54 results about "Agglomerative hierarchical clustering" patented technology

A method extracts planes from three-dimensional (3D) points by first partitioning the 3D points into disjoint regions. A graph of nodes and edges is then constructed, wherein the nodes represent the regions and the edges represent neighborhood relationships of the regions. Finally, agglomerative hierarchical clustering is applied to the graph to merge regions belonging to the same plane.

The invention provides a method for learning and anomaly detection of trace modes by utilizing much feature information of a trace. Firstly, in the trace mode learning phase, similarities of motion directions and spatial positions between traces are considered at the same time, a typical trace motion mode is extracted by hierarchical agglomerative clustering, and is provided with high cluster accuracy; and the time efficiency is greatly improved through constructing a Laplacian matrix and reducing the dimensionality of the matrix. Then in the abnormity detection phase, a distribution area of scene starting points is learned through a GMM model, a moving window is used as a basic comparing element, differences of a trace to be detected and a typical trace in position and direction are measured by defining a position distance and a direction distance, and an on-line classifier based on the direction distance and the position distance is established. That the trace belongs to a starting point abnormity, a global abnormity or a local abnormity is determined online through a multi-feature abnormity detection algorithm; and due to the fact that starting point, direction and position feature differences are considered at the same time, and the global abnormity and the local child segment abnormity are considered, the learning and anomaly detection method based on multi-feature motion modes of the vehicle traces is higher in abnormity recognition rate when being compared to traditional methods.

The invention relates to a two-stage hybrid particle swarm optimization clustering method, which is mainly used for solving the problems of greater time consumption and low accuracy of the conventional particle swarm optimization K-mean clustering method when the number of dimensions of samples is higher. The technical scheme disclosed by the invention comprises the following steps: (1) reading a data set and the number K of clusters; (2) taking statistics on information of dimensionality; (3) standardizing the dimensionality; (4) calculating a similarity matrix; (5) generating a candidate initial clustering center; (6) performing particle swarm K-mean partitional clustering; and (7) outputting a particle swarm optimal fitness value and a corresponding data set class cluster partition result. According to the two-stage hybrid particle swarm optimization clustering method disclosed by the invention, the first-stage clustering is firstly performed by adopting agglomerative hierarchical clustering, a simplified particle encoding way is provided, the second-stage clustering is performed on data by particle swarm optimization K-mean clustering, the advantages of hierarchical agglomeration, K-mean and particle swarm optimization methods are integrated, the clustering speed is accelerated, and the global convergence ability and the accuracy of the clustering result of the method are improved.

A process planning method based on similarity theory comprises the following steps of: i) completing a part characteristic information model by adding a characteristic combination description on the basis of traditional single characteristic description, and building an description of part characteristics; ii) indicating entities formed by relevant information of relevant characteristic processing procedures regarding the characteristics as core by process elements, and generating a most basic module and a procedure characteristic of process planning; iii) building a case indexing structure based on the characteristics, generating similar process information by the similarity judgment of a part characteristic multi-branched tree, and performing the process similarity judgment; iv) performing procedure combinations by an agglomerative hierarchical clustering method, so as to generate a procedure cluster; and v) sequencing all the process elements in the procedure cluster according to the precedence relation, so as to ensure that all the process elements in the procedure cluster can be processed on a machine tool.

A system and method of identifying outliers in item categories are described. A pairwise similarity measurement may be determined between each item listing in a plurality of item listings based on a comparison of at least one feature of each item listing. At least one outlier among the plurality of item listings may be determined using the pairwise similarity measurements. The feature(s) may comprise at least one feature from a group of features consisting of: a title, an image, a price, an attribute, and a description. Each item listing in the plurality of item listings may belong to the same leaf or non-leaf category in a network-based marketplace or publication system. The outlier(s) may be determined using at least one clustering algorithm. The clustering algorithm(s) may comprise an agglomerative hierarchical clustering algorithm and / or a density-based clustering algorithm.

