Patents
Literature
Hiro is an intelligent assistant for R&D personnel, combined with Patent DNA, to facilitate innovative research.
Hiro

35results about How to "Clustering results are stable" patented technology

Global K-means clustering method based on feature weight

A global K-means clustering method based on a feature weight is disclosed. The implementation process of the method comprises the following steps of: firstly, clustering data into one class, wherein the optimal clustering centre is the centre of mass of all the data; secondly, figuring out the minimal data point of a target function through calculation and taking the data point as the initial clustering centre of a next class; thirdly, performing iterative updating by a K-means method with the feature weight to obtain the optimal clustering centre of clustering into two classes; and finally, performing updating iteration by orderly increasing the number of the clustering centres by the same method until the clustering of K classes predetermined is completed, thus completing the whole process of clustering all the data points into K classes. In the invention, the new global K-means clustering method based on the feature weight is established by combining the global K-means method with the K-means with the feature weight, and the clustering result is very stable; and in contrast with the experimental results of several K-means clustering methods, the effectiveness and the robustness of the clustering method disclosed by the invention are proved.
Owner:XIDIAN UNIV

Clustering control method of intelligent wireless sensor network based on DPSO (Discrete Particle Swarm Optimization)

The invention relates to a clustering control method of an intelligent wireless sensor network based on DPSO (Discrete Particle Swarm Optimization). The clustering control method comprises the following steps of: (1) receiving ID (identification) and position information of n sensor nodes of the intelligent wireless sensor network; (2) randomly initializing the speed and the position of each particle; (3) calculating a fitness function value of each particle to obtain the smallest fitness function value and the corresponding particle position of all of the particles; (4) updating the speed and the new position of the particle; (5) recalculating the fitness function value of the generation according to the updated particle position by each particle and selecting the particle with the smallest fitness function value as a particle with global optimum; and (6) judging whether iterative round number reaches the maximum iterative round number, if not, returning to the step (4), if so, separating each optimal cluster head and cluster nodes thereof according to the number of the sensor node at the position of the global best particle to form the cluster structure division with global optimum. The invention has uniform cluster and stable cluster result, and is beneficial to reliable communication.
Owner:ZHEJIANG UNIV OF TECH

Entropy weight-based global K-means clustering method

The invention relates to an entropy weight-based global K-means clustering method. The method is realized through the steps of: firstly, clustering data into a class, wherein the optimal clustering center is the centroid of all data; next, determining the minimum data point of an objective function through computation, and taking the data point as an initial clustering center of a next class; then iterating and updating by utilizing an entropy weight-based K-means method to get the optimal clustering centers when the data are clustered into two classes; and sequentially increasing the number of the clustering centers by adopting the same method to iterate and update until the clustering of the set K class is completed, and in this way, completing the whole process of clustering all data points into the K class. According to the method provided by the invention, the global K-means clustering method is combined with K-means with entropy attributes, a new entropy weight value-based global K-means clustering method is constructed, and the clustering results are quite stable; and compared with the experimental results of varieties of K-means clustering methods, the effectiveness and the robustness of the clustering method provided by the invention are proven.
Owner:XIDIAN UNIV

Multi-channel spectrum clustering method based on local density estimation and neighbor relation spreading

The invention discloses a multi-channel spectral clustering method based on local density estimation and neighbor relation spreading. The multi-channel spectrum clustering method based on local density estimation and neighbor relation spreading mainly solves the problem that an existing clustering method cannot carry out clustering on data distributed unevenly in density. The multi-channel spectrum clustering method based on local density estimation and neighbor relation spreading comprises the steps that local density of a sample is estimated and is used as data characteristics and dimension lifting is carried out on original data; a distance matrix, a threshold value and a similarity matrix are calculated, and a neighbor relation matrix is initialized; the neighbor relation matrix and the similarity matrix are updated, and similarity of samples of a subset is updated by the adoption of a local maximum similar value, and an accurate affinity matrix is obtained; a similarity matrix and a normalized Laplacian matrix are calculated; a spectrum matrix is normalized, and a clustering result is obtained through the K-means algorithm. Compared with an existing clustering technology, the multi-channel spectrum method based on local density estimation and neighbor relation spreading enables a more real similarity matrix to be obtained, the clustering result is more accurate and the robustness is better.
Owner:JIANGNAN UNIV

