124 results about "Affinity propagation" patented technology

The invention provides a method for automatically screening influential users on social media networks. The method comprises the following steps of: modeling for users, interested objects and interaction relationship between the users and the interested objects by a hypergraph model; automatically learning to obtain a connotative interest topic by using a regularization topic probability model of hypergraph restraints and taking the similarity relationship between the content information and the content information of the interested objects as restraints; and performing topic influence sorting on each user and interest object, propagating topic influence by using a similarity propagation model, the users and the interest objects on the hypergraph and the hyperedge between the users and the interest objects until to reaching a steady state, and then sorting to obtain the influential users under specific topics. By the method, the distribution of the user influence in the social media network can be truly and accurately reflected.

The invention discloses a clustering analysis-based intelligent fault diagnosis method for an antifriction bearing of a mechanical system. A diagnosis model is trained firstly, comprising the following steps: collecting standard vibration signal samples of five fault and normal bearing states of an outer ring, an inner ring, a rolling body and a holding frame; decomposing signals, extracting original vibration signals as well as time domain and frequency domain characteristics of decomposed components to obtain an original characteristic set; removing redundancy by means of a self-weight algorithm and an AP (Affinity Propagation) clustering algorithm to obtain Z optimal characteristics; classifying sample statuses by means of the AP clustering algorithm to obtain a well-trained diagnosis model. A fault diagnosis is performed by the following steps: collecting real-time vibration information of a bearing, decomposing the signals, extracting the optimal characteristics determined by the model, importing the AP clustering algorithm to cluster parameters based on the diagnosis model, comparing with the Z characteristics known in the model to obtain a category of a current unknown signal, so as to complete the fault diagnosis. According to the clustering analysis-based intelligent fault diagnosis method disclosed by the invention, both EEMD (Ensemble Empirical Mode Decomposition) and WPT are utilized to decompose the vibration signals, more refined bearing status information can be acquired, the self-weight algorithm and the AP clustering algorithm increase intelligence of the diagnosis, and therefore accurate diagnosis is ensured.

The invention discloses an image segmentation method based on super pixel clustering. More specifically the image segmentation method includes the steps of: 1, segmenting images by using a Simple Linear Iterative Clustering (SLIC) algorithm, and generating super pixels; 2, improving a construction mode of a similarity matrix about the super pixels, and fusing a color characteristic and a texture characteristic through non-symmetry of the similarity matrix; 3, clustering the super pixels through an Affinity Propagation (AP) clustering algorithm based on the similarity matrix; and 4, reaching the purpose of image segmentation by adding spatial information of the super pixels and dividing a disconnected region into different types of super pixel groups by means of breadth-first traversal. The image segmentation method based on the super pixel clustering has good segmentation effect and a fast convergence speed. Target objects can be effectively segmented without arrangement of target quantity.

The invention discloses a holographic touch interactive exhibition system with multisource input and intelligent information optimizing functions. The system senses surrounding environment information by the aid of a plurality of sensors, a smoke sensor is started to monitor fire without audiences, faces in videos are recognized when the audiences are present, proper broadcasting information is recommended, and the identity sense of customers is enhanced. The exhibition system introduces an interactive mode, and the audiences can realize man-machine interaction by the aid of a holographic touch screen to acquire information that the audiences want to know. Besides, by the aid of an intelligent information mining algorithm and features extracted by a color co-occurrence matrix, finding of a facial expression recognition area of interest and feature extraction are combined into one step, the faces are clustered by the aid of an affinity propagation clustering algorithm, and the broadcasting information is changed according to emotion. By the aid of the exhibition system, the information can be effectively pushed according to ages and genders of the audiences in a complicated background environment, sufficient exhibition can be realized, the audiences can be guided to more comprehensively know the information, and the identity sense of the audiences is enhanced.

