propagation process, when a plurality of labels with highest frequencies exist in neighbor nodes of the updated nodes, randomly selecting a label with highest frequency to the label of the corresponding node; and after multi-step iterative updating, ensuring that the closely connected nodes have the same specific label values. According to the social network group division method provided by an embodiment of the invention, user groups are divided by improving the label propagation algorithm according to edge clustering coefficient attributes of a user relationship graph, and a division result has relatively high application values for network public opinion monitoring, commercial user mining and the like.

The invention discloses a key protein predicating method based on priori knowledge and network topology characteristics. Based on the analysis of the topological relation between known key proteins, the close relation between key proteins is found; edge clustering coefficients are taken as parameters for assessing the close degree of two key proteins; and parts of known key proteins and the shared-clustering coefficients of neighbor nodes and the known key proteins are utilized for predicating new key proteins. The key protein predicating method is simple to implement and unknown key proteins can be accurately predicated just according to PPI (protein-protein interaction) information and information of parts of known key proteins, the method is applicable to not only non-weighted PPI networks but also weighted PPI networks, and the problems such as high cost and time wasting of a chemical experiment method are solved.

A marine observation big data visualization analysis method based on a complex network comprises the steps of performing grid division on original marine observation big data, constructing daily average data in a grid into a single Gaussian model and a mixed Gaussian model, and obtaining nodes represented by probability feature vectors; Determining the similarity between any two nodes in the single Gaussian network and the multi-Gaussian network to obtain a similarity matrix; And setting a threshold value to obtain an adjacent matrix, calculating the degree, the clustering coefficient and thenode betweenness of each node according to the adjacent matrix, and visualizing or drawing the degree, the clustering coefficient and the node betweenness on double logarithm coordinates or on a map.According to the invention, the Gaussian mixture model is combined with the complex network theory for the first time; The invention provides a marine observation big data analysis and visualization method, the fluctuation of ocean motion reflected on the data is restored to the maximum extent, and model parameters are used for expressing high-dimensional ocean data, so that the defect that a network model constructed on the basis of Pearson similarity can only measure time sequence data is overcome, and the calculation speed is also improved.

The invention relates to a clustering and multi-hop communication method of network nodes of a wireless sensor, in particular to a clustering and multi-hop communication method of a wireless sensor network based on a small-world model. The algorithm utilizes the small-world model to cluster the sensor network, which improves the clustering coefficient at the network side and enables that the sensor network has favorable performances of small-world network and large clustering coefficient; nodes in each cluster and average hop numbers are restricted so that the load of each cluster can be balanced; and the node with largest residual energy can take turns a cluster head so that the energy of all nodes can be oppositely balanced. The invention can balance the energy consumption of the networknodes in the whole situation, avoids the situations that the network is partitioned into a plurality of complementary connections as a part of nodes are premature failure, prolongs the life cycle ofthe whole network and improves the robustness of the network simultaneously.

The invention discloses a protein biological network motif identification method integrating topological attributes and functions. The protein biological network motif identification method integrating the topological attributes and functions (Ecc-GOSS) is based on the biology significance of a motif, and comprehensively evaluates the biological significance of protein-protein interaction by integrating edge clustering coefficients and semantic similarity of GO phrases. The protein biological network motif identification method integrating the topological attributes and functions is easy to implement, a great amount of network motifs with biological significance can be identified accurately according to PPI information and gene ontology information, and the protein biological network motif identification method integrating the topological attributes and functions has good robustness on high-percentage false positive prevailing among the large-scale data of the protein-protein interaction.

The present invention discloses a group degree based sorting method and model evolution method for important nodes on complex network. The method comprises: first, obtaining each order of group degree of each node on a complex network; then calculating each order of overall group degree of the complex network; normalizing each order of overall group degree; calculating a weight of each order of group degree according to a normalization result; finally, each node performing weighting on each order of group degree of the node according to the weight, wherein a result is an importance value of the node; and sorting each node according to the importance value. During model evolution, each time a new node is added, a connecting node of the new node is selected according to the importance values of existing nodes. According to the sorting method and model evolution method provided by the present invention, the importance value of the node is calculated based on the group degree, the obtained node importance sequence is better in line with the actual situation of the network, and the average path length and cluster coefficient property obtained through model evolution both have obvious advantages.