Harmonic duty allocation method based on AP cluster data screening

The invention provides a harmonic duty allocation method based on AP cluster data screening, and the method employs an AP clustering method. Through the clustering of harmonic impedance at a PCC place, the method finally achieves the estimation of system-side harmonic impedance through employing a person correlation coefficient and the sign symbol of estimated harmonic impedance, thereby achieving the harmonic duty allocation. The AP clustering method does not need to define a class number in advance, continuously searches proper clustering centers in an iteration process, automatically recognizes the positions and number of the class centers from data points, and enables the sum of the similarity of all data points to the nearest class representation point to be the greatest. In the beginning, the AP clustering method takes all the data points as the class centers, and achieves the clustering process through the information transmission among the data points. The AP clustering method is a deterministic clustering algorithm, and clustering results of a plurality of independent operations are very stable.
Owner:YUNNAN POWER GRID CO LTD ELECTRIC POWER RES INST

Density-based partitioning and clustering method for K center points in data mining

The invention relates to a density-based partitioning and clustering method for K center points in data mining. The method comprises the following steps that 1, a needed data set is given, and the clustering number K is determined; 2, the densities and average density of data objects are calculated; 3, the minimum density distance value of each data object in the data set is calculated; 4, the minimum density distance values of the data objects in the data set are descendingly sorted, and K data objects corresponding to the minimum density distance values are selected as a clustering center from large to small according to the determined clustering number K, wherein the densities of the K data objects are larger than the average density; 5, the data objects in the data set are distributed to an initial clustering center closest to the data objects, and a clustering result is obtained. The high-quality center points can be selected, subsequent iteration updating steps in a K-means algorithm are not needed, computation complexity is lowered, classification accuracy is improved, high stability is achieved, and operation efficiency is improved.
Owner:无锡中科泛在信息技术研发中心有限公司

User cluster analysis method in customer care system

The invention discloses a user cluster analysis method in a customer care system. The method is characterized by establishing a systematic structure based on the customer care system, an application procedure of data mining and a brand-new algorithm. The method can overcome the defects in user clustering methods of current customer care systems, is scientific and proper, has high accuracy and strong universality, excellent effects, simplicity and reliability.
Owner:NORTHEAST DIANLI UNIVERSITY

Reusable spacecraft surface impact damage feature recognition method

ActiveCN112233099AImprove accuracyMake up for the lack of obvious defect classificationImage enhancementImage analysisMulti objective optimization algorithmFeature extraction
The invention discloses a reusable spacecraft surface impact damage feature recognition method. The method comprises the steps of representing a thermal image sequence of pixel points collected by a thermal infrared imager by a three-dimensional matrix; finding a temperature peak value in the three-dimensional matrix M; performing temperature division on the row of the frame number where the temperature peak point is located; performing temperature division on the column of the frame number where the temperature peak point is located; extracting transient thermal response in a blocking and step-by-step manner; adopting a mean shift algorithm to automatically classify the extracted typical transient thermal responses; adopting a decomposition multi-objective optimization algorithm based ongradient search weight adjustment to select representations of each type of transient thermal response to form a matrix Y; converting the three-dimensional matrix into a two-dimensional matrix, and performing linear change on the two-dimensional matrix by utilizing the matrix Y to obtain a two-dimensional image matrix R (x, y); and performing feature extraction on the two-dimensional image matrixR (x, y) by using a one-dimensional Ostu segmentation algorithm. According to the method, through the uniformly distributed solution sets, the difference and similarity are comprehensively considered,and the defect feature extraction accuracy is improved.
Owner:中国空气动力研究与发展中心超高速空气动力研究所

Aerospace heat-proof material impact damage characteristic type determination method

