47results about How to "Improve clustering accuracy" patented technology

The invention discloses a fast multi-view clustering method and system for anchor point diagram decomposition based on logarithm sparse constraint, a storage medium and computer equipment, and belongs to the field of machine learning and data analysis, and the method mainly comprises the following steps: reading a vectorized multi-view data set, and generating anchor points in all views; constructing an anchor point diagram between the original data and an anchor point set formed by the anchor points; performing anchor point diagram decomposition on the anchor point diagram to obtain a basis matrix and a coefficient matrix, performing log norm constraint on the basis matrix, and removing redundant information; and performing K-means on the coefficient matrix to obtain a clustering label, and generating a clustering result. According to the method, the clustering precision and efficiency can be remarkably improved when a large-scale multi-view clustering task is processed, and the method can be widely applied to the field of data analysis.

The invention discloses a self-supervised clustering method for massive single cell data. The method comprises the following steps: 1) downloading single cell RNA sequencing data from a disclosed gene expression database; 2) preprocessing the single-cell RNA-seq data, wherein the preprocessing comprises filtering, standardization and feature selection; (3) carrying out preliminary clustering on the preprocessed single cell RNA sequencing data by utilizing a rapid Louvain algorithm; 4) screening representative cell points from the preliminary clustering result by using an index of clustering representativeness; 5), after representative cell points are screened out from each preliminary cluster, a Transform model is used, and a final clustering result is obtained.According to the method, the rapid Louvain algorithm and the Transform neural network are combined, the efficiency and accuracy of clustering analysis are remarkably improved, and an efficient and accurate solution is provided for analysis of mass single cell data.

The embodiment of the invention relates to a layout processing method, electronic equipment and a computer readable storage medium. The method comprises the following steps of: determining the similarity of each sub-layout on the basis of geometrical characteristics of graphs in a plurality of sub-layouts formed by segmenting the layout, wherein the sub-layouts have the same size; pre-classifying the sub-layouts based on the similarity to determine a classification number; determining a geometric feature value of the geometric feature of the graph based on the geometric feature of the graph; and clustering the sub-layouts based on the classification number and the geometric feature value to determine the type of each sub-layout. According to the technical scheme provided by the invention, more efficient clustering and higher clustering precision can be realized.

The invention discloses an electric vehicle charging aggregation and adjustment prediction method and system based on integrated clustering, and relates to the technical field of electric vehicle and power grid interaction, and the method comprises the steps: collecting the charging demand data of an electric vehicle, carrying out the classification and clustering, and integrating different types of clusters; analyzing the charging demand data and screening outlier samples; establishing an electric vehicle charging load characteristic index model, and predicting a charging load curve of the electric vehicle; evaluating the theoretical potential of the electric vehicle participating in the demand response; and optimizing a power grid access mode based on the electric vehicle charging load characteristic index and the theoretical potential of the demand response. The method integrates the advantages of a plurality of clustering algorithms, and improves the accuracy and reliability of data analysis. By integrating different clustering algorithms, the respective defects can be complemented, so that the clustering precision is improved, the effective identification of the user behavior and the charging mode of the electric vehicle is ensured, and the stability of the overall clustering result is improved.

The invention relates to an environmental signal blocking method based on characteristic information entropy, which comprises the following steps of: blocking partial environmental signal radiation source signals through a wave limiter according to frequency parameter information, comparing and matching signals detected in a complex electromagnetic environment with a radiation source identification database, and identifying the radiation source identification database; blocking non-combat signals matched with information in a database, clustering unrecognized radiation source signals based on carrier frequency, pulse width and arrival angle parameter information, calculating characteristic information entropy of pulse width parameters according to a clustering result, comparing the characteristic information entropy of the pulse width parameters with a set threshold value, and determining whether the radiation source signals cannot be recognized or not according to the characteristic information entropy of the pulse width parameters. The radiation source signal composition condition of the pre-sorting clustering result is judged, so that the number of radiation sources in different categories is estimated; in a word, the method has the advantages that the algorithm processing pressure can be relieved, the two-dimensional parameter joint clustering accuracy is improved, and environmental signals can be quickly and effectively filtered out.