The invention aims at the actual demands for parallel solving of electric power system problems and provides a distribution type parallel load flow calculation method based on hierarchical clustering automatic partition. Firstly, automatic partition is achieved through an agglomerative hierarchical clustering algorithm; secondly, partition parallel calculation is achieved through a decoupling method between areas; finally, global convergence is achieved based on iteration convergence conditions, and therefore the power distribution network parallel load flow solving method which is suitable for characters of a power distribution system, strong in parallelism and small in data dependency among tasks can be achieved. According to the distribution type parallel load flow calculation method, clustering advantages can be fully developed, work which needs to be done by ultra-large type computers in the prior art can be finished at low cost and high speed, the computing speed is obviously improved under the precondition that computational accuracy is guaranteed, and the demand for real-time performance of analysis and control is met.

The invention discloses a method for predicting crystallizer breakout databased on agglomerative hierarchical clustering, and belongs to the technical field of steel metallurgy continuous casting detection. The method comprises the following steps that firstly, a sticking breakout / normal work condition sample database is established, the sticking breakout temperature and the normal work conditiontemperature are selected, and a sample database comprising a sticking breakout sample set and a normal work condition sample set is constructed; secondly, random sample set hierarchical clustering isconducted, equivalent samples are selected from the sticking breakout sample set and the normal work condition sample set at random, a random sample set is formed by the samples and online actual measurement temperature samples, and hierarchical clustering is conducted on the random sample set; and thirdly, breakout recognition and alarming are conducted, whether the online actual measurement temperature samples belong to the sticking breakout class cluster or not is detected, and accordingly breakout can be recognized and predicted. By means of the method, limitation of manual parameter defining in the predicting process is avoided, whether the online actual measurement temperature samples comprise the breakout features or not is judged only through respective features of the sticking breakout and the normal work condition temperature, and the beneficial effects that the detection principle is clear, the operation speed is high, and the breakout recognition accuracy rate is high are achieved.

The invention discloses a test-case selection method based on user sessions and a hierarchical clustering algorithm. The method includes the following steps: acquiring server access logs, and carryingout sorting according to time; carrying out preprocessing and clustering to form a user session sequence set; calculating similarity distances among all user session sequences through using an improved user-session-sequence comparison algorithm; employing the improved condensing hierarchical clustering algorithm to cluster the user session sequences, and outputting final clustering results of test cases; and optimizing selection of the test cases through deleting redundant test cases. According to the method of the invention, representative user operation sequences can be quickly mined from the large number of server access logs to use the same as test cases, automation of test-case generation and optimization of test-case selection are realized, and subsequent work of automated functiontests of a server, performance tests, user behavior analysis and the like is facilitated.

The present invention discloses an object segmentation method based on multiple-instance learning and graph cuts optimization. The method comprises the first step of carrying out salient model construction by adopting a multiple-instance learning method on training images, and predicting packages and instances of a testing image by using a salient model, thus to obtain a saliency testing result of the testing image; a second step of introducing the saliency testing result of the testing image into a graph-cut frame, optimizing the graph-cut frame according to instance characteristic vectors and marks of the instance packages, acquiring a second-best solution of graph cuts optimization, and obtaining precise segmentation of an object. According to the method provided by the present invention, the saliency testing model is constructed by using the multiple-instance learning method and thus is suitable for images of specific types, the saliency testing result is used into an image segmentation method based on the graph theory so as to guide image segmentation, a graph cut model frame link is optimized, an agglomerative hierarchical clustering algorithm is adopted for solving, the segmentation result can thus well accords to semantic aware output, and an accurate object segmentation result can be obtained.

The invention discloses a transformer fault diagnosis method based on improved hierarchical clustering. The method comprises the steps of (1) establishing a transformer vibration signal acquisition system, carrying out data sampling on a transformer based on the established system to obtain a transformer vibration signal, (2) obtaining a transformer vibration signal through the Hilbert-Huang transform method, and (3) carrying out hierarchical classification on the vibration signal characteristic values of a tested transformer through an improved agglomerative hierarchical clustering algorithm, and determining the state of the tested transformer. According to the improved agglomerative hierarchical clustering algorithm, based on a traditional agglomerative hierarchical clustering algorithm, after objects are combined, the data of an initial distance matrix is directly used to carry out data processing, and a technical effect of the rapid and effective diagnosis of a transformer fault state is realized.

