76results about How to "Efficient clustering" patented technology

The subject innovation provides for systems and methods to facilitate weighted suffix tree clustering. Conventional suffix tree cluster models can be augmented by incorporating quality measures to facilitate improved performance. Further the quality measure can be employed in determining cluster labels that show improvements in accuracy over conventional means. Additionally “stopnodes” can be defined to facilitate traversing suffix tree models efficiently. Quality measurements can be determined based in part on weighting factors applied to terms in a vector model, said terms being mapped from a suffix tree model.

The present invention relates to broadcasting / multicasting of media content over a communication network using shared bandwidth available from peer-to-peer networking. The system of the present invention includes a plurality of broadcast devices, a plurality of receiving devices, a plurality of databases, and a control center. The control center is the central nerve of the network, and provides a number of services, including but not limited to channel control, ad insertion, conditional access, program guide services, and the like. The broadcast device converts media content, which can be television, radio, and other data, received from various content providers, into digital data packets, having a suitable format for transmission over the Internet. Each receiving device will request the relevant packets, decode the received packets, and display or present the media content contained in the packets via an associated device. Packets may be received directly from the broadcast devices or from peers (other receiving devices) on the network.

The subject innovation provides for systems and methods to facilitate weighted suffix tree clustering. Conventional suffix tree cluster models can be augmented by incorporating quality measures to facilitate improved performance. Further the quality measure can be employed in determining cluster labels that show improvements in accuracy over conventional means. Additionally “stopnodes” can be defined to facilitate traversing suffix tree models efficiently. Quality measurements can be determined based in part on weighting factors applied to terms in a vector model, said terms being mapped from a suffix tree model.

The present invention discloses a tensor-based user track mining method. The tensor-based user track mining method comprises: (1) acquiring history track data of a user; (2) dividing data with time difference exceeding a predetermined time threshold in the history track data to form a plurality of segments of continuous track data; (3) extracting an arrest point of the user in each segment of track on account of each segment of continuous track data; (4) dividing the arrest point into a start point and a destination point and acquiring a corresponding road segment in sequence by a map matching method; (5) building a three-dimensional tensor by using the arrest point and the road segment sequence; (6) finding a related hotspot road segment between the start point and the destination point for a user search request (S,Q); and (7) calculating a recommended path according to a road segment weight set. The tensor-based user track mining method according to the present invention has the advantages as follows: only longitude and latitude of each of the start point and the destination point are provided for searching the hotspot recommended path between the start point and the destination point for the user search request; and the user does not need to understand the background implied data structure.

The invention belongs to the cross-media information technology field and particularly is an image optimization clustering method based on the typical correlation analysis. The invention mainly adopts the typical correlation to analyze while considering content characteristics of media data in various modes, maps the characteristics of the media data in various modes to an isomorphism sub-space of a united dimension through the sub-space mapping algorithm and obtains the final clustering result through optimizing clustering algorithm. The invention overcomes single-mode characteristic limitation in the multimedia field where only data is used, effectively solves the isomerism problem of the media data in various modes on the bottom layer characteristics, realizes the united measurement of the media object information between various modes, obtains results which are more accurate, more effective and more comforted to the needs in the large scale image data and has a wide application value in the cross-media processing and the retrieval field.

The invention relates to a network-based system for analyzing opinion information in a discrete text, belonging to the field of network information safety. The system comprises the following modules: a discrete text information acquisition module which acquires network information in a preset analysis cycle, a discrete text information tracking and restoring module which restores ellipsis and remote anaphora in the original content to obtain a text which contains a relatively complete text structure and semantic information, a semantic information mining and characteristic extracting module which realizes semantic information mining and characteristic extracting on text information by utilizing a latent semantic indexing technology, an opinion information clustering module which realizes information clustering by combining a niche genetic algorithm with a K-Means method, a hot opinion event discovery module which mines the hot opinion in the obtained topic and event, and a background information processing and data supporting center which analyzes data and provides a repertoire specially for a network, new words in the network, the existing class information and the existing hot topics. By applying the invention, the problem that information analysis is influenced as the text structure of the existing network opinion information is incomplete, ellipsis and remote anaphora are more and the new works in the network are more is solved, and the accuracy for discovery of the opinion and hot event is improved by adopting a high-efficiency clustering method.