The invention discloses a method for microblog hot event online detection based on emotion analysis. According to the method for microblog hot event online detection based on emotion analysis, emotion extraction is conducted on online microblog texts according to emotion analysis, the bursting state of emotions and emotion texts is detected according to the improved Kleinberg algorithm, hot topics are extracted according to an affinity propagation clustering algorithm in the bursting period, so that only a small part of texts with a large amount of effective information participate in event extraction, the data volume is reduced, meanwhile a large amount of data noise is filtered out, and therefore the purpose that online real-time detection of the microblog hot events is achieved.

The invention discloses a Wi-Fi-based indoor positioning navigation method. The Wi-Fi-based indoor positioning navigation method includes an acquisition stage, an online positioning stage and a navigation stage. According to the acquisition stage, RSS fingerprint information which is acquired from reference points in the same direction is clustered through an affinity propagation clustering algorithm, so that different classes and class leaders can be formed, and a fingerprint database can be established and is stored in a server. According to the online positioning stage, a handheld terminal receives the RSS fingerprint information and sends the same to the server; the server performs class matching on the received RSS fingerprint information and fingerprint information of the server; a fingerprint graph matrix is formed according to a matched class leader set and is transmitted to the handheld terminal for precise positioning; norm minimization processing and corresponding weighing are performed on rough positioning information through using an improved compressed sensing model; and the positioning of coordinates can be completed. According to the navigation stage, a shortest path is searched through utilizing a quantum cellular ant algorithm; and a position is tracked and navigated in real time through using Kalman filtering system. The Wi-Fi-based indoor positioning navigation method of the invention has the advantages of simple algorithms, small workload of affine clustering and matching, short response time and small positioning error.

The invention discloses a method, equipment and a device for improving an intelligent customer service response rate, and a storage medium. The method comprises the following steps that: obtaining thenon-response questions of intelligent customer service, and classifying the non-response questions to obtain knowledge type questions; calculating a first similarity among knowledge type questions; according to the first similarity, through an affinity propagation clustering algorithm, clustering each knowledge type question to obtain a plurality of clustering clusters; taking the clustering center of each clustering cluster as a target standard question, and obtaining a target standard answer corresponding to the target standard question; and adding the target standard question and the corresponding target standard answer into an intelligent customer service knowledge base. In the method, the knowledge type questions in the non-response questions are clustered through the affinity propagation clustering algorithm, the clustering center of each obtained clustering cluster is taken as the target standard question to be added into the intelligent customer service, and the clustering center is a question with the highest representativeness in each clustering cluster so as to improve the intelligent customer service response rate and improve user experience.

The invention relates to a novel power system load curve clustering method. The novel power system load curve clustering method comprises the following steps: step S1, obtaining daily load curve active power data in five or more days of loads or transformers to be classified from a monitoring, controlling and data-collecting system; step S2, taking a mean value of all workdays at every moment of each load or transformer as a user typical load curve according to the daily load curve active power data obtained in the step S1; step S3, carrying out standardization treatment on the user typical load curve obtained in the step S2; step S4, culturing the user typical load curve by using an affinity propagation algorithm to obtain preliminary classification results; step S5, calculating a user load form characteristic index, and carrying out min-max standardization treatment to obtain a standardized load form characteristic index; and step S6, carrying out secondary classification on the preliminary classification results in the step S4 based on the standardized load form characteristic index obtained in the step S5.

The invention belongs to the technical field of network security and particularly relates to an APDE-RBF (Affinity Propagation Differential Evolution-Radial Basic Function) neural network based network security situation prediction method. The APDE-RBF neural network based network security situation prediction method comprises the steps of dividing and clustering sample data by utilizing an AP clustering algorithm to obtain the number of nodes of hidden layers of the center and network of the RBF; obtaining population diversity by using AP clustering, changing a zoom factor and a crossover probability of a DE algorithm adaptively and optimizing the width and connection weight of the RBF; meanwhile, performing chaotic search on elite individuals and population diversity center of each generation of population in order to avoid falling into local optimization and jumping out of a local extreme point; and determining a final RBF network model, inputting a test dataset and outputting a situation prediction value. The APDE-RBF neural network based network security situation prediction method aims at improving the prediction precision for the network security situation while enhancing the generalization ability.