The invention provides a group abnormal behavior detection method based on an air monitoring platform. Firstly, light flow vectors of feature points are appropriately corrected by estimating depth information of an image to reduce a target movement speed estimation error caused by the perspective phenomenon, then the light flow vectors of the feature points are clustered, and target detection under a moving camera is achieved by combining a background movement consistency law. Abnormal behaviors are detected by adopting a double-Gauss mixed model, and model parameters are solved by using an expectation maximization algorithm. Finally, misjudgment is verified by adopting a time queue mechanism, space coordinates of the abnormal feature points are clustered by means of a simplified agglomerative hierarchical clustering algorithm, the isolated abnormal feature points are removed, and abnormal groups are marked. The validity of the method is verified by experiments in multiple scenes.

The invention discloses a travel endpoint identification method based on a multi-layer agglomeration hierarchical clustering algorithm. According to 4G communication signaling data of a user mobile phone collected by a communication operator, a three-layer algorithm model is proposed: an equal-time-distance interpolation algorithm, an agglomeration hierarchical clustering algorithm and a ping-pongdwell correction algorithm are used for extracting a user trip end point. Finally, the travel endpoint information of the user all day is further sorted out to form the travel time-space sequence completed by the individual. The invention utilizes the feature that the positioning frequency of 4G signaling data is higher than that of 2G signaling data, exerts the identification advantage of the multi-layer condensation hierarchical clustering algorithm, solves the shortcomings of the traditional single-layer algorithm in identifying actual travel endpoints, and realizes the intelligent identification of the endpoint information of residents' traffic travel by using the signaling data in the background of 3G / 4G-LTE technology. This method can be used for large-scale, automated information collection of residents' travel endpoints.

The invention relates to the field of image recognition, in particular to a method for identifying shielded fruits in a natural environment. The method comprises: preprocessing, solving an optimal shielding threshold value by utilizing a distance fluctuation curve of adjacent manual marking points on a region edge and a mass center, judging a fruit shielding region and an unshielded edge, and carrying out random three-point circle determination on non-adjacent marking points on the unshielded edge; and taking the obtained circle center point set as the input of an improved cohesion hierarchical clustering classifier based on heap sorting, solving the circle center mean value of each class in the result, and performing circle fitting on the circle center mean value and the farthest markingpoint on the edge of the fruit to realize the identification of the shielded fruit. Compared with the prior art, the recognition method is good in real-time performance, high in recognition rate and small in fluctuation range.

The invention provides a satellite telemetry data clustering method based on time series special point. The satellite telemetry data clustering method comprises steps that step1, a Zscore algorithm is used for preprocessing of an original time series set X=(x1, x2...Xl); step2, an SPSegmentation segmentation expression method is used to extract all of special points of every original time series Xi to form a special point series SPSi=(xi(tp1), xi(tp2).. .xi(tpk)), which is used to replace the original series to be used as clustering input; step3, the corresponding time position supplementing and aligning processing of any two special point series SPSi and SPSj is carried out, and then the time positions of the special points of the sequences corresponding to the various elements of the two SPS on an original time axis are aligned with each other; step 4, the similarity coefficient calculation of the aligned special point series is carried out, and because the sequence aligning is carried out, most of similarity measurement calculations are usable, and finally, a PSPS_Dist similarity matrix is acquired; step 5, an agglomerative hierarchical clustering algorithm is adopted to realize time series clustering.

The invention proposes a speaker marking method, and belongs to the technical field of voiceprint recognition, mode recognition and machine learning. The method includes three stages: in the first stage, voice data to be tested are divided into fragments with equal length through an i-vector probabilistic linear discriminant analysis-agglomerative hierarchical clustering method, and then the fragments are clustered into categories of which the total number is equal to that of speakers; in the second stage, prior probability that a fragment belongs to a speaker is obtained; and in the third stage, iteration is performed through a variational Bayes hidden Markov method based on soft decision, when a system converges, a speaker to which a fragment belongs is calculated, and speaker marking is finished. The speaker marking method provided by the invention combines advantages of two speaker marking methods, and can effectively improve the accuracy rate of speaker marking.