The invention relates to electroencephalographic signal extraction, in particular to an electroencephalographic signal processing method based on isolated component automatic-clustering process. In order to improve signal-to-noise ratio for inducing an electroencephalographic signal, to remarkably improve the separability of the electroencephalographic signal within an action period under the stimulation of different tasks, and to facilitate the extraction and analysis of signal characteristics and the recognition of the task mode, the invention adopts a technical proposal which is in particular as below: the electroencephalographic signal processing method based on isolated component automatic-clustering process comprises the following steps: firstly, adopting an online maximum information algorithm (Informax algorithm) based on an information maximization criterion to sequentially carry out isolated component analysis (ICA) on stimulated and induced multi-channel signals, constructing a large component sample set Y with all obtained components, calculating the mutual information of the components, and finally adopting a total intra-class distance minimization criterion to carry out clustering process on a mutual information distance matrix so as to obtain class tags of all the components.

A method for indexing a suffix tree in a social network includes: scanning an input string and dividing the string into partitions each having a common prefix; performing no-merge suffix tree indexing on the divided partitions; storing information on the partitions on which no-merge suffix tree indexing is performed; storing suffix nodes of the no-merge suffix tree; and establishing a prefix tree. The performing no-merge suffix tree indexing includes: generating a set of suffixes having the common prefix in the input string; generating a suffix set from the set of suffixes and storing the suffix set; and building the suffix set as a sub-tree.

The invention discloses a distributed clustering method for Internet micro-content. The present invention adopts a multi-machine distributed clustering method. The main control machine divides the micro-content to be processed into multiple small files, and distributes these small files to multiple clustering machines for clustering operations. A single clustering machine performs meta-clustering on each assigned small file, and then merges these meta-clustering result files to obtain the corresponding stand-alone clustering merged file, and then sends it to the main control machine. After receiving the stand-alone clustering and merging files sent by each clustering machine, the master control machine extracts micro-content representative points from each stand-alone clustering and merging file, performs meta-clustering on these micro-content representative points again, and generates a new clustering category items, and merge the corresponding categories to get the final clustering result. The invention can accurately and rapidly cluster massive Internet micro-contents, and is an efficient and practical distributed clustering method.

The invention discloses a tracking measurement set partitioning method for multiple extended targets based on density analysis and spectrum clustering. The method is mainly used for solving the problems that in noisy environments, the number of the multiple extended targets is unknown, a changing measurement set is difficult to partition and the calculating cost is high. The method comprises the steps of constructing a density distribution function for the measurement set by adopting a Gaussian kernel, then, selecting a density threshold according to a density histogram technology, filtering noise wave measurements out of the measurement set, constructing a noise wave measurement data set removed similarity matrix by introducing an affinity propagation technology, finally, carrying out Laplace spectrum transform on the similarity matrix, and clustering by adopting a K-mean algorithm. The method has the advantages that the measurement set of the multiple extended targets can be accurately partitioned, and the calculating cost is reduced, so that the tracking performance for the multiple extended targets is improved, and the design requirements of actual engineering systems are met.

Provided is a color image clustering segmentation method based on the multi-scale perception characteristic of human vision. The method is characterized by comprising the following steps: firstly, segmenting a CIELAB color space into two parts through a cylinder with (a, b) as the circle center and Rn as the radius, wherein a=0 and b=0; secondly, segmenting an image into segments with a certain density and a certain size according to the traditional image segmentation clustering algorithm; thirdly, calculating the average color vector value of each clustering segment and projecting each vector onto the ab plane; fourthly, calculating the length of the vector, projected onto the ab plane, of the average color vector value of each clustering segment; fifthly, classifying the clustering segments into different measure spaces according to the lengths of the vectors; sixthly, calculating the included angle between the vectors of every two adjacent segment classes according to the formula shown in the specification; seventhly, clustering the segments meeting conditions with the formula as the criterion; eighthly, repeating the third step to the six step until convergence. By means of the method, the clustering effect and the anti-jamming capability of the image are improved.

