3211 results about "Cluster result" patented technology

The invention discloses a clustering method based on mobile object spatiotemporal information trajectory subsections. The clustering method based on mobile object spatiotemporal information trajectory subsections comprises the steps that the three attributes of time, speed and direction are introduced, and a similarity calculation formula of the time, speed and direction is provided for analyzing an internal structure and an external structure of a mobile object trajectory; firstly, according to the space density of the trajectory, the trajectory is divided into a plurality of trajectory subsections, then the similarities of the trajectory subsections are judged by calculating differences of the trajectory subjections on the space, time, speed and direction, finally, trajectory subsections in a non-significant cluster are deleted or are merged into adjacent significant clusters on the basis of a first cluster result, and therefore an overall moving rule is displayed on the clustering spatial form. According to the clustering method based on the mobile object spatiotemporal information trajectory subsections, the clustering result is improved, higher application value is provided, a space quadtree is adopted to conduct indexing on the trajectory subsections, clustering efficiency is greatly improved under the environment of a large-scale trajectory number set, and trajectories can be effectively clustered.

The invention relates to a news event testing method based on metadata analysis, belonging to the technical field of data mining. The invention is characterized in that: a multi-dimensional vector model is used to represent news documents, and the time characteristics of news are given adequate consideration during the weight calculation of the characteristic representation, and an improved IDF (inverse document frequency) calculation method for news characteristic words. During calculating the similarity among different pieces of news, such information as the time, categories and specific contents of the news are taken into comprehensive consideration, the news documents are preprocessed by extracting the keywords, and thus the dimensionality of the vector is greatly reduced. Subsequently, the news reports are clustered by means of the hierarchical clustering method, and the clustering result tree is then partitioned to cluster the new reports and accordingly matched with relevant news events. Compared with the prior event testing method, the invention has the advantage of a higher F value, which is used as a standard to assess the quality of clustering.

The invention discloses a semi-supervised classification method of unbalance data, which is mainly used for solving the problem of low classification precision of a minority of data which have fewer marked samples and high degree of unbalance in the prior art. The method is implemented by the following steps: (1) initializing a marked sample set and an unmarked sample set; (2) initializing a cluster center; (3) implementing fuzzy clustering; (4) updating the marked sample set and unmarked sample set according to the result of the clustering; (5) performing the self-training based on a support vector machine (SVM) classifier; (6) updating the marked sample set and unmarked sample set according to the result of the self-training; (7) performing the classification of support vector machines Biased-SVM based on penalty parameters; and (8) estimating a classification result and outputting the result. For unbalance data which have fewer marked samples, the method improves the classification precision of a minority of data. And the method can be used for classifying and identifying unbalance data having few training samples.

The invention discloses a road network-based spatio-temporal trajectory clustering method. The method comprises the steps of data acquisition, spatio-temporal trajectory expression, spatio-temporal similarity measurement, sub-trajectory clustering and clustering result output. Through a trajectory recording device, spatio-temporal trajectory data of a moving object is acquired; a spatio-temporal trajectory model is built based on trajectory expression of a line segment; a trajectory file is output after linear interpolation and semantic expansion and is subjected to feature point selection; sub-trajectories are divided through feature points to perform trajectory reconstruction; a network distance between the sub-trajectories is calculated as a spatio-temporal similarity measurement basis; sub-trajectory clustering is realized by applying a label propagation algorithm; and finally a clustering result is output. According to the method, similarity and abnormal features in the spatio-temporal trajectory data can be extracted by performing clustering analysis on various spatio-temporal trajectory data, and meaningful trajectory modes in the data can be conveniently discovered.