The invention discloses an end-to-end anchor point multi-view image clustering method under the guidance of global principal components, and particularly relates to the field of image processing. Comprising the following steps: acquiring a sub-view sample matrix of each view, and determining a feature mapping matrix; obtaining a global anchor point matrix and an anchor point diagram matrix; obtaining an anchor point fuzzy label matrix in the anchor point diagram matrix, and obtaining a sample fuzzy label matrix based on the anchor point fuzzy label matrix; establishing a target function based on the sub-view sample matrix, the feature mapping matrix, the global anchor point matrix, the anchor point diagram matrix, the anchor point fuzzy label matrix and the sample fuzzy label matrix; the feature mapping matrix is determined again, the sub-view sample matrixes of all the views are converted into the unified potential space again through the feature mapping matrix until the target function converges, and a final sample fuzzy label matrix is output. Unification of objective functions can be realized, so that model optimization is facilitated to obtain a globally optimal solution, and clustering precision is improved.

The invention discloses an automatic unauthorized vulnerability detection system and method, and the system comprises a user behavior collection module, a user behavior graph analysis module, an attack case generation module, and an execution and report module. A user group behavior graph is generated, pruning, subgraph division, function point and common point segmentation and key path discovery are executed, test cases for unauthorized vulnerabilities are more efficiently and accurately generated, a large number of useless test cases are filtered, the cases are automatically executed, and the test efficiency is improved. And a test execution result is judged by utilizing a page similarity calculation method without a common subtree, and a detection result is fed back, so that the efficiency and the accuracy of detecting the unauthorized vulnerability are greatly improved.

The invention discloses a risk assessment method and device, a computer readable storage medium and electronic equipment. The method relates to the field of artificial intelligence, and comprises the following steps: obtaining target data of N users, the target data comprising user data and credit data of the users; the N users are clustered according to the target data of the N users through a target clustering algorithm, M user clusters are obtained, and an initial clustering center set of the target clustering algorithm is determined based on an eagle optimization algorithm; and for each user cluster, evaluating the credit risk of the first user in the user cluster to obtain a risk evaluation result, and determining the risk evaluation result as a risk evaluation result corresponding to each user in the user cluster. Through the credit risk assessment method and device, the problem that the accuracy of credit risk assessment is affected due to poor clustering accuracy caused by randomly determining the initial clustering center of the clustering algorithm when the credit risk of the user is assessed based on the clustering algorithm in the related technology is solved.

The invention discloses a hyperspectral image clustering method and system based on a spatial-spectral joint dynamic hypergraph, and the method comprises the steps: determining a low-rank representation coefficient of a hyperspectral image, and obtaining a Laplace regular term of the hyperspectral image hypergraph and the weight of a hyperedge; according to the Laplacian regular term and the weight of the hyperedge, constructing a similarity-maintained low-rank subspace representation optimization model of the hypergraph structure; iteratively solving the low-rank subspace representation optimization model for keeping the similarity of the hypergraph structure to obtain a final low-rank representation coefficient, determining a Laplacian regular term of a new hypergraph and the weight of a hyperedge according to the low-rank representation coefficient solved in each round in the iteration solving process, and calculating the weight of the hyperedge by utilizing the Laplacian regular term of the new hypergraph. Performing the next round of solution; and determining a clustering result of the hyperspectral image according to the final low-rank representation coefficient. And the clustering precision is improved.

The present application discloses an image clustering method, an electronic device and a computer-readable storage medium, which is applied to a cloud storage system, and comprises: performing image clustering processing on at least two acquired images to obtain multiple initial image sets; determining a historical clustering set matching a target object of each image from an acquired historical clustering set library; constructing an undirected graph based on a historical set tag set formed by each initial image set and a historical set tag of a historical clustering set matching a target object in the initial image set; performing set merging processing on each initial image set based on the undirected graph to obtain a target image set. This can improve the accuracy of image clustering.