The invention discloses a facial expression recognition method based on facial motion unit combination features. The facial expression recognition method based on the facial motion unit combination features is used for solving the technical problem of low recognition rate of a single facial motion unit with an existing facial expression recognition method based on facial motion units. The technical scheme includes that the facial expression recognition method based on the facial motion unit combination features comprises the steps of building a large-scale facial expression data base, carrying out clustering of each category of facial expressions and corresponding training samples by using the affinity propagation (AP) clustering algorithm, judging arbitrary unit (AU) combinations in each sub-category, determining the number of the sub-categories under the same facial expression by combining main AU combinations, generating the number of the categories of the training sample by combining the sub-categories of all facial expressions, and carrying out classification training by using the support vector machine (SVM) method. According to the facial expression recognition method based on the facial motion unit combination features, the recognition rate of the single facial motion unit is improved, namely that average recognition rate of a single AU unit is improved from 87.5% in prior are to 90.1% and by 2.6%.

The invention is suitable for the technical field of intrusion detection and provides an intrusion detection method and device. The intrusion detection method includes subjecting an intrusion detection data set KDD CUP99 to data standardizing to obtain a high-dimension vector set with n dimensions; subjecting the high-dimension vector set to dimension reduction to obtain a low-dimension vector set with m dimensions, wherein m is smaller than n; utilizing a semi-supervised affinity propagation clustering algorithm to process the low-dimension vector set to obtain a clustering result; determining a normal cluster and at least one abnormal cluster in the clustering result according to data size in obtained clusters; acquiring data of the abnormal clusters, wherein the data are abnormal data obtained by intrusion detection. By the intrusion detection method, probability distribution of data class is enabled to be close to original distribution obtained by using all attributes as much as possible, calculating speed of the subsequent clustering algorithm is increased greatly, and abnormality detection accuracy of an intrusion detection model is improved.

The invention discloses a power grid user classification method based on a load characteristic index weighted clustering algorithm. Load curve clustering is introduced, the problem of high complexityof an existing algorithm is solved, load characteristic indexes are built, the dimension of a traditional AP (affinity propagation) algorithm is reduced to improve the efficiency of the algorithm, similarity calculation is improved by weight, and convergence judgment is performed according to DB (Davies-Bouldin) indexes. In order to overcome the shortcomings of subjective weight, characteristic index contribution degree evaluation rules are designed, the weight of the load characteristic indexes is determined objectively and adaptively by an entropy weight method to measure the distinction degree of the characteristic indexes for clustering results, weight assignment reasonability is improved, and accuracy of user classification results is finally improved.

The invention discloses an extensible partition method for associated flow graph data, comprising the following steps: (1) preprocessing graph data and converting a node as the node ID; (2) adding each preprocessed side which is existed in an ID form in FIFO and waiting to enter into a sliding window to process; 3) completing the sliding window and calculating the PageRank value of each node according to a mixing approximation PageRank method; (4), tracking the collection of the initial node corresponded by each node in a subgraph, which is composed of sides in the sliding window and obtaining associated value of each node; (5) obtaining a plurality of centroids and corresponding clusters by adopting an affinity propagation clustering algorithm to all nodes in the window; (6) obtaining a plurality of partitioning results which are not of uniform size via associating repeated iteration of a clustering method and storing the partitioning results whose scales are less than threshold value in the sliding window; (7) transmitting the partitioning results whose scales reach or exceed the threshold value to the appropriate storage node by using a data distribution method; and finishing the graph data partitioning. Compared with the prior art, the method provided by the invention has higher partitioning quality.

The invention discloses a multi-target collaborative salient-region detection method based on similarity propagation and relates to the field of multimedia information processing and computer vision. The method includes the following steps: with respect to a plurality of image files which are input, segmenting each image into superpixels and performing pairwise measurement on inter-superpixel similarities and establishing a superpixel similarity graph model; realizing inter-image superpixel similarity matrix bipartite matching through the superpixel similarity graph model; performing superpixel graph model similarity diffusion so as to obtain a similarity graph model; and calculating a saliency map through the similarity graph model. The method performs collaborative salience detection on a plurality of targets in a plurality of images which are input so that detection precision is improved; through a parallelization similarity propagation algorithm, the operation speed is reduced and needs in practical application are met; and experiment results show that the method obtains a more accurate detection result during a shorter calculation time.