The invention discloses a distributed knowledge data mining device and method used for a complex network. The distributed knowledge data mining device adopts a distributed computing platform which is composed of a control unit, a computing unit and a man-machine interaction unit, wherein the innovation key is to finish the calculated amount needed by a multifarious clustering algorithm in the data mining by different servers so as to improve the efficiency of the data mining. Aiming at different knowledge data, the degrees of relation and the weights of knowledge data also can be computed by applying different standards, so that a more credible result is obtained. A second-level clustering mode is adopted in the knowledge data mining process; the result of the first-level clustering is relatively rough, but the computing complexity is very low; and the computing complexity of the second-level clustering is relatively high, but the result is more precise. By combining the first-level clustering with the second-level clustering efficiently, the distributed knowledge data mining device improves the time complexity and clustering precision greatly in comparison with the traditional first-level clustering mode. According to the invention, as a visual and direct exhibition network structure and a dynamic evolutionary process are adopted, references are provided for the prediction in the fields of disciplinary development and hotspot research.

The invention discloses an image appearance based loop closure detecting method in monocular vision SLAM (simultaneous localization and mapping). The image appearance based loop closure detecting method includes acquiring images of the current scene by a monocular camera carried by a mobile robot during advancing, and extracting characteristics of bag of visual words of the images of the current scene; preprocessing the images by details of measuring similarities of the images according to inner products of image weight vectors and rejecting the current image highly similar to a previous history image; updating posterior probability in a loop closure hypothetical state by a Bayesian filter process to carry out loop closure detection so as to judge whether the current image is subjected to loop closure or not; and verifying loop closure detection results obtained in the previous step by an image reverse retrieval process. Further, in a process of establishing a visual dictionary, the quantity of clustering categories is regulated dynamically according to TSC (tightness and separation criterion) values which serve as an evaluation criterion for clustering results. Compared with the prior art, the loop closure detecting method has the advantages of high instantaneity and detection precision.

A Markov Random Field (MRF)-based technique is described for performing clustering of images characterized by poor or limited data. The proposed method is a statistical classification model that labels the image pixels based on the description of their statistical and contextual information. Apart from evaluating the pixel statistics that originate from the definition of the K-means clustering scheme, the model expands the analysis by the description of the spatial dependence between pixels and their labels (context), hence leading to the reduction of the inhomogeneity of the segmentation output with respect to the result of pure K-means clustering.

The embodiment of the invention discloses a malicious traffic detection implementation method and device based on deep learning. The method can include: obtaining the flow session of malicious code through dynamic sandbox technology; mapping the traffic session of malicious code to a gene map and extracting the map features; using the map features of traffic session for clustering and marking theclustering results by a malicious code family; training the preset depth learning model by using the tagged gene map of the malicious code family to establish the malicious traffic detection model; using the malicious traffic detection model to detect real-time network traffic, and realizing malicious traffic detection. Through the embodiment, many problems existing in the conventional detection technology such as artificial feature extraction difficulty, privacy disclosure, encryption and confusion difficult to identify, machine learning feature manual extraction and the like are solved to acertain extent, and the method has the characteristics of high robustness, high speed, high accuracy, low false alarm rate, cross-platform detection and the like.

In learning for pattern recognition, an aggregation of different types of object image data is inputted, and local features having given geometric structures are detected from each object image data inputted. The detected local features are put through clustering, plural representative local features are selected based on results of the clustering, and a learning data set containing the selected representative local features as supervisor data is used to recognize or detect an object that corresponds to the object image data. The learning thus makes it possible to appropriately extract, from an aggregation of images, local features useful for detection and recognition of subjects of different categories.

The invention provides a method and system for clustering movement tracks of vehicle objects in road network space. The method includes acquiring positioning data of different vehicles on a road in real time, wherein the positioning data includes longitude data, latitude data and course angle data; converting the longitude data, the latitude data and the course angle data into projection coordinates including x-axis coordinates, y-axis coordinates and course angles; performing network gridding on the projection coordinates, obtaining a plurality of partitions and marking the partitions; dispersing data in each partition into a plurality of sub partitions according to the course angle data, obtaining the maximal radius of neighborhood through calculation and obtaining an E domain through calculation based on the maximal radius of neighborhood; performing clustering in the E domain through a DBSCAN algorithm and obtaining a clustering result. According to the invention, based on mass vehicle positioning data information, through improvement of the DBSCAN algorithm, course angles are added into the algorithm, so that the clustering effect can be improved substantially.