The invention discloses a microblog recommendation method and device based on the data mining technology. The method includes the steps that microblogs are classified according to content; the click rate of selected users to each class of microblogs in a preset period of time is obtained, and according to the click rate of each class of microblogs, a microblog interest model of the users in the time period is obtained through calculation; the users are clustered according to the final microblog interest model, and microblog candidate sets recommended to all classes of the clustered users are determined according to the final microblog interest model; the microblogs in the microblog candidate sets are recommended to the users. In this way, the microblogs can be classified and sequenced in complex microblog information, the users are clustered, and corresponding microblogs which user groups with different features are interested in are recommended according to analysis results. The method and device can be used for accurately recommending hot microblogs which the users are interested in to the users by microblog websites in time.

The invention provides a method of identifying cell types based on single-cell RNA sequencing data. According to the method, a low-rank representation model of a high-dimensional matrix is effectivelycombined with a graph regularization theory; an optimization model is constructed by considering the global structure and local structure characteristics of data, the model is solved by adopting an alternating direction multiplier method (ADMM) to obtain a reliable inter-cell similarity matrix, and then the similarity matrix is clustered by adopting a spectral clustering method, so that single cells are clustered, and the cell types are identified. According to the method, the clustering effect of the single-cell RNA sequencing data can be remarkably improved.

The invention discloses a remote sensing image super-resolution reconstruction method based on fuzzy kernel classification and an attention mechanism. Firstly, high-resolution and low-resolution optical remote sensing images corresponding to a certain region are given, and a test sample and a training sample are divided; secondly, performing fuzzy kernel estimation on all low-resolution images inthe data; then, fuzzy kernels of all samples in the training set are used for K-means clustering; classifying the high-resolution image pair and the low-resolution image pair of the test set by usinga clustering model; and then constructing a neural network model based on an attention mechanism, setting absolute value errors of the high-resolution image and the low-resolution image as a loss function, obtaining an optimal model according to a test set reconstruction result, finally reconstructing an input image according to the model, and outputting a final result graph. According to the method, the peak signal-to-noise ratio of the reconstructed image can be improved, the robustness is high, and the definition of image edge details is improved.

The invention relates to a face detection method and a face detection system. The method includes: judging presence of an unmarked group of face images in internal memory; if presence is confirmed, determining emptiness of cache; if emptiness is affirmative, storing another group of face images and corresponding average characteristic values in the internal memory into the cache and using as first face images and first average characteristic values; if emptiness is negative, using corresponding average characteristic values of the unmarked group of face images in the internal memory as secondaverage characteristic values of second face images, and comparing the second average characteristic values with the first average characteristic values in the cache in terms of similarity; if similarity is affirmative, obtaining new average characteristic values as new first average characteristic values from the second average characteristic values and the first average characteristic values inthe cache; if similarity is negative, selecting three best face images in the cache and the corresponding first average characteristic values and outputting. A plurality of continuous groups of images of a same person can be effectively clustered and separated from those of other persons, so that a function of outputting one group of images for one same person is realized, and timeliness in face detection and acquisition is improved.

The invention provides a cell clustering method. The cell clustering method is characterized by comprising the step of calculating isolation between every two adjacent cells in a wireless communication network cell aggregation; the step of setting the isolation threshold value theta and judging whether a boundary exists between the two cells; the step of dividing the cell aggregation in which the boundaries exist between every two adjacent cells into sub-clusters Si in the wireless communication network cell aggregation U, wherein i=1, 2,...,m and the sub-clusters Si meet the formula: U=S1 UpsilonS2 Upsilon...Upsilon Sm; the step of calculating the connectivity among the sub-clusters; the step of setting a connectivity threshold value N and judging whether the sub-clusters need to be combined or not according to the connectivity threshold value N; when the connectivity degree between every two adjacent sub-clusters is larger than the connectivity threshold value N, the two sub-clusters are combined. The limitation tolerance for an interference link is ruled according to the connectivity threshold value among the sub-clusters, and on the premise that the interference among cell base stations is effectively avoided, the requirement for improving the self-adaptability of the cell network service can be improved.

The invention discloses an image division method based on an inter-class maximized PCM (Pulse Code Modulation) clustering algorithm. The method comprises the following steps: carrying out classified labeling on pixel points of an input image according to a gray value; obtaining a clustering label when a clustering analysis method is used for dividing a target image; and carrying out performance evaluation on a label obtained by image division and an original label according to an evaluation index by a clustering method. The novel inter-class maximized PCM clustering algorithm considers the inter-class penalty, and parameters are adjusted and adjusted to enlarge the distance between class centers, so that the optimal classification of the pixel points in the image is realized.