The invention relates to an affinity propagation clustering-based integrated classifier constructing method. The method comprises the following steps: S1, preprocessing data; S2, obtaining a characteristic distance negative matrix; S3, clustering characteristics by utilizing an affinity propagation clustering algorithm; S4, constructing base classifiers; S5, repeating the step S4 till the quantity of the base classifiers is up to a preset value; S6, screening the base classifiers; and S7, integrating the base classifiers. The method disclosed by the invention has the advantages that the method can be matched with existing characteristic filtering methods and has a broader application prospect; genes are grouped by utilizing affinity propagation clustering through adopting bicor correlation coefficients as a relevance maxim and characteristic subspaces are constructed in a random selection way on this basis, so that better base classifiers with diversity can be obtained; the base classifiers are fused by utilizing a majority voting method; and therefore, by adopting the method disclosed by the invention, better classifying effect can be obtained and the classifying performance is stable at the same time.

The invention provides a novel indoor positioning method based on a fingerprint cluster. The novel indoor positioning method comprises a hierarchical indoor positioning method based on an affinity propagation cluster algorithm and a nearest neighbor algorithm which are combined together. Collected indoor fingerprint information is preclassified by effectively using the similarity of indoor fingerprints with close geographical positions, and fingerprints in a fingerprint database are classified into different subsets; furthermore, the time for positioning resolving is greatly shortened by a method for hierarchically resolving positioning results according to the fingerprint sets which are obtained by classification and have characteristics; meanwhile, the precision and the stability of an indoor positioning system are guaranteed.

The invention relates to a similarity propagation and popularity dimensionality reduction based mixed recommendation method. According to the similarity propagation and popularity dimensionality reduction based mixed recommendation method, sparse data are processed in two phases; firstly, neighbors of the sparse data are expanded due to constant iteration of similar matrixes of users, resources and Tags through a similarity propagation method and accordingly elements for zero are filled; then a score algorithm in a search engine is introduced to calculate the Tag popularity in consideration of the problem that original data is provided with meaningless rubbish Tags, the tags with the popularity smaller than a certain threshold value are deleted to simplify data to perform dimensionality reduction on the matrix; recommendation results are diversified and the sparsity and cold starting problem can be relieved to some extent due to the fact that the recommendation based on contents and the collaborative filtering recommendation are combined. The similarity propagation and popularity dimensionality reduction based mixed recommendation method has the advantages of solving the problem of data sparsity in the individual recommendation process and being high in recommendation result accuracy, high in accuracy and high in reliability.

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

The invention discloses a polarized SAR (synthetic aperture radar) image classification method based on improved affinity propagation clustering. The problem of low classification accuracy in the existing unsupervised polarized SAR classification method is mainly solved. The method comprises the implementation steps of: carrying out four component decompositions on each pixel point, extracting four scattering powers of each pixel point; dividing an image according to the obtained scattering powers to obtain four classes; equally dividing each obtained class into 20 small classes; clustering the 20 small classes in each class by the improved affinity propagation clustering to obtain the pre-classification result of the image; and finally, carrying out iterative classification on a pre-classified image by a Wishart classifier to obtain the final classification result. Compared with the classical classification method, for the method disclosed by the invention, the division on a polarized SAR image is stricter; the classification effect is better; the computation complexity is small; and the polarized SAR image classification method can be used for carrying out terrain classification and target identification on the polarized SAR image.