The invention discloses a hyperspectral image clustering method, system and device, a medium and a product, and relates to the field of image data processing. The method comprises the following steps: acquiring a hyperspectral image; adopting a superpixel segmentation method to convert the hyperspectral image into a superpixel-level graph; the superpixel-level graph is regional-level graph data; obtaining a clustering model; and inputting the super-pixel-level graph into a clustering model to obtain a clustering result. Wherein the clustering model comprises an adaptive filter graph encoder, a graph embedded clustering self-training decoder and an isomorphic enhancement structure learning module; the adaptive filter graph encoder is used for extracting low-frequency features and high-frequency features; the graph embedding clustering self-training decoder is used for carrying out self-training and obtaining a clustering result by utilizing a k-means algorithm; and the isomorphic enhancement structure learning module is used for dynamically adjusting an input graph structure according to different clustering tasks. According to the invention, the clustering precision of the hyperspectral image can be improved.

The invention relates to the technical field of intelligent traffic, and discloses a method and system for identifying clusters with road lane change features, and the method comprises the following specific steps: S1, carrying out the clustering of lane lines of a target road to be subjected to lane change; s2, extracting geometric information of the cluster 1; s3, judging whether the obtained geometric information is in a list form or not; S4, obtaining long edges and short edges; s5, obtaining a long side point sequence list; s6, carrying out space coverage judgment to obtain a lane changing pre-selection list; s7, re-organizing lane line information; s8, judging whether the lane line information is greater than a threshold value or not; s9, obtaining a point sequence of geometric information of the lane line, and obtaining a long side and a short side of the point sequence; further obtaining a point sequence list; s10, further screening a point sequence list; and S11, obtaining all families with road lane changing characteristics. The method solves the problems that in the prior art, a prior sample is relied on, and a lane line with lane changing characteristics is difficult to accurately recognize, and has the advantages of being capable of improving clustering precision, flexible and easy to use.

The present disclosure relates to the field of computer technology, and specifically provides a vehicle data clustering method and a vehicle trajectory restoration method. The vehicle data clustering method includes: obtaining at least two clustering results obtained according to vehicle data clustering processing, and restored trajectory information corresponding to each of the clustering results; for any target clustering result among the at least two clustering results, based on the similarity between the target clustering result and other clustering results, determining a candidate clustering set from the other clustering results; wherein the other clustering results are clustering results other than the target clustering result among the at least two clustering results; based on the restored trajectory information of each candidate clustering result in the candidate clustering set and the similarity with the restored trajectory information of the target clustering result, updating the target clustering result. The disclosed method improves the clustering accuracy of vehicle data.

The invention discloses a user load curve analysis method and system based on Bayesian nonparametric clustering and a time covariance structure, and relates to the technical field of power system data analysis, and the method comprises the steps: collecting intelligent electric meter data, and carrying out the preprocessing of the original intelligent electric meter data; calculating a covariance matrix of the preprocessed data; calculating a mahalanobis distance between the load curves based on the covariance matrix; and constructing a distance dependent Chinese restaurant process model based on mahalanobis distance, performing link allocation and clustering parameter inference in combination with Gibbs sampling, and performing adaptive clustering to obtain a user load curve analysis result. According to the method, the problem of singularity of a covariance matrix under a high-dimensional small sample is solved by adopting Leidoit-Wolf shrinkage estimation, complete Bayesian inference is realized, the clustering number is adaptively determined, and manual intervention is not needed; while high clustering precision is maintained, calculation complexity is significantly reduced, and the method is suitable for large-scale actual data.

This application relates to an image processing method, apparatus, electronic device, and storage medium, which can be applied to the field of mapping. The method includes: acquiring an image of an object to be clustered and a target image of an object; determining a first object image to be segmented and a second object image to be unsegmented from the image of the object to be clustered and the target image of the object; the size of the first object image is larger than the size of the second object image; segmenting the first object image to obtain at least two sub-images; inputting the second object image and each sub-image into an image matching model for image matching processing to obtain a first matching result; and determining a target image set of the image of the object to be clustered based on the first matching result. According to the technical solution of this application, the accuracy and efficiency of image clustering can be improved.

