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

30 results about "Pattern clustering" patented technology

Content-based and time-evolving social network analysis

System and method for modeling a content-based network. The method includes finding single mode clusters from among network (sender and recipient) and content dimensions represented as a tensor data structure. The method allows for derivation of useful cross-mode clusters (interpretable patterns) that reveal key relationships among user communities and keyword concepts for presentation to users in a meaningful and intuitive way. Additionally, the derivation of useful cross-mode clusters is facilitated by constructing a reduced low-dimensional representation of the content-based network. Moreover, the invention may be enhanced for modeling and analyzing the time evolution of social communication networks and the content related to such networks. To this end, a set of non-overlapping or possibly overlapping time-based windows is constructed and the analysis performed at each successive time interval.
Owner:IBM CORP

Searching system based on encyclopedic data extracting integration

The invention relates to a searching system based on encyclopedic data extracting integration. The searching system is characterized by comprising a data extracting module, a data integrating module and a data searching module, wherein the data extracting module is used for extracting encyclopedic webpage from internet, positioning and initially filtering tables in various encyclopedic webpage, positioning and extracting the tables based on visual features, uniformly converting the extracted tables into a list mode, classifying the tables with the same feature parameters into a sort, extracting and identifying the classification information for the tables of each sort, and storing the classified information into an information database and an XML database; the data integrating module is used for classifying and marking according to sorts, adopting an integrating method to merge the tables with the same attribute into the same mode library, clustering the mode information in each mode library, and outputting mode cluster and recommended mode; and the data searching module is used for searching corresponding table information from the information database, and outputting searching results and the recommended mode.
Owner:PEKING UNIV

Mobile phone with system failure prediction using long short-term memory neural networks

Mobile phones and methods for mobile phone failure prediction include receiving respective log files from one or more mobile phone components, including at least one user application. The log files have heterogeneous formats. A likelihood of failure of one or more mobile phone components is determined based on the received log files by clustering the plurality of log files according to structural log patterns and determining feature representations of the log files based on the log clusters. A user is alerted to a potential failure if the likelihood of component failure exceeds a first threshold. An automatic system control action is performed if the likelihood of component failure exceeds a second threshold.
Owner:NEC CORP

Generalized pattern recognition for fault diagnosis in machine condition monitoring

A generalized pattern recognition is used to identify faults in machine condition monitoring. Pattern clusters are identified in operating data. A classifier is trained using the pattern clusters in addition to annotated training data. The operating data is also used to cluster the signals in the operating data into signal clusters. Monitored data samples are then classified by evaluating confidence vectors that include substitutions of signals contained in the training data by signals in the same signal clusters as the signals contained in the training data.
Owner:SIEMENS AG

Pattern clustering-based parallel network flow characteristic detection method

The invention discloses a pattern clustering-based parallel network flow characteristic detection method, which comprises the following steps: selecting a matching algorithm set of patterns, selecting length dividing points of the patterns and dividing a pattern set into short pattern subsets and long pattern subsets; determining the number of processing units for processing the short pattern subsets and the long pattern subsets; copying a number of texts to be detected and inputting each text to be detected into the corresponding processing unit of each pattern subset respectively, wherein the number of the copied texts is equal to the total number of the long and short pattern subsets; in combination with the processing results of each pattern subset, judging whether an attack pattern exists in the texts to be detected or not; and repeating the steps to continuously detect data traffic transmitted by a quick network flow to be detected. The method is an extensible total solution for network flow characteristic detection, can be applied to various levels of performance requirements and the pattern sets of various scales, and is of profound value for systems for high-performance content detection, intrusion detection, virus protection and unified threat management, network information monitoring and the like.
Owner:EASYWAY

Digest index generation method for time sequence key value type industrial process data

The invention discloses a digest index generation method for time sequence key value type industrial process data. The method comprises the steps that S1, the time sequence key value type industrial process data is acquired; S2, smooth noise preprocessing is performed on acquired time sequence data to obtain time sequence data with timestamps; S3, a symbolic aggregate approximate representation method is adopted to represent the time sequence data obtained after preprocessing; and S4, results obtained after symbolic aggregate approximate representation are subjected to mode clustering, and theresults obtained after mode clustering are made into indexes by the adoption of a prefix algorithm. The method has the advantages that based on the data preprocessing method, the symbolic aggregate approximate representation method and the prefix tree algorithm are fused to form the digest index generation method for the time sequence key value type industrial process data; and through the method, the dimension of the original time sequence data can be lowered, features of the original data are effectively extracted, and the digest index generation method is realized by the adoption of the prefix tree algorithm.
Owner:CHONGQING UNIV

A software defect repair template extraction method based on clustering analysis

The invention discloses a software defect repair template extraction method based on clustering analysis, belonging to the field of software maintenance. The steps are as follows: firstly, a fine-grained modification