The invention relates to an evolutionary multi-objective optimization community detection method based on affinity propagation. The method is characterized by comprising the first step of starting the evolutionary multi-objective optimization community detection method based on affinity propagation, the second step of reading in an adjacent matrix of a network, the third step of setting correlation parameters, the fourth step of selecting a similarity measuring method based on signals, the fifth step of selecting the clustering method of affinity propagation, the sixth step of selecting a multi-objective evolutionary algorithm, the seventh step of obtaining a preliminary partition result based on the fifth step and obtaining a further partition result based on the sixth step, and the eighth step of finishing the evolutionary multi-objective optimization community detection method based on affinity propagation. According to the evolutionary multi-objective optimization community detection method based on affinity propagation, a complex network can be partitioned more precisely and rapidly.

The invention discloses an image automatic marking method. The method comprises the following three steps of: selecting some representative images from a personal photo gallery through an affinity propagation clustering algorithm to describe cluster visual information, so that the training image number can be greatly reduced; reasonably marking the representative images according to the visual information and a semantic annotation consistency principle, so that the image searching and managing performance can be improved; and automatically marking the other images through a random walk algorithm with a restarting function.

The invention relates to an online social network multisource point information tracing system and a method thereof. The method comprises the following steps: arranging a part of observation nodes in a network so as to acquire the range that messages are transmitted in the network and the moments that the messages are received, namely, mapping the timeliness and the spatiality of observation nodes that messages are received for multiple times into a network structure firstly, primarily confirming a source point range, and confirming a source point alternative set and a time delay alternative set by using a restart random walk algorithm; converting a positioning problem into a clustering problem by virtue of similarity of source points and single observation nodes in the alternative set in terms of time and space, designing a clustering algorithm based on affinity propagation learning, finding the optimal representative point set, and confirming the number and the positions of the source points. By adopting the online social network multisource point information tracing system and the method thereof, the nodes are sufficiently utilized to receive time dimension information of the messages, under the condition that the state information of all user nodes in the network is not acquired, the number and the positions of propagation source points can be relatively accurately confirmed, harmful information propagation can be effectively controlled, and social harmony and stability can be maintained.

The present invention relates to a handset sensor-based group behavior analysis method, and belongs to the technical field of computer network applications. According to the technical scheme of the invention, a handset acceleration signal and a touch screen sensor signal are adopted as a data source for behavior analysis. After that, original data are subjected to noise treatment through a sliding average filter. The features of pre-processed data are extracted in the time domain and in the frequency domain. In this way, a dimension reduction method combined with the mutual information and the linear discriminant analysis is constructed for the dimension reduction treatment of extracted features. After the dimension reduction treatment, corresponding feature vectors are input into a behavior classification model, and then a user behavior recognition result is obtained. Furthermore, the access time and the access frequency are selected as group behavior attributes. Finally, based on the affinity propagation algorithm, the data of user group behaviors are clustered, so that the group behavior analysis is completed.

The invention discloses an ensemble classification method based on randomized greedy feature selection, and belongs to the field of bioinformatics and data mining. The method is used for classifying gene expression data related to plant stress response. The method includes the following steps that 1, randomness is introduced into a traditional greedy algorithm to conduct feature selection; 2, a weighting local modular function serving as a community discovery evaluation index in a complex network is used as heuristic information of the randomized greedy algorithm; 3, base classifiers are trained in each feature subset with a support vector machine algorithm; 4, clustering partition is conducted on the base classifiers with an affinity propagation clustering algorithm; 5, base classifiers serving as class representative points in the cluster are used for conducting integration, and an ensemble classification model is formed with a simple majority voting method. By means of the method, whether plant samples are stressed or not can be recognized according to gene expression data, and the microarray data classification precision is greatly improved; besides, the algorithm is high in generalization capability and has very high stability.