ActiveCN112215830AImprove accuracyMake up for the lack of obvious defect classificationImage enhancementImage analysisMulti objective optimization algorithmFeature extraction
The invention discloses an aerospace heat-proof material impact damage characteristic type determination method. The method comprises the steps: representing a thermal image sequence of pixel points collected by a thermal infrared imager by a three-dimensional matrix; finding out a temperature peak point in the three-dimensional matrix; performing temperature division on the row of the frame number where the temperature peak point is located; performing temperature division on the column of the frame number where the temperature peak point is located; extracting transient thermal responses ina block-by-block and step-by-step manner; using a mean shift algorithm to automatically classify the extracted typical transient thermal responses; using a decomposition multi-objective optimization algorithm based on adaptive weight adjustment to extract representatives of each type of transient thermal response to form a matrix Y; converting the three-dimensional matrix into a two-dimensional matrix, and performing linear transformation on the two-dimensional matrix by utilizing the matrix Y to obtain a two-dimensional image matrix R (x, y); and performing feature extraction on the two-dimensional image matrix R (x, y) by using a one-dimensional Ostu segmentation algorithm. According to the method, the defects of a traditional clustering method in setting the clustering number and selecting the initial clustering center are overcome, and the accuracy of defect feature extraction is guaranteed.
Owner:中国空气动力研究与发展中心超高速空气动力研究所

S-transform and 2DPCA-combined coherent unit grouping method and system

The present invention relates to an S-transform and 2DPCA-combined coherent unit grouping method and system. According to the method and system, a WAMS phasor measurement device is utilized to obtainthe real-time power angles of the units of a system; fast S-transform is adopted to convert the power angle signals of each generator to a time-frequency feature modulus matrix; the 2DPCA is adopted to perform dimensionality reduction on the time-frequency feature modulus matrix so as to convert the time-frequency feature modulus matrix into a low-dimensional feature index matrix; and the featureindex matrix is inputted to a self-organizing neural network for clustering and recognition. With the method and system of the present invention adopted, online unit grouping can be performed on the system in real time under different operation modes according to the change of different faults and different fault locations, and the time-frequency domain information of the power angles is comprehensively considered. The method and system have the advantages of high recognition accuracy and stable clustering results. With the method and system adopted, necessary premises can be provided for thesimplification of a power grid, the determination of the oscillation center of the system and islanding control, and the safe, efficient and stable operation of the power grid can be ensured.
Owner:STATE GRID FUJIAN ELECTRIC POWER CO LTD +4

Text data stream clustering algorithm based on affinity propagation

The invention discloses a text data stream clustering algorithm based on affinity propagation. The text data stream clustering algorithm is characterized by including the following steps: 1, carrying out dimension reduction processing on a text data set to obtain a corresponding text vector set; 2, obtaining clustering centers of all moments, and completing the clustering algorithm. By means of the text data stream clustering algorithm, the accuracy and the robustness of the algorithm can be improved without assigning the number of clusters in advance, and therefore the requirements for solving practical problems are met.
Owner:HEFEI UNIV OF TECH

Distributed large-scale face clustering method and device

The embodiment of the invention discloses a distributed large-scale face clustering method and device. The embodiment of the invention provides a technical scheme. The method comprises the following steps: clustering to-be-clustered face pictures in batches and summarizing clustering results; obtaining a clustering set and a non-clustering set; extracting a set number of face pictures from each class of the clustering set to form a representative class, calculating the similarity distance between each unclustered face picture and each representative class; and obtaining a similarity distance set, clustering the unclustered human face images into the most similar representative classes based on the similarity distance set, determining a link relationship between each unclustered human faceimage and the corresponding representative class according to a set link threshold, combining the representative classes based on the link relationship, and outputting a combination result. By adopting the technical means, the data calculation amount of large-scale face data clustering can be reduced, the memory consumption is reduced, and the face clustering efficiency is improved on the premiseof ensuring the stability of the clustering result.
Owner:GUANGZHOU PCI TECH SOFTWARE DEV CO LTD +1