The invention discloses a clustering method, a system and a medium for automatically confirming the number of clusters based on a coefficient of variation, wherein, the density value of each data point in a data set is calculated, the density index is calculated according to the density value, and the data point with the largest density index is selected as a first clustering center; Calculating the shortest distance between each data point and the existing clustering center, calculating the probability that each data point is selected as the clustering center according to the shortest distance, and preselecting the clustering center according to the roulette disc method; Until the set cluster centers are selected, the initial cluster centers selected are used for k-means clustering to generate a corresponding number of clusters; Calculate the average intra-cluster coefficient of variation and the minimum inter-cluster coefficient of variation, then calculate the difference between theaverage intra-cluster coefficient of variation and the minimum inter-cluster coefficient of variation, compare the difference with the set value, and if the difference is less than the set value, merge the two clusters with the minimum inter-cluster coefficient of variation; Until the difference is greater than or equal to the set value, the clustering result is output.

The invention provides an energy consumption calculating and scheduling method of an urban rail traffic system. The method includes the steps of establishing a multi-layer network model of the urban rail traffic energy consumption system on the basis of the urban rail traffic system; calculating topological properties and data properties of nodes as input, wherein the topological properties include the degrees, betweenness and clustering coefficients of the nodes, and the data properties include the energy consumption and strength of the nodes; fusing the topological properties and data properties of the nodes through an OWA operator to obtain the weights of the nodes in the urban rail traffic system; calculating the energy efficiency emergence state, namely the energy efficiency representation parameter, of the urban rail traffic system in combination with the weights of the nodes through Choquet integrals on the basis of the emerging theory; regulating and controlling the energy efficiency of the urban rail operation system on the basis of the energy efficiency representation parameter of the urban rail system. For the complex-structure urban rail traffic system, the energy efficiency of the urban rail system can be calculated by means of the method, and the reasonable regulation and control of the energy efficiency of the urban rail operation system are realized in combination with the calculated energy efficiency.

The invention discloses a method for optimizing outburst-prevention technical measures for a coal and gas outburst mine. The method comprises the following specific steps: 1), various outburst-prevention technical measures are made according to the on-site condition of a coal mine; 2), a value of a cluster evaluation index under each outburst-prevention technical measure is determined; and a cluster evaluation index matrix is established; 3), the outburst-prevention technical measures made in the step 1) are divided into multiple grades, and white functions of each cluster evaluation index in different grades in the technical schemes are determined; 4), the weight coefficient of each cluster evaluation index is determined; 5), index fixed weight cluster coefficients are solved, and a fixed weight cluster coefficient matrix is established; 6), the largest coefficient in the fixed weight cluster coefficient matrix is found, and the corresponding outburst-prevention technical measure is the optimal technical scheme. With the adoption of the method, the optimal outburst-prevention technical measure for the coal mine can be acquired, the gas treatment cost is reduced, the drilling workload is reduced, drilling construction time is shortened, and a basis is laid for increase of gas suction capacity on a coal seam and comprehensive treatment of gas in the mine.

The invention discloses a method for measuring the contribution degree of a developer during the development of open source software, and the method comprises the following steps of: S11: selecting an open source project and downloading full version control information in the open source project; S12: disposing data and setting up data sheet SvnInformation; S13: structuring a developer network; S14: calculating the degree of the developer; S15: calculating the document submit times of the developer; S16: calculating the clustering coefficient of the developer; and S17: calculating the contribution degree of the developer. The method can be used for providing a practical management method for the project manager, and can be used for performing active and effective digital management on the situation of the development project and the contribution degree of the developer.

The invention provides a circuit health ranking evaluation method in combination with dependency relation and gray clustering technology, and belongs to the technical field of fault diagnosis. The method comprises the following steps: firstly determining an observation index and the health rank of the circuit, establishing a dependency graphical model to obtain a dependency matrix, using the dependency matrix to determine the optimal observation index and the weight thereof, and establishing a whitening function of the observation index through simulation, and finally acquiring an observation value of a circuit to be detected object, performing clustering coefficient computation to finish the health ranking evaluation. The dependency matrix is used for extracting the optimal observation index and determining the corresponding index weight, so that the weight can reflect the sensitive degree of the index to the circuit stage change, the whitening function is obtained through the simulation, and the gray clustering health ranking is implemented according to measured data; the evaluation result is accurate and meets the actual condition.