The invention discloses a method for generating a personalized question practice plan for intelligent integration of wrong questions, comprising: collecting students' wrong question text data, dividing the wrong question text data into different wrong question categories by adaptive clustering; extracting knowledge points involved in different wrong question text data based on co-occurrence information of word segmentation results of the wrong question text data in different wrong question categories; calculating the similarity of knowledge points between wrong question text data in combination with knowledge points and difficulty statistics of the wrong question text data, and constructing a wrong question knowledge point similarity matrix; constructing an answer tendency matrix according to students' historical answer question data and students' unanswered question data in a question bank and performing matrix decomposition to obtain a semantic feature vector of each question; constructing a personalized question practice plan generation model to calculate students' answer tendency on different questions, and selecting the first several unanswered questions from high to low according to the answer tendency to generate a personalized question practice plan.

The application discloses a laser radar point cloud object detection method and system, comprising the following steps: obtaining a three-dimensional point cloud of a laser radar, and after filtering interference point clouds and compensating for missing point clouds, obtaining a to-be-segmented point cloud; generating a two-dimensional distance image from the to-be-segmented point cloud; setting an angle constraint; the angle constraint is that the included angle between the line connecting two adjacent points in the two-dimensional distance image and the line connecting the radar center and the point farther away from the radar center among the two adjacent points is not less than an angle threshold; setting a distance constraint; the distance constraint is that the distance between the two adjacent points in the two-dimensional distance image is not greater than a distance threshold; traversing the points in the two-dimensional distance image, judging whether the two adjacent points satisfy the angle constraint or the distance constraint, if any constraint is satisfied, marking the two adjacent points as the same object, otherwise, marking the two adjacent points as different objects, and thus obtaining an object segmentation result. The object can be quickly detected under sparse point cloud data.

The invention discloses a local time-sharing weather classification and prediction method based on historical photovoltaic output data. The method comprises the following steps: firstly, dividing effective photovoltaic output in a day into a plurality of time periods, constructing a fluctuation feature vector, and describing the fluctuation of the photovoltaic output by adopting the fluctuation feature vector; and then, clustering the fluctuation feature vector sequences in each sub-time period by using a K-means algorithm to realize classification marking of historical time-sharing weather types. Thirdly, performing multivariable phase space reconstruction on the historical fluctuation feature sequence, establishing a convolution-long and short-term memory network prediction model, and inputting a phase space reconstruction result into the convolution-long and short-term memory network prediction model to obtain a sub-time period fluctuation feature sequence of a prediction day; and finally, according to the distance between the sub-time period fluctuation feature sequence of the prediction day and the clustering center value of each typical weather type, determining the weather type of the divided time period of the prediction day, and solving the problem that the existing weather prediction method is insufficient in precision and space / time resolution.

This application provides a mask layout correction method, system, medium, program product, and terminal based on a combination of membrane computing and the DBSCAN algorithm. The method involves obtaining an initial layout, performing preliminary clustering on the initial layout using a preset clustering method to obtain several partitions and their corresponding density and contour features; obtaining initial density and contour parameters for DBSCAN clustering; iteratively optimizing the initial clustering parameters of the DBSCAN clustering using a preset membrane computing algorithm based on the density and contour features of each partition to obtain the optimal clustering parameters for each partition; performing DBSCAN clustering on each partition based on the corresponding optimal clustering parameters to obtain the core region, boundary region, and sparse region of the initial layout; and correcting the core region, boundary region, and sparse region based on a preset correction strategy to obtain the corrected layout of the initial layout. This improves the accuracy of mask layout correction.