The invention discloses a clustering method for question sentences in a question-and-answer platform and a system thereof. The technical scheme is as follows: the question sentences in the question-and-answer platform is analyzed according to the semantic feature of the question sentences to obtain analysis results; the semantic feature comprises the question type and the comparison feature of the question sentences and thesaurus correlative to the content of the question sentences; and aiming at the question sentences which is analyzed by the semantic feature, a clustering algorithm for evaluating the semantic similarity of the question sentences is adopted to obtain clustering results of the question sentences in the question-and-answer platform. The system comprises a question sentences analysis module and a clustering algorithm module. Aiming at the problem that the clustering method for the question sentences in the question-and-answer platform and the system thereof are not existed in the prior art, the technical scheme of the invention fills the gap, thereby not only realizing fast and exact clustering method and system in the question-and-answer platform, but also improving user experience.

The invention belongs to the field of network security visual analysis and relates to a website abnormal access behavior detection method based on the visual analysis. The method includes the steps of carrying out preprocessing on log data of a web server, utilizing a visual method to display position, time and content information of the data, utilizing an animation effect to display access events, carrying out clustering analysis on access users, carrying out acquisition and calculation on data attributes, and carrying out pattern discovery on abnormal access behaviors through combination of observation to visual results and clustering results and manual analysis. Compared with a traditional pure-machine computation method, the website abnormal access behavior detection method based on the visual analysis is capable of enabling a user to understand more clearly and visually, fully utilizes human intelligence, finds good balance between intelligent degree and human involvement, and is favorable for improving efficiency of solving problems.

The invention relates to a target reconstruction method based on a geometric constraint and belongs to the computer vision field. The method comprises the following steps of through a structure from motion (SFM) method, acquiring initial point cloud; through image characteristic point clustering, acquiring a classification result of characteristic points, wherein the classification result means aneighborhood relation of similar portions in an image; carrying out normal characteristic clustering of the initial point cloud, and using a corresponding relation between the classification result ofthe image characteristic points and an initial point cloud clustering result to define a geometric structure of the initial point cloud; using the geometric structure to acquire a sparse portion in the initial point cloud, defining the portion as a ''hole'', and then using a combined structure constraint of a ''hole'' area to carry out fitting of a space plane and a curved surface through an RANSAC method and a least square method; and sampling a fitted surface, adding an acquired three-dimensional point into the initial point cloud so as to acquire a dense point cloud model, and finally using a Poisson surface to reconstruct and acquire a three-dimensional model of a target. Through an experiment result, implementation of the method is verified and a good effect is achieved.

The invention discloses an alarm correlation analysis method in a communication network, targeting topology research in a tree-hierarchy structure network. The method comprises the following steps: in accordance with the space-time relativity of malfunctions occurring to network nodes, defining space-time relativity of upper layer network nodes in the tree-hierarchy structure network; based on the space-time relativity of the upper layer network nodes, clustering upper layer nodes in the tree-hierarch network; based on clustering result, dividing the total alarm database into a plurality of sub alarm databases; based on the attributes of alarm items, such as alarm occurrence frequency, alarm priority and alarm failure type, determining weight of each alarm item; utilizing the weighted Apriori correlation rules algorithm, conducting correlation rules mining on each sub alarm database. The method of the invention aims at addressing the problems of alarm relativity analysis of tree-hierarchy structure, and can effectively mine alarm correlation rules of interest among a large amount of alarm information.

The invention relates to a short-term load prediction method based on a clustering and sliding window. The method comprises the following steps of: preprocessing electric power load data; clustering historical data of a prediction user by utilizing a clustering algorithm, and adjusting clustering parameters; selecting k data from near to far of the prediction time in a category, containing most data, in clustering results to form a sliding window k; predicting the k selected data by utilizing a combination model based on the sliding window, and acquiring a primary prediction result; and correcting the primary prediction result of the combination model according to meteorological factors to obtain a final load prediction result. Compared with the prior art, the method has the advantages of high prediction precision, good adaptability and the like.