The invention discloses a power supply area dividing method based on improved gray clustering, including the following steps: S1, establishing an evaluation index system; S2, collecting and preprocessing secondary index data; S3, determining the number of gray classes and the center point thereof; S4, calculating the whitening weight function of each gray class; S5, calculating the integrated weight vector v of the secondary indexes; S6, calculating the clustering coefficient of each secondary index about each gray class; and S7, getting the division result of a power supply area. By improving the most important parts, namely, the index weight calculation method and the whitening weight function design, in gray clustering, the power supply area division result is more accurate. Through reasonable division of the power supply area, optimal construction guidance schemes can be presented for distribution networks after division, in order to improve the utilization rate of assets and promote the overall quality of the power grid.

The invention provides a transformer direct-current magnetic bias evaluation method, which comprises the following steps: firstly, obtaining an actual exciting current characteristic quantity of a transformer, and determining a relative degradation value of the actual exciting current characteristic quantity based on the actual exciting current characteristic quantity; and determining a whiteningweight of the actual exciting current characteristic quantity in a preset gray state based on the relative degradation value; secondly, determining a variable weight coefficient of the actual excitingcurrent characteristic quantity based on the relative degradation value, and determining a final weight vector of the actual exciting current characteristic quantity according to the variable weightcoefficient; and finally, determining the sum of clustering coefficients of the actual exciting current characteristic quantity in the preset gray state based on the whitening weight and the final weight vector, determining the maximum value of the sum of the clustering coefficients, and determining the evaluation result of the transformer based on the gray state where the maximum value of the sumof the clustering coefficients is located. By adopting the evaluation method, the reliability and accuracy of evaluation can be improved.

The invention discloses a calculation method for predicting key proteins by combining multiple data features. According to the method, the features such as aggregation features, co-expression features, functional similarity and positional consistency of key proteins are analyzed; and the edge clustering coefficient of a protein interaction network, the Pearson correlation coefficients of gene expression values, the semantic similarity indexes of gene ontology terms and protein subcellular localization statistical features are effectively integrated. The method of the invention is simple and easy to use; four kinds of data, such as protein interaction relationship data, gene expression spectrum data, gene ontology term information data and protein subcellular localization data information are inputted; and test results indicate that the method of the invention can significantly improve the prediction accuracy and efficiency of the key proteins in the protein interaction network comparedwith an existing method.

The invention discloses an anti-attack detection method based on a network node topological structure. The method comprises the following steps: S1, importing a network and selecting a node as an attack object; s2, calculating five network topology properties: clustering coefficient, betweenness centrality, approximate centrality, feature vector centrality and neighbor node average value; s4, constructing a feature vector space; s5, attacking the network by using an anti-attack method; s6, extracting five network topology properties from the attacked network and constructing a vector space; and S7, adopting a classifier model random forest in machine learning, and verifying the feature vectors extracted in the S4 and the S6 by adopting a reservation method to obtain classification precision. The invention further provides an anti-attack detection system based on the network node topological structure. According to the method, whether the nodes are attacked by a certain countermeasure attack method or not is detected through the topological properties of the multiple nodes in the network, the complexity of the detection algorithm is reduced, the method is universally suitable for various attack methods, and high detection precision is obtained.

The invention provides a virtual network mapping method and relates to the communication field. The virtual network mapping method comprises the following steps of introducing a clustering coefficient theory and improving the clustering coefficient theory, calculating a virtual network node clustering coefficient weight and a physical network node clustering coefficient weight; generating a breadth-first search tree according to the virtual network node clustering coefficient weight; successively mapping virtual nodes in the breadth-first search tree; and successively mapping the virtual network link to a physical link. The virtual network mapping method can reduce number of virtual network mapping physical links and furthermore improves request accepting rate and revenue-cost ratio of the virtual network.

The invention discloses an evaluation method based on a variable-weight clustering method of a gray whitening function. The evaluation method comprises the following steps of establishing n clustering units i which represent the number of evaluated objects, wherein n=1, 2, 3, 4, ..., n, and i belongs to I; m clustering indexes j which represents evaluation indexes, wherein j=1, 2, 3, 4, ..., m, and j belongs to J; and s gray evaluations k, wherein k=1, 2, 3, 4, ..., s, and k belongs to K; performing dimensionless standardization processing on original data which are evaluated by the clustering units i;, obtaining a sample matrix of target object according to the sample values of the clustering units i about the clustering indexes j, wherein the elements in the matrix are clustering whitening values of an evaluating object; constructing a plurality of gray whitening weighting functions fjk about the clustering indexes j, wherein each gray whitening weighting function represents an evaluation grade; and determining each clustering coefficient for evaluating the gray, and obtaining the gray of the evaluating object, thereby finishing evaluation.