This invention relates to the field of e-commerce data processing technology, specifically to a method and apparatus for multi-platform e-commerce order aggregation based on dynamic clustering rules. The technical solution includes: collecting order data from e-commerce platforms and performing standardized processing to obtain platform order data; extracting various clustering feature values ​​from a clustering feature system; matching these values ​​with rules in a dynamic clustering rule set to determine the candidate cluster to which the order belongs; calculating the comprehensive similarity between orders; determining whether the comprehensive similarity is greater than or equal to a preset similarity threshold, and assigning the order to that cluster; performing missed order detection for orders not clustered; and outputting the order results for the clusters. This significantly improves clustering accuracy, effectively reduces missed orders, and minimizes operating and transportation costs. It also significantly improves order processing efficiency, adapting to the needs of large-scale order processing. The system operates stably and supports efficient processing of massive amounts of data.

The application discloses a multi-target heuristic single cell deep clustering method, and the multi-target heuristic single cell deep clustering algorithm is obtained by improving the scDHA algorithm. The multi-target heuristic single cell deep clustering algorithm firstly adopts a stacked Bayesian autoencoder algorithm and an Autoencoder algorithm to jointly perform multiple dimension reductions on single cell transcriptome sequencing data after preprocessing, then adopts a K-means algorithm and an Ultra-Scalable Spectral Clustering algorithm to jointly cluster multiple dimension reduction data to generate multiple clustering results, and finally adopts a dynamic positive and negative sine optimization algorithm to calculate a unique clustering result from the multiple clustering results. The method has the advantages of fast clustering speed, small hardware memory consumption, low hardware cost, high clustering precision and strong robustness.

This invention relates to the field of data processing technology, specifically to an artificial intelligence-based production line scheduling optimization method. The method obtains the adjustment time and corresponding change magnitude based on the data change characteristics of parameter adjustment curves; it obtains the process difference significance based on the distribution characteristics of the adjustment times of parameter adjustment curves for all orders under the same process and the change magnitudes of all adjustment times; it obtains the target process based on the difference significance of all processes; and it obtains the comprehensive difference degree based on the difference characteristics of the parameter adjustment curves of the target process between different orders. This invention clusters all orders based on the comprehensive difference degree and order production time to obtain different order clusters; and it schedules and allocates production lines based on all order clusters, which can reduce equipment wear and tear and improve production efficiency.

The invention discloses a power grid database security protection method and device, equipment and a medium, and relates to the technical field of data security access, the method comprises the following steps: identifying a plurality of groups of historical access logs of a power grid database to obtain a plurality of historical access behaviors, and analyzing and extracting to obtain corresponding access behavior characteristics; performing density-based clustering analysis on all the access behavior characteristics to obtain a plurality of behavior modes; predicting a risk coefficient of each behavior pattern based on a long short-term memory network model, and classifying the plurality of behavior patterns into a plurality of risk levels; matching a first behavior mode corresponding to the plurality of behavior modes according to the current access behavior, determining a protection strategy according to a first risk level determined by the first behavior mode in combination with the behavior intention of the current access behavior, and performing security protection; wherein the behavior intention is obtained based on analysis of the current access behavior. According to the method, the access behavior can be accurately identified, so that the power grid database is reasonably and safely protected.

The invention discloses an open world target detection category discovery method based on semi-supervised contrast clustering. The method comprises the following steps: acquiring an original data set of the open world and a new data set of the open world; constructing an open world target detection category model based on semi-supervised contrast clustering for identifying known categories of the open world targets and clustering and identifying unknown categories of the open world targets; based on the training data in the original data set of the open world and the training data of the new data set of the open world, training an open world target detection category model based on semi-supervised contrast clustering to obtain a trained open world target detection category model based on semi-supervised contrast clustering; and inputting test set data in the new data set of the open world into the trained open world target detection category model based on semi-supervised contrast clustering to realize recognition of a new target category of the open world. The method provided by the invention can adapt to newly appearing object categories in a dynamic environment.

The application discloses a kind of dam deformation prediction method, equipment, storage medium and product, the prediction method includes preprocessing to dam monitoring data;The environmental data after pretreatment is extracted to depth feature, and environment characteristic is obtained;The reservoir water level after pretreatment and environment characteristic are carried out cluster analysis, and according to the time domain division of time factor and dam survey point displacement according to cluster analysis result, obtain the cluster division result of different time periods;According to each cluster division result, respectively construct sample data set;Each sample data set is used to train and test Transformer-LSTM model respectively, and the dam deformation prediction model of each time period is obtained;Obtain the data to be predicted, and the dam deformation prediction model corresponding to the time period of the data to be predicted is used to predict the prediction data, and the dam deformation prediction result is obtained.The dam deformation prediction precision of the application is improved.