User clustering method under massive MIMO system

A user clustering method under a massive MIMO system comprises the following steps: S1, finding and eliminating noise users; S2, obtaining three-dimensional features of the users, and calculating a similarity matrix of the three-dimensional features; S3, performing user clustering by using the AP algorithm according to the similarity matrix; and S4, learning and predicting the state of the user clustering using Q learning to obtain a final clustering result. The method can effectively perform clustering processing on the users, thereby reducing the dimension of the system, reducing the processed dimension to a certain extent, reducing the interference, and improving the resource utilization.
Owner:XIAMEN UNIV +1

Virus sample clustering method and device, electronic equipment and storage medium

The invention discloses a virus sample clustering method, a virus sample clustering device, electronic equipment and a storage medium, which are used for clustering attacked viruses so as to process the attacked viruses and improve the safety. The clustering method comprises the following steps: acquiring behavior characteristics of a virus sample to be clustered; on the basis of a preset reference behavior feature set, generating a behavior identification group for the behavior features of the virus sample, wherein each reference behavior feature in the preset reference behavior feature set has a unique behavior identification; based on the behavior identification group of the virus sample, generating a group characteristic value for indicating a virus family to which the virus sample belongs; based on the group characteristic value of the virus sample, generating a family characteristic value used for indicating a virus family to which the virus sample belongs, each virus family comprising at least one virus family; and clustering the virus samples based on the group feature values and the ethnic group feature values of the virus samples.
Owner:TENCENT TECH (SHENZHEN) CO LTD

Energy storage optimal configuration method based on user side BESS full life cycle

The invention discloses an energy storage optimal configuration method based on a user side BESS full life cycle, and relates to a power and capacity optimal configuration technology of a battery energy storage system and a Fuzzy C-Means clustering method in the field of machine learning. The method comprises the steps of data preparation, typical day acquisition, optimal configuration model constraint condition selection, user-side energy storage optimal configuration model establishment and evaluation index selection and solution, a Fuzzy C-Means clustering method is used for typical day selection, an obtained clustering center is taken as a typical day, and a fuzzy membership degree is assigned to each class cluster for each data point and iteratively updated. The clustering result is stable, the volatility is small, the initial value sensitivity is low, in the model solving process, the model is ingeniously converted into an LP problem, the solving speed is high, the obtained solution is a globally optimal solution, and the example result shows that the model has high practical significance, and the commercial development of energy storage can be effectively promoted.
Owner:国网重庆市电力公司营销服务中心 +2

Method and system for stable and efficient self-adaptive clustering

The invention discloses a method and a system for stable and efficient self-adaptive clustering. The method comprises the following steps of: a, obtaining a set p of input data from p1 to pn, in which n input data is included in the set, and obtaining a threshold value theta of cluster radius; b, adding pi and input data in the set with a distance from the input data pi of smaller than the threshold value theta in a candidate cluster Cpi corresponding to the input data pi, in which the input data pi represents the i-th input data in the set; and c, defining m input data in the candidate cluster Cpi, using a function d (pi, pij) as a distance between the two input data pi, pj, and calculating the input data pi as the probability of a cluster center. The method and the system are applied for establishing a stable and efficient self-adaptive cluster system; the amount of final clusters is not required to be preset, so that the system has calculation efficiency and can realize calculation complexity of o(m2), and the method and the system can be suitable for various mobile intelligent terminals at present.
Owner:WUXI TSINGHUA NAT LAB FOR INFORMATIONSCI & TECH INTERNET OF THINGS TECH CENT

Text clustering intelligent evaluation method based on hybrid clustering

The invention discloses a text clustering intelligent evaluation method based on hybrid clustering. The method comprises the following steps: preprocessing a text set to obtain all feature words in an original text set; deleting high-frequency and low-frequency feature words in the text set to obtain a pre-selected feature subset; calculating corresponding weights of all texts in the original text set by adopting a TF-IDF method; performing binary coding on each feature word in the feature subset, and generating a matrix for the feature words after text feature pre-selection; setting an adaptive value function, performing feature reselection on the feature subset, and selecting a global optimal individual with an optimal adaptive value; decoding the global optimal individual to obtain a final feature subset T; using a TF-IDF method to express a vector on the T, performing normalization and standardization, using a hybrid clustering method to perform text clustering, and selecting a final result of clustering; and generating a word cloud for each cluster in the clustering result. The method has the advantages of being good in clustering effect and small in calculation amount.
Owner:SOUTH CHINA AGRI UNIV