The invention relates to the technical field of computer network security, and provides a link discovery technique based network intrusion prediction method. The method comprises the following steps: acquiring the network data of network base-points, and carrying out processing on the network data so as to generate target data; calculating correlation coefficients among the network base-points; calculating the weighted degrees of the network base-points; calculating the weighted clustering coefficients of the network base-points; calculating the weighted comprehensive feature values of the network base-points; and sequencing the weighted comprehensive feature values of the network base-points, and finding out a key network base-point. By using the method disclosed by the invention, the problem that because a network is large in data size and streamed, a key network base-point with a high intrusion risk can not be quickly and accurately found is solved; and the method disclosed by the invention is small in calculated amount, quick in response speed and high in accuracy, and can effectively improve the network security protection capability.

An OSS project developer prediction method based on Email networks comprises the following steps that 1) different types of Email networks are constructed; 2) different network node ordering algorithms are used to calculate the different nodes and obtain corresponding characteristic values, and a network topology property is used to obtain the characteristic vector centrality and clustering coefficient of each node; 3) the characteristic values obtained via different algorithms and ranks of topology property parameters are normalized and serve as sample features; 4) part of the nodes serves as a sample and input to a machine learning classifier, and a Bayesian algorithm is used for learning; and 5) residual node samples are predicted. The method is aimed at the characteristic that an OSS project includes a lot of participants but a few core developers, developers in different OSS projects can be predicted effectively, and compared with a network node ordering algorithm, the accuracy is improved substantially.

The invention relates to a method for extracting a hyponymy relation of field terms from wikipedia. The method comprises the following steps of (1) using a wikipedia page corresponding to the field name as the starting page, carrying out the breadth-first traversal with the depth of 3, utilizing an URL (uniform resource locator) regular expression to filter the hyperlink not directing to the field term, and respectively storing the traversed page and hyperlink as the page text collection and the binary group collection; (2) obtaining the bidirectional link feature, edge betweenness feature and clustering coefficient feature from the binary group collection; obtaining the anchor text location feature and anchor text context feature from the text collection, and building five-dimensional feature vectors; (3) using a Random Forest classifier to carry out binary classifying on the hyperlink in the binary group collection according to the hyponymy relation and the non-hyponymy relation. The method has the advantage that the text feature and the hyperlink topology feature are comprehensively applied, so the hyponymy relation can be automatically extracted from the wiki.

The invention relates to a summary method of a complex network diagram. The complex network diagram is recorded as G=(V, E), V is a set of nodes, E is a set of edges. The method comprises the steps of1, selecting a subgraph H=(V ', E') from the complex network graph, wherein the formula of the sub-graph is shown in the specification, and the formula of the sub-graph is shown in the specification;for the edge formed by any two nodes in the sub-graph, the following conditions are met: sup(e(u, v),H) is greater than or equal to k-2, marking the subgraphs as dense subgraphs; Step 2, compressingall nodes in each dense subgraph into a node, marking the node as a super point, and marking edges associated with all the super points as super edges; and step 3, forming a new node set by all the super points and other nodes except the super points, forming a new edge set by the super edges and other edges except the super edges, and forming a summary graph by the new node set and the new edge set. The data scale can be effectively reduced, the phenomenon of dense nodes and associated edges in the network layout is reduced, and the power and clustering coefficients of the original graph areapproximately kept; and the overall structure of the data can be clearly displayed.

The invention discloses a privacy protection method and a privacy protection system oriented to triangular data release in a graph. The method comprises the following steps: deleting edges of originalgraph data to obtain a new graph with a threshold value lambda of the number of triangles connected by a single node; The sensitivity upper bounds of the histogram of the number of triangles and thenumber of corresponding nodes are calculated to determine the amount of noise added and the histogram of the distribution of the number of triangles after noise addition is issued. The sensitivity upper bounds of the histogram of the number of triangles and the number of corresponding nodes are calculated. The cumulative histogram sensitivity upper bound of the number of triangles and the number of corresponding nodes is calculated, and the cumulative histogram of the noised triangles is published. The local clustering coefficients are divided into k groups, and the sensitivity upper bounds ofthe clustering coefficients and the distribution histograms corresponding to the number of nodes in each group are calculated, and the distribution histograms of the clustering coefficients after noising are published. The sensitivity upper bound of cumulative histogram of clustering coefficients after grouping is calculated, and the cumulative histogram of clustering coefficients after noising is published. The invention publishes the triangular calculation result of the large graph data on the premise of ensuring privacy, and has certain usability and security.