The invention discloses a structural representation and clustering combined learning method and system for multi-source image fusion. The method comprises the steps that samples are selected from a sample matrix set of multi-source images to construct an initial dictionary matrix set, and an initial transformation matrix set, a weight vector, a slack variable, a Lagrange operator, a penalty factor and the number of iterations # imgabs0 # are generated; calculating a transformation matrix set, a structure characterization matrix, a slack variable, a weight vector, a Lagrange operator, a penalty factor and a structure characterization matrix of # imgabs 1 # iteration, and if a preset termination condition is satisfied, performing k mean value processing on the structure characterization matrix of # imgabs 2 # iteration to obtain a clustering result; and otherwise, adding 1 to the number of iterations # imgabs3 #, and returning to continue iteration. According to the method, complementarity information among multi-source image data can be effectively mined, structural characterization with high discrimination can be obtained, and the method has the advantages of high universality and robustness, high calculation efficiency, noise interference resistance and high recognition precision.

The present invention provides a clustering method, device, electronic device, and computer-readable storage medium, wherein the clustering method includes: clustering a feature set in a database based on a first similarity threshold, thereby obtaining multiple first-class clusters; merging at least two first-class clusters in the multiple first-class clusters based on a second similarity threshold and a constraint condition; wherein the second similarity threshold is less than the first similarity threshold. This can solve the problem of multiple levels of one target appearing at a high threshold and improve clustering accuracy.

The disclosed embodiment proposes a clustering method and device, an electronic device and a storage medium, the method comprising: obtaining a neural network for correcting the similarity of facial features, the neural network being trained based on difficult samples; the difficult samples being samples in at least one category of face data set; processing images to be clustered based on the neural network to obtain similarity correction values ​​of facial features of the images to be clustered; correcting feature similarity using the similarity correction values ​​to obtain corrected feature similarity; the feature similarity represents the similarity of facial features between the images to be clustered and images of the corresponding category; clustering the images to be clustered based on the corrected feature similarity to obtain clustering results of the images to be clustered and the images of the corresponding category.

The invention relates to a multi-view cancer subtype data clustering method based on a convolutional auto-encoder and knowledge distillation, and belongs to the field of machine learning. The method comprises the following steps of: rearranging characteristics of multi-omics data by using a Pearson correlation coefficient, reducing the redundancy of the data and improving the representativeness of the characteristics; and inputting the rearranged multi-omics data into a convolutional auto-encoder model, and training a teacher model to capture a high-dimensional structure and features of the data. Knowledge in the teacher model is transmitted to a lightweight student model by using a knowledge distillation algorithm, and the student model is trained by optimizing a loss function (including reconstruction error and distillation loss). And based on the trained student model, performing clustering analysis on the multi-view cancer subtype data by using KL divergence to obtain a final clustering result.

The present invention belongs to the technical field of unsupervised image recognition, and discloses a deep image clustering method, system, device, medium and terminal, wherein the deep image clustering method comprises: inputting a group of pictures, obtaining multiple sets of codes for each picture through an encoder in a SIMO autoencoder; inputting the multiple sets of codes for each picture and the codes after information fusion of these codes into the same decoder to obtain reconstructed pictures; calculating the reconstruction loss of the reconstructed pictures and the pictures obtained after data enhancement of the input pictures, and performing pre-training of the network; after the pre-training is completed, clustering the fused codes to obtain the comparative clustering loss, adding it into the training process for joint training, and realizing parameter optimization; after the joint training is completed, feature extraction is performed on the entire data set, and the fused codes are obtained and input into the traditional clustering algorithm to obtain the clustering result. The present invention can simultaneously extract semantic information from multiple angles of a single image, thereby increasing the richness of information.