The invention relates to a network application encrypted traffic recognition method and device based on protocol attributes, and belongs to the technical field of computer network service security. The device comprises an offline training module and an online identification module. The offline training module is composed of a data set obtaining module, a message type fingerprint establishment module based on a second order Markov chain and a certificate length clustering module. A training set is obtained through a data set obtaining module. Application fingerprints are obtained and stored according to the training set by the message type fingerprint establishment module based on the second order Markov chain; clustering results and application certificate cluster distribution probability are obtained and stored according to the training set by the certificate length clustering module. The offline training module is composed of a network traffic capturing module and a recognition module. The recognition module matches the network traffic obtained by the capturing module with a stored application fingerprint library one by one; moreover, the certificate clustering results are taken into consideration, thus obtaining a recognition probability; the recognition result is an application corresponding to the highest probability. Compared with the prior art, the method and the device have the advantage of improving the recognition accuracy and efficiency.

The invention discloses power utilization mode classification and control method based on user behavior characteristics. With an improved secondary clustering model built by use of a secondary clustering method, the load point of each day of the same user in one year in an industrial park is taken as a characteristic vector, the daily power utilization characteristics of the user can be concluded from a clustering result, and a plurality of typical power utilization modes of the enterprise user can be provided, and therefore, basis can be provided for load prediction, fault diagnosis, electricity pricing and the like in the industrial park; furthermore, the optimal plane power utilization mode in demand side management can be selected by virtue of optimization function modeling on load data; the model is advantageous for a power supply company to reduce the loss of electricity selling profit as much as possible under the premise of guaranteeing power supply; at last, a user power utilization behavior mode library in the industrial park built on the basis is capable of comparing a new settling enterprise inconvenient to model with the modelled typical user mode and obtaining the load characteristics of the new settling enterprise by virtue of analogizing, and therefore, the planning efficiency of the park can be improved.

The invention discloses a bus load prediction method. The method comprises the following steps: correcting abnormal values in historical load data by use of a transverse comparison method, and determining key influence factors of bus load by use of a grey relation projection method; putting load curves with similar features in the same category by use of an improved K-Means clustering method to get a plurality of typical load patterns, building a random forest classification model, and establishing the mapping relationship between influence factors and clustering results; for each load pattern, training a plurality of prediction models by use of a multivariate linear regression method; and determining the category of a day under test, and selecting a matching regression model to realize load prediction. A data mining method is introduced to analyze the changing rules of bus load, and a prediction model library is built. Model matching is realized based on the category of a day under test. The accuracy and real-time performance of short-term bus load prediction are improved. More accurate decision support is provided for power grid planning and real-time dispatching.

The invention provides a cyclic neural network short-period load predication method based on information entropy clustering and an attenuation mechanism. The method comprises the following steps of analyzing characteristics which affact the power load; calculating the information entropies of all characteristics to the load by means of an xgboost algorithm; performing cluster analysis based on theinformation entropy of each characteristic as the weight on the historical data of a predicated area by means of a clustering algorithm; selecting a cluster which is nearest to a predicating day weight from the clustering results, and forming a time sequence T according to a sequence that the time to the predicating time reduces from longest to shortest; using the time sequence T as an encoder ofthe attention cyclic neural network, and obtaining a predication result by a decoder. Compared with the prior art, the cyclic neural network short-period load predication method has advantages of high predication precision and high self-adaptability.

The invention discloses a traffic flow state judgment method based on multiple-cross-section vision sensing clustering analysis. Traffic flow data are acquired by a PTZ video camera arranged on the road side and are adopted to judge the expressway road traffic flow state according to a clustering analysis method. By means of the traffic flow state judgment method, traffic flow data easy to acquire like speeds and flow are combined with upstream traffic flow data and downstream traffic flow data to achieve clustering analysis, acquired clustering results are clear, and certain fault tolerance exists. In practical application, the clustering number can be modified according to specific conditions, and clustering results are simplified. By means of the method, traffic condition division methods and critical data suitable for characteristics of current expressways are provided, and traffic flow conditions are accurately and comprehensively reflected.