The invention discloses a power utilization load prediction method based on adaptive hierarchical time sequence clustering. The method comprises the steps: 1), sequentially dividing load sequences on a window time segment according to the rise, reduction and leveling-off features of a power utilization load after quantification; 2), carrying out the agglomerative hierarchical clustering of the load sequence features through employing a hierarchical clustering method; 3), carrying out the prediction of the load through the hierarchical idea, enabling the latter load sequence in a prediction load sequence group to serve as a prediction load of a next moment, and enabling the predicted load sequence to serve as the current latest load sequence; 4), dynamically adjusting a quantification factor, a time window and a clustering parameters through a feedback method, and completing the power utilization load prediction. Therefore, the method is more precise in a supershort period, effectively reduces the storage price of original data in load prediction, and plays a support role in scientific and accurate power dispatching of an intelligent power grid.

The invention relates to a method for optimum design of gear reducers on the basis of a clustering multi-objective estimation of distribution algorithm, and aims at solving the problems that the local search ability of the existing multi-objective estimation of distribution algorithm is not sufficiently utilized in the multi-objective optimization problem solving process, abnormal solutions are directly discarded in the solving process, the population diversity is easy to lose and too much calculation overhead is used for constructing an optimum probability model. The method comprises the following steps of: firstly dividing a population into a plurality of local classes by utilizing an agglomerative hierarchical clustering algorithm; randomly selecting a unity from each local class to form a global class; and constructing a Gaussian model for each unity to approach a population structure and carrying out sampling to generate a new unity, wherein the mean value of the Gaussian model is the unity, and a covariance matrix is a covariance matrix of the local class where the unity is located or a covariance matrix of the global class. The method disclosed by the invention is used for the field of spaceflight.

The invention discloses a method for clustering and style migration of ultrasound images, and the method comprises the following steps: 1, carrying out the data preprocessing, and converting the ultrasonic image into a data vector; 2, performing a contrast experiment by using three clustering algorithms: a K-means clustering algorithm, a condensation hierarchical clustering algorithm and a spectral clustering algorithm, and selecting an optimal algorithm; 3, selecting an image with large style difference according to a clustering experiment result; 4, designing a network structure and a loss function based on the CycleGAN; and 5, designing a style migration experiment for the benign and malignant data set, and analyzing an experiment result. According to the clustering and style migrationalgorithm designed by the invention, the influence of noise on an experimental result is effectively avoided, and the adaptability of the used method to noise image data is presented.

The invention relates to a standby relay node deployment method and system in an intelligent distribution communication network. The method comprises: obtaining the relay node information in an intelligent distribution communication network; utilizing an agglomerative hierarchical clustering algorithm to divide the intelligent distribution communication network into a plurality of communication groups; and deploying relay nodes for each communication group if all the communication groups meet preset constraint conditions. The method and system employ an agglomerative hierarchical clustering algorithm to divide relay nodes in the intelligent distribution communication network into a plurality of communication groups, and deploy relay nodes for each communication group, allowing the intelligent distribution communication network to isolate a failed relay node from a network operation scope after failure of the relay node so as to avoid network chain failure response caused by failure of few relay nodes, and meanwhile allowing standby relay nodes to replace original relay nodes when some relay nodes fail so as to eliminate the unfavorable influence caused by relay node failure.

The invention relates to the technical field of artificial intelligence, and discloses a classification model construction method and device, computer equipment and a readable storage medium. The method comprises the steps: recognizing the degree of difference between stock rules through a condensation hierarchical clustering algorithm, and summarizing the stock rules with the degree of differencelower than a difference threshold to generate a rule set; extracting a judgment index and a judgment result of each stock rule in the rule set, and performing duplicate removal on the judgment indexand the judgment result to obtain a rule index and a rule result; constructing a classification model taking a rule result in the rule set as a classification result and taking a rule index as a judgment factor; and creating a configuration page according to the rule indexes in the rule set, and associating the configuration page with the classification model corresponding to the rule set. According to the method, only one classification model needs to be called to calculate the calculation data corresponding to the plurality of similar stock rules, so that the complexity of rule use is greatly reduced, and the problem that a current user difficultly finds an accurate rule to develop a service needing to be solved at present is solved.