Short text clustering method and terminal device

The invention is applicable to the technical field of information processing, and provides a short text clustering method and a terminal device, and the method comprises the steps: carrying out the preprocessing of a short text set, and obtaining all texts in the short text set; Calculating the similarity between all the texts and other texts according to all the texts; Determining a clustering center of all the texts according to the similarity between all the texts and other texts; And performing clustering processing on all the texts according to the clustering center. According to the embodiment of the invention, the problem that clustering depends on selection and initial division of an initial clustering center in the prior art, so that a clustering result may be different from realdistribution of data set samples, an error result is obtained, or the clustering is very difficult to converge can be solved.
Owner:HEBEI UNIV OF ENG

Clustering algorithm based on minimal spanning tree

The invention discloses a clustering algorithm based on a minimal spanning tree. The clustering algorithm includes the steps of S1, inputting a data set to be clustered and the number of categories K; S2, constructing a minimal spanning tree of the data set; S3, traversing the minimal spanning tree sequentially in a descending order of priority values of nodes; S4, merging between nodes; S5, calculating attributes of a new node after merging; S6, judging whether to end this traversal or not; and S7, judging whether to end the clustering. According to the invention, a skeleton of data points is constructed using the minimal spanning tree, multiple times of traversals and node merging are carried out according to the priority values of the nodes, and the clustering not is ended until the total number of nodes is equal to the number of categories K.
Owner:SHANGHAI NORMAL UNIVERSITY

Data missing value filling method based on functional dependence and clustering

The invention provides a data missing value filling method based on functional dependence and clustering, which comprises the following steps: checking a data set to be processed to obtain a complete data subset and an incomplete data subset; a function dependency set is obtained from the complete data subset by adopting an HYFD algorithm, and function dependencies are sorted in an ascending order; judging whether missing attributes in the incomplete tuple exist in an RHS set of a function dependency set or not; if the missing attribute exists in the RHS set, the incomplete tuple is filled by using the complete tuple in the complete data subset; if the missing attribute does not exist in the RHS set, performing clustering processing on the complete data subset through an improved AP clustering algorithm; using a KNN algorithm to process the incomplete tuple, and finally using the complete tuple to perform filling; the accuracy of the missing value filling algorithm is effectively improved.
Owner:ZHENGZHOU UNIVERSITY OF LIGHT INDUSTRY

Clustering-based small sample load prediction method and device, equipment and storage medium

The invention discloses a clustering-based small sample load prediction method and device, equipment and a storage medium, and the prediction method comprises the steps of: carrying out feature extraction of a historical power load and a to-be-predicted power load to obtain a feature vector; performing integrated clustering on the historical power load and the to-be-predicted power load according to the obtained feature vector to obtain a clustering result; carrying out noise reduction on the clustering result by adopting a wavelet noise reduction algorithm, and carrying out equalization processing on the data after noise reduction to obtain time sequence data with a preset length; inputting the time sequence data with the preset length into the second-order long-short-term memory neural network to obtain a prediction result of the power load; wherein the second-order long-short-term memory neural network is subjected to data training of historical power load and to-be-predicted power load. The method still has excellent prediction performance under the condition that the to-be-predicted power load is scarce.
Owner:GUANGDONG POWER GRID CO LTD +1

Clustering control method of intelligent wireless sensor network based on DPSO (Discrete Particle Swarm Optimization)