The invention discloses a 3D point cloud data processing method. A ground filtering method for a 3D point cloud is achieved by establishing a 3D raster map and fitting a ground plane curve, a data structure is simple, the obtained ground plane curve is accurate and reliable, and filtering effect and real-timeliness are very good. A provided partition method adopts a method for clustering search windows in the cylindrical coordinate raster map, the calculation amount in the clustering process is greatly decreased, real-timeliness is good, and a clustering result is accurate. A provided training sample marking method is formed by combining point cloud partition with an appropriate display and storage method and is easy to achieve, multiple categories of samples can be marked in point cloud data of each frame, and sample marking efficiency is greatly improved.

The invention discloses an information searching method based on a public security domain knowledge ontology model, belonging to the searching technical field of natural language controlled words in the public security domain. The method disclosed by the invention comprises the steps of 1) establishing an analysis data warehouse, and implementing cluster analysis for the analysis data warehouse to obtain six basic elements; 2) dividing data in the analysis data warehouse into six categories according to a cluster result; 3) clustering each category of the data to obtain an element dimension of each category of the basic elements; 4) clustering the data in each element dimension to obtain a classification property of the data; 5) determining names of the controlled word categories according to a clustering result, dividing the public security data into the corresponding controlled word categories to obtain a controlled work bank; 6) establishing multi-dimensional quote marks for each controlled word; 7) searching the controlled word which is related to an input word in the controlled word bank according to an index number. The invention can automatically search vocabularies which are related to the target vocabularies, and can solve the problem that hidden information in the public security industry is hard to use and relate.

The invention belongs to the technical field of radar signal processing and discloses a multi-target clustering method for a high resolution millimeter wave radar. The method comprises the following steps of: obtaining signal-to-noise ratios of plots detected by the radar, setting a signal-to-noise ratio detection threshold, and discarding plots with signal-to-noise ratios below the signal-to-noise ratio detection threshold in the plots detected by the radar to obtain effective plots; sorting the effective plots according to the signal-to-noise ratios from high to low to obtain sorted effective plots; obtaining a relative distance and a relative angle of each effective plot and the radar and obtaining a spatial right coordinate position and a speed of each effective plot; clustering the sorted effective plots to obtain a plurality of clusters; and calculating the position, the size, and the speed of the center point of a target corresponding to each cluster. The multi-target clusteringmethod for the high resolution millimeter wave radar has the advantages of realizing a target point cloud cluster identification of the high-resolution radar, having no lag in clustering results, andcapable of accurately calculating the recognition target and the target information.

The disclosed apparatus uses a training data generation apparatus 2, which generates training data used for creating characteristic expression extraction rules. The training data generation apparatus 2 includes: a training data candidate clustering unit 21, which clusters a plurality of training data candidates assigned labels indicating annotation classes based on feature values containing respective context information, and a training data generation unit 22 which, by referring to each cluster obtained using the clustering results, obtains the distribution of the labels of the training data candidates within the cluster, identifies training data candidates that meet a preset condition based on the obtained distribution, and generates training data using the identified training data candidates.

The invention discloses a clustering method for network behavior habits based on K-means and LDA (Latent Dirichlet Allocation) two-way authentication. According to the clustering method, webpage properties, keywords and frequency in internet browsing records of persons are utilized to combine with a K-means algorithm, an LDA document topic extracting model and an annealing algorithm. The clustering method comprises the following steps: firstly, performing K-means algorithm clustering and LDA document topic extracting model generation on a staff-label-frequency set and a person browsing record-person-keyword set; secondly, storing and calculating an intermediate result, and then performing K-means and LDA two-way authentication by using the annealing algorithm; calculating a global best topic-classification label sequence, and optimizing a network behavior habit clustering result by taking the global best topic-classification label sequence as a reference. By means of the K-means and LDA two-way authentication, the sensitivity to person-classification labels is improved; by using the annealing algorithm, the optimizing efficiency of the clustering result can be improved, and further the clustering accuracy is improved.