The invention relates to a method for obtaining interval parameters of typical working conditions of harmonic users. The method comprises the following steps of obtaining maximum values and minimum values of a plurality of monitoring time segments of any phase in harmonic detection data of users to form an interval data sample set; performing interval principal component analysis on the interval data sample set by adopting CPCA, and selecting a sample characteristic quantity according to a variance contribution rate; calculating the similarity between data points for the sample characteristic quantity, and defining input parameters to perform affinity propagation clustering until an algorithm is subjected to iterative convergence; calculating deviation degree indexes of interval sample points of the sample characteristic quantity, and when the deviation degree indexes are smaller than a set threshold, defining the deviation degree indexes as transitional working conditions and removing the sample characteristic quantity and corresponding original data; outputting an original sample set according to a working condition division result which does not contain the transitional working conditions; and calculating typical parameters of harmonics of the working conditions. According to the method, the harmonic pollution level of an electric network after user access and running can be assessed more reasonably, thereby facilitating analysis and control on the harmonic users.

The invention discloses a stock index tracking and prediction method based on social network clustering. The method comprises the following steps: acquiring indices of a previous month and a current month and component stock data from a third-party database, and carrying out data cleaning, so that in-sample data and out-of-sample data that can be used for researches are obtained; calculating metric distances through correlation coefficients among component stocks to construct a social network among the component stocks, carrying out the network clustering through the adaptive affinity propagation clustering algorithm, extracting a clustering center of each cluster to form a stock pool, utilizing an index tracking optimization model to achieve the optimized tracking of stocks in the stock pools on a benchmark index, and determining the optimized weight of index tracking; and applying the stock pools and the optimized weight obtained through in-sample training to the index tracking for the out-of-sample data, so that a predicted index is obtained. The invention further provides a stock index tracking and prediction system. The method and system provided by the invention has the advantages that correlation among the constructed stock pools is low, tracking errors are small, replication results are highly stable, and accurate index tracking is achieved.

The invention discloses a road congestion discovery method based on a higher-order Markov model. The method includes two steps: probability transfer matrix calculation and congestion state prediction;during probability transfer matrix calculation, obtaining appropriate road section state classification by employing affinity propagation clustering, and then calculating a transfer probability matrix of the road section state in different time intervals; and during congestion state prediction, predicting the congestion state with the combination of the transfer probability matrix according to the state of road section at which a moving object is located and adjacent road sections. According to the method, the transfer probability matrix is calculated through the affinity propagation clustering algorithm in advance and then stored in a database, when a model is required for prediction, the probability matrix is directly read from the database, the transfer probability matrix is calculatedin advance and stored, the matrix can be directly inquired during prediction, the speed of real-time prediction is greatly increased, and the classification accuracy can be improved by employing affinity propagation clustering so that the correctness of the probability transfer matrix is improved.

The invention relates to a power user grouping method and system based on the DTW distance and an affinity propagation clustering algorithm, and the method comprises the steps: collecting power quality monitoring data and user active power data, and constructing a feature matrix representing a time sequence change rule; calculating the feature similarity between each user and the power quality index, and quantifying the association degree of the user features and the power quality features; on the basis of the association degree of the user characteristics and the power quality characteristics, performing power quality characteristic similarity clustering calculation on different users, and classifying the users with similar association degrees into the same class; and providing corresponding electric energy with different electric energy qualities for users with different electric energy quality characteristics. According to the invention, user characteristic classification based on the aspect of power quality is realized, and a technical support is provided for lean power quality service of users.

The invention discloses an XML (extensible markup language) document spectrum clustering method based on affinity propagation in the technical field of Web data management. The method disclosed by the invention comprises the following steps: representing an XML document by virtue of a characteristic vector formed by an XML path; then calculating the initial similarity between every two XML document vectors to obtain an initial similarity matrix W, and further determining an initial affinity relation matrix N; then correcting the similarity between every two implied similar XML document vectors by adopting an affinity propagation algorithm, so as to obtain a final similarity matrix A; and finally determining the clustering number and clustering result by applying a first specified method according to the final similarity matrix A. According to the method disclosed by the invention, the initial similarity matrix between XML documents obtained by adopting the traditional similarity calculation method is corrected by adopting the affinity propagation algorithm, and similarity between the implied similar XML documents can be reflected; and the method disclosed by the invention is independent of sequence of the XML documents and is applicable to the clustering of XML document retrieval results which are arranged in any sequence.