The invention relates to a clustering control method of an intelligent wireless sensor network based on DPSO (Discrete Particle Swarm Optimization). The clustering control method comprises the following steps of: (1) receiving ID (identification) and position information of n sensor nodes of the intelligent wireless sensor network; (2) randomly initializing the speed and the position of each particle; (3) calculating a fitness function value of each particle to obtain the smallest fitness function value and the corresponding particle position of all of the particles; (4) updating the speed and the new position of the particle; (5) recalculating the fitness function value of the generation according to the updated particle position by each particle and selecting the particle with the smallest fitness function value as a particle with global optimum; and (6) judging whether iterative round number reaches the maximum iterative round number, if not, returning to the step (4), if so, separating each optimal cluster head and cluster nodes thereof according to the number of the sensor node at the position of the global best particle to form the cluster structure division with global optimum. The invention has uniform cluster and stable cluster result, and is beneficial to reliable communication.
Owner:ZHEJIANG UNIV OF TECH

Opportunistic signal spatial alignment of multi-user bi-directional relay system

A base station-user networking mode in a UDN network is used for the following scenarios: a communication system comprising a plurality of base stations and a plurality of users divides the base stations into different clusters, and the users in the clusters adopt a zero-forcing precoding scheme to perform downlink transmission. The method is characterized by comprising the following two operationsteps: (1) a base station clustering stage: taking geographical coordinate information of base stations as input, dividing the base stations into different clusters by utilizing a mean vector offsetalgorithm, and ensuring that the base stations with closer geographical positions are in the same cluster; and (2) a user network access stage: after the base stations are clustered, the base stationsin the same cluster send the same reference signal, the user receives and calculates a reference signal power value, and when an antenna constraint condition is satisfied, base station-user cooperation finds a network access mode which maximizes the sum of user reference received power values through a greedy algorithm. Compared with the traditional clustering method, the base station clusteringmethod provided by the invention has the advantages that the result is more stable and effective, and the practicability is higher.
Owner:BEIJING UNIV OF POSTS & TELECOMM

A Recognition Method of Impact Damage Feature on Surface of Reusable Spacecraft

ActiveCN112233099BImprove accuracyMake up for the lack of obvious defect classificationImage enhancementImage analysisMulti objective optimization algorithmFeature extraction
The invention discloses a reusable spacecraft surface impact damage feature identification method. The temperature is divided into the row of the frame number; the temperature is divided into the column of the frame number where the temperature peak point is located; the transient thermal response is extracted in blocks and steps; the mean shift algorithm is used to automatically classify the extracted typical transient thermal response; A decomposition multi-objective optimization algorithm based on gradient search weight adjustment is used to select the representative of each type of transient thermal response to form a matrix Y; change the three-dimensional matrix into a two-dimensional matrix, and use the matrix Y to linearly change it to obtain a two-dimensional image matrix R (x, y); use the one-dimensional Ostu segmentation algorithm to perform feature extraction on the two-dimensional image matrix R(x, y). Through the uniformly distributed solution set, the present invention comprehensively considers the difference and similarity, and improves the accuracy of defect feature extraction.
Owner:中国空气动力研究与发展中心超高速空气动力研究所

Super-pixel-based prototype spectrum set generation method

The invention discloses a prototype spectrum set generation method based on superpixels, and belongs to the field of hyperspectral image processing, and the method specifically comprises the steps: firstly, collecting spectrum wave bands of a certain region, taking each wave band as an image, and forming an image set X; then, 40% of wave bands are randomly selected, pixel points in the image corresponding to the wave bands are averaged, and an image Y is formed; thirdly, SSIM structure indexes of the images in the image set X and the image Y are calculated one by one to serve as scores of the images, and the scores are arranged in a descending order; re-selecting the first 40% wave band to regenerate the image Y, and repeatedly calculating the score of each image in the image set X until the wave band is stable; and finally, selecting the first three wavebands with the highest score, inputting the first three wavebands into a superpixel segmentation algorithm to extract superpixel small blocks at fixed intervals, taking the average spectrum of each superpixel small block as an initial clustering center of a k-means clustering algorithm, and performing clustering to obtain a prototype spectrum set of the region. According to the method, the calculation amount of the clustering process is reduced, and the clustering result is stabilized.
Owner:BEIJING UNIV OF POSTS & TELECOMM +1
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products