The invention discloses an aviation fault text abstract generation method and device based on a fault causal atlas. The method comprises the following steps: step 1, generating a directory atlas through an original fault causal atlas; 2, obtaining a feature vector of a target text sample; 3, generating text candidate abstracts by using the fault causal atlas model; 4, decoding the text generation model; and 5, generating an aviation fault text abstract. According to the method, firstly, a catalog atlas is obtained for a fault causal atlas which is constructed in advance through an agglomerated hierarchical clustering method, then structural features obtained in advance are introduced in the encoding stage and the decoding stage through the fault causal atlas, and finally the needed aviation fault text abstract is generated. According to the method, information extraction is carried out by utilizing a document chapter structure, fault logic positioning is carried out by applying a fault cause and effect graph, and the problem that an existing generative abstract model cannot meet chapter directory structure constraints is solved.

The invention provides an image tampering detection method, electronic equipment and a storage medium. The method comprises the steps of extracting feature points and feature vectors from a to-be-detected image; matching the feature points to obtain a matching point pair set; using a hierarchical clustering algorithm to perform condensation hierarchical clustering on the matching points in the matching point pair set to obtain a clustering tree diagram; obtaining different class clusters by utilizing different clustering numbers; utilizing an elbow rule to obtain an optimal class cluster set;selecting two class clusters from the optimal class cluster set in sequence, and executing the following steps on the two class clusters: judging whether the number of matching point pairs between thetwo class clusters is greater than a preset threshold value or not; if so, selecting a matching point pair from the two class clusters; calculating an affine transformation matrix according to the matching point pairs; searching matching point pairs by utilizing the affine transformation matrix; and determining a copying and pasting area according to the matching point pair. According to the method and the equipment provided by the invention, the algorithm complexity is reduced, the algorithm operation efficiency is improved, and the copying and pasting area is positioned more intelligently and accurately.

The invention discloses a network connectivity correction method for intelligent optimization of a recovery path of a power failure system. The method comprises the following steps of A) establishing an initial to-be-connected graph by each recovered line, a charged node and a target node; B) performing repeated combination until connected sub-graph aggregation is finished to form connected sub-graphs; C) searching for a shortest connection path that connects all the connected sub-graphs by utilizing a prim algorithm, and establishing a connected graph containing all to-be-recovered power supply points; and D) setting a line state on the connection path on the connected graph to be 1 according to the connected graph containing all the to-be-recovered power supply points, thereby realizing connectivity correction. The method for effectively correcting non connected individuals into connected individuals according to a connectivity correction algorithm based on an agglomerative hierarchical clustering algorithm and the prim algorithm is high in calculation speed, small in number of iterations, high in convergence speed, relatively high in stability and relatively small in calculation result fluctuation, and has very high adaptability and good application prospects.

A process planning method based on similarity theory comprises the following steps of: i) completing a part characteristic information model by adding a characteristic combination description on the basis of traditional single characteristic description, and building an description of part characteristics; ii) indicating entities formed by relevant information of relevant characteristic processing procedures regarding the characteristics as core by process elements, and generating a most basic module and a procedure characteristic of process planning; iii) building a case indexing structure based on the characteristics, generating similar process information by the similarity judgment of a part characteristic multi-branched tree, and performing the process similarity judgment; iv) performing procedure combinations by an agglomerative hierarchical clustering method, so as to generate a procedure cluster; and v) sequencing all the process elements in the procedure cluster according to the precedence relation, so as to ensure that all the process elements in the procedure cluster can be processed on a machine tool.