The invention relates to a network intrusion detection system based on machine learning. The system comprises a network intrusion data acquisition module, an intrusion data classification processing module, a network training initialization data selection module, a generalized neural network module, a network training data selection module and a result output module; results are obtained through intrusion data acquisition, classification, selection, training and prediction for result output and alarming. The system has the beneficial effects that the data is from network intrusion data, and the algorithm aims to effectively cluster the intrusion data. Compared with a traditional algorithm, the intrusion detection accuracy and effectiveness are improved, a computer network security protection system is perfected, the security coefficient of the network is greatly improved, and the labor cost is saved.

Provided are a data clustering apparatus and method, which can rapidly and accurately cluster data. The data clustering apparatus includes an index discriminating unit discriminating an index corresponding to an input position of new data input to a space for data clustering, including a lattice-type segmented space having lattice unit spaces set with different indexes, and a clustering unit creating a new cluster in the discriminated index using the input new data as a representative value when a cluster is not created at the discriminated index.

The invention relates to a hierarchical clustering-based log audition method and device. The method comprises the following steps of: dividing a log into a first part and a second part; respectively determining vectors of the first part and the second part; and clustering the log by utilizing the vectors of the first part and the second part of the log so as to obtain a clustering result of the log, wherein the first part comprises attributes expressed by uniform structures in the log and the second part comprises attributes expressed by non-uniform structures in the log. According to the method and device provided by the invention, log audition is carried out by utilizing a hierarchical clustering method so as to carry out clustering on logs, so that abnormal log information in logs of people is mined.

The invention discloses a graph-data-oriented projection clustering method. The method comprises the following steps: for a graph data set D to be processed, obtaining representative subgraph patternsof all subgraphs in the graph data set D through a depth-first search algorithm; for the representative subgraph patterns, obtaining Top-k diversity subgraph patterns of the graph data set D, and enabling all of the Top-k diversity subgraph patterns to be generated into a Top-k diversity subgraph pattern set RS; carrying out projection matching on each subgraph in the graph data set D and each characteristic sub-graph in the Top-k diversity subgraph pattern set RS and obtaining a characteristic matrix of the graph data set D; and with adaptive entropy serving as a clustering objective function, carrying out clustering processing on the characteristic matrix through a graph projection clustering algorithm to obtain a clustering result. The method enables the clustering result of the graphdata set to be more accurate and higher in diversity; and high-dimensional data processing effect is better.

The invention provides a pedestrian detection method. The pedestrian detection method comprises the steps that (1) a laser range finder is installed, (2) laser cloud point data of the laser range finder are set by a laser scanner module, (3) a laser data preprocessing module is established, (4) obstacles in front of a vehicle are detected in real time through a pedestrian detection module, and (5) pedestrian targets are distinguished in real time through a pedestrian distinguishing module. The invention further provides a detection system for realizing the pedestrian detection method. The detection system for realizing the pedestrian detection method comprises a power supply, a DSP, a data circuit, a power circuit and the laser range finder, wherein the laser range finder is connected with a network interface of the DSP through the data circuit, the power supply supplies power to the laser range finder through the power circuit, and the laser scanner module, the laser data preprocessing module, the pedestrian detection module and the pedestrian distinguishing module are arranged inside the DSP. The pedestrian detection method and the detection system of the pedestrian detection method have the advantages that the detection speed is high, and the reliability and robustness are high.

The invention provides a data clustering analysis method based on Grassmann manifold, and relates to a spatial data clustering method. The method comprises the following processes of inputting N data points {x}<n><i=1> and the clustering number K to calculate the distance among the data points; constructing a Laplacian matrix L=D<-1 / 2>SD<-1 / 2>, wherein the D is the diagonal matrix, and D<ii>=[sigma]<J=1><n>S<ij>; calculating the feature vectors v<1>, v<2> to v<k> corresponding to the k maximum feature values of the Laplacian matrix L, and constructing a matrix V=[ v<1>, v<2> to v<k>] which is an element of a set R<nk>; and regarding each row of Y as a point in an R<k> space, and performing classification by using a K means algorithm. The data clustering analysis method based on Grassmann manifold has the advantages that the data distributed on different sub spaces can be effectively clustered; data sets with complicated geometrical structures can be analyzed; and the effective clustering is performed on the manifold space.