The invention discloses a method for clustering gene expression data based on manifold learning, and the method provided by the invention comprises the following steps: acquiring a gene expression data matrix A through an acquisition system, and preprocessing the gene expression data matrix A by using a local linear smoothing algorithm; introducing the preprocessed data matrix A, and constructing a weighted neighborhood figure G in a three-dimensional space; taking the shortest path between two points as the approximate geodesic distance between two points; calculating a two-dimensional embedded coordinate by using an MDS (minimum discernible signal), and mapping the three-dimensional data matrix A to a two-dimensional visual space; and carrying out clustering on the two-dimensional visual space subjected to mapping by using a k-mean clustering algorithm so as to obtain the clustering result. The clustering method has the characteristics of low calculating cost, capability of eliminating high-order redundancies, suitability for pattern classification tasks, and the like; and by using the method disclosed by the invention, the current states of cells, the effectiveness of medicaments to malignant cells, and the like can be discriminated effectively according to the clustering result. The invention also provides a system for clustering gene expression data based on manifold learning.

The invention provides an information search method and system based on interactive document clustering. The method comprises the following steps that a document set is horizontally partitioned and preprocessed; word frequency statistics is conducted, and high-frequency words constitute a characteristic word set; vector space representation of documents is generated, the distances between the documents are calculated, and a similarity matrix is generated; a Laplacian matrix is generated, the number of clusters and a representation matrix are determined according to intervals between proper values of the Laplacian matrix, secondary clustering is conducted, and initial distance results are obtained; users conduct interactive operation on the initial distance results, new characteristic words are mined through chi-square statistics, a vector space is reconstructed, and the clustering process is repeated; finally, clustering results are shown to the users, and therefore the users obtain different categories of search results. According to the information search method and system, a semi-supervised learning approach in which the users intervene is adopted, the documents are clustered and analyzed, and the users obtain the different categories of search results.

The invention discloses a recommendation method and device of tour routes. The recommendation method and device are used for solving the problems that scenic spots recommended by existing tour recommendation services are scattered and a single tour route is recommended. The recommendation method comprises the steps that a server receives a tour route search request submitted by a client; according to search conditions carried by the search request, scenic spots meeting the search conditions are searched for; according to describing information of the searched scenic spots and the number of tour days, the scenic spots are clustered; clustering results of the scenic spots are recommended to the client. By the adoption of the technical scheme, search results of the scenic spots can be intelligently fed back according to the search conditions of users, the relevance among the scenic spots is improved by means of clustering, the scattered scenic spots are converted into the relevant scenic spots, dynamic search can be achieved, and the problem that the scenic spots are single is solved.

The invention discloses an adaptive spatial clustering method, comprising the following steps of: (1) preprocessing spatial data and selecting features; (2) creating a Delaunay triangulation network according to spatial attribute; (3) performing clustering analysis operations according to the spatial attribute; (4) turning to a step (5) if a spatial solid obstacle is needed to be further considered, and turning to a step (6) if a thematic attribute is needed to be considered, otherwise, ending the spatial clustering operations; (5) introducing a spatial obstacle layer, performing overlap analysis on the spatial obstacle and the side length of the Delaunay triangulation network between the entities in each spatial cluster, and breaking the side length if the spatial obstacle is intersected with the side length; (6) performing the thematic attribute clustering by an improved density-based spatial clustering method; (7) visualizing the clustering result, and outputting the clustering result. The adaptive spatial clustering method is simple and convenient to operate, high in degree of automation, high in calculation efficiency, perfect in functions, strong in applicability and the like, and can effectively improve capability of spatial clustering analysis to excavate deep-seated geoscience rules.