The invention discloses a target segmentation method based on multi-instance learning and graph-cut optimization: Step 1: use the multi-instance learning method to carry out saliency model modeling on the training image, and use the saliency model to carry out saliency model modeling on the packages and examples in the test image Predict, get the saliency detection result of the test image; Step 2: Introduce the saliency detection result of the test image into the graph cut framework, optimize the graph cut framework according to the example feature vector and the label of the example package, and solve the suboptimal graph cut optimization solution to obtain an accurate segmentation of the target. The present invention adopts the method of multi-instance learning to establish a saliency detection model to make it suitable for specific types of images, and uses the results of saliency detection in an image segmentation method based on graph theory to guide image segmentation, and the frame link of the graph cut model It is optimized and solved by agglomerative hierarchical clustering algorithm, so that the segmentation results can better conform to the output of semantic awareness, and obtain accurate target segmentation results.

The invention relates to a two-stage hybrid particle swarm optimization clustering method, which is mainly used for solving the problems of greater time consumption and low accuracy of the conventional particle swarm optimization K-mean clustering method when the number of dimensions of samples is higher. The technical scheme disclosed by the invention comprises the following steps: (1) reading a data set and the number K of clusters; (2) taking statistics on information of dimensionality; (3) standardizing the dimensionality; (4) calculating a similarity matrix; (5) generating a candidate initial clustering center; (6) performing particle swarm K-mean partitional clustering; and (7) outputting a particle swarm optimal fitness value and a corresponding data set class cluster partition result. According to the two-stage hybrid particle swarm optimization clustering method disclosed by the invention, the first-stage clustering is firstly performed by adopting agglomerative hierarchical clustering, a simplified particle encoding way is provided, the second-stage clustering is performed on data by particle swarm optimization K-mean clustering, the advantages of hierarchical agglomeration, K-mean and particle swarm optimization methods are integrated, the clustering speed is accelerated, and the global convergence ability and the accuracy of the clustering result of the method are improved.

The invention discloses a clustering-based automatic driving lane changing scene classification method and a clustering-based automatic driving lane changing scene identification method. The classification method comprises the following steps: 1) collecting data of each sample point needing lane changing scene classification; 2) preprocessing the data in the step 1); (3) taking each preprocessed sample point as a cluster, calculating the distance between the sample point of each cluster and the sample points of all other clusters by adopting an agglomerated hierarchical clustering algorithm, and carrying out combined clustering on two clusters with the closest distance; and 4) taking the clusters which are combined and clustered together as lane changing scenes, and outputting clustering results. According to the automatic lane changing scene classification method and the automatic lane changing scene identification method, an accurate and reliable basis is provided for further improving an automatic lane changing algorithm and an optimization function, and scene classification identification with high efficiency, low cost and high scene coverage rate is realized.

The invention relates to an enhanced cohesion hierarchical clustering-based distribution network user load feature classifying method, which is characterized by comprising steps: calculating a daily load curve characteristic quantity according to an active power curve and a reactive power curve of users; obtaining a daily load characteristic quantity set (see the specification) of N users, an enhanced damping coefficient gamma and a similar coefficient matrix P(X) among all points; forming all groups of merging routes into a merging route set Sg(s), and calculating a hierarchical clustering cohesion process by using a value iteration algorithm; obtaining a group of routes with a minimal similar coefficient value weight sum in the merging route set Sg(s). By using the clustering enhanced cohesion hierarchical clustering algorithm for the characteristic quantities, return values of all results of each layer of clustering are calculated, the cohesion merging route with the maximal return value is selected, the accuracy of the clustering algorithm is improved, defects such as a sensitive initial value, occurrence of a continuous error, and integral deviation of the result of the hierarchical clustering are avoided, and certain measures are adopted to prevent the influence of a singular value on the results.

The present invention relates to a method and system for deploying a backup relay node in an intelligent power distribution communication network. The method includes: acquiring relay node information of the intelligent power distribution communication network; For multiple communication groups; if all communication groups meet the set constraints, then deploy a backup relay node for each communication group. The invention adopts the agglomerative hierarchical clustering method to divide the relay nodes in the intelligent power distribution communication network into communication groups, and configures relay nodes for the communication groups, so that the intelligent power distribution communication network can not only isolate the relay nodes after they fail out of the operating range of the entire network, avoiding network cascading failure reactions caused by the failure of a small number of relay nodes, and after some relay nodes fail, the backup relay nodes will continue to work instead of the original relay nodes, eliminating the The adverse effects brought by the failure of the successor node.