The invention discloses a thalamus function partitioning method based on subspace feature learning. The thalamus function partitioning method comprises the following steps: firstly, carrying out fibertracking by using diffusion tensor imaging to obtain internal structure connection information of the brain of a living body, and extracting complex nonlinear thalamus cortex features by using fine cortex partitions to form structure connection features; then, using the deep subspace network and the hidden subspace mapping of the added self-expression feature learning features to extract low-dimensional subspace characteristics; and finally, performing spatial constraint on voxel features to reduce the influence of noise, better reflecting a spatial topological structure, enriching the extraction of spatial information, constructing an affinity matrix, and obtaining functional partitions by using a normalized segmentation method. According to the thalamus function partitioning method, theinfluence of noise can be reduced, and the topological structure of voxel space can be better reflected, and extraction of space information is enriched, and thalamus function partitions can be efficiently obtained.

The invention provides a wireless sensor network optimization method for environment monitoring. The wireless sensor network optimization method can solve the problems of establishment and reconstitution of clusters including sensor nodes, cluster heads and aggregation nodes in a wireless sensor network on the basis of a genetic algorithm. The wireless sensor network optimization method includes the steps of introducing concepts of active nodes, common nodes and dormant nodes, comprehensively considering the requirements for connectedness restriction of a network, energy conservation and the like, calculating a fitness value of each node through a fitness function, selecting the node with the highest fitting value as the cluster head node, and achieving automatic switching of the nodes in the active state, in the common state and in the dormant state. The wireless sensor network optimization method can broaden network monitoring scope, improve utilization rate of network node energy in the environment monitoring, and greatly prolongs the service life of the network.

The present invention relates to a heterogeneous data cluster generation apparatus and method and a data clustering method and apparatus, and more particularly, to a data clustering method and apparatus which cluster data measured by different sensors into a number of groups. Aspects of the present invention provide an apparatus and method for generating clusters by putting together heterogeneous data which are values measured by different types of sensors. Aspects of the present invention also provide an apparatus and method for generating clusters by setting indices in order to effectively cluster multi-dimensional data, massive data, or scattered data.

The present invention relates to broadcasting / multicasting of media content over a communication network using shared bandwidth available from peer-to-peer networking. The system of the present invention includes a plurality of broadcast devices, a plurality of receiving devices, a plurality of databases, and a control center. The control center is the central nerve of the network, and provides a number of services, including but not limited to channel control, ad insertion, conditional access, program guide services, and the like. The broadcast device converts media content, which can be television, radio, and other data, received from various content providers, into digital data packets, having a suitable format for transmission over the Internet. Each receiving device will request the relevant packets, decode the received packets, and display or present the media content contained in the packets via an associated device. Packets may be received directly from the broadcast devices or from peers (other receiving devices) on the network.

The embodiment of the invention discloses a distributed comparison clustering method and a device, electronic equipment and a storage medium. The method according to an embodiment of the invention comprises the following steps: receiving sub-tasks equally divided by clustering tasks in a task queue through each computing node; processing the received sub-tasks by the computing node; and sending acomparison clustering result generated by processing to a result queue, continuously getting new sub-tasks from the task queue for processing until all the sub-tasks equally divided corresponding to one clustering task are processed, and finally summarizing the comparison clustering result of clustering each sub-task by the management node to finish the comparison clustering task. And the computing nodes process one subtask according to the processing process of the subtask and then get a new subtask until the comparison clustering task is completed, so that the computing power of each computing node is balanced and fully utilized, and the data comparison clustering is more efficient.

The embodiment of the invention provides a user data processing method, device and equipment and a readable storage medium. The method provided by the embodiment of the invention comprises the steps of obtaining behavior characteristic data of a user; screening the behavior characteristic data based on the ARPU value of the user to obtain behavior characteristic data of the target user with the ARPU value in a preset range; preset clustering algorithm, performing clustering analysis processing on the behavior characteristic data of the target user; obtaining multiple user clusters, the marketing object group is determined based on the ARPI value of the user; according to the method, the user cluster is divided into the multiple user clusters, the multiple user clusters are obtained throughreal-time and efficient clustering according to the multi-dimensional behavior characteristic data of the user, the user clusters are divided more accurately and more efficiently, and therefore the marketing strategy used for improving the ARPU value of the user can be further provided for the different user clusters in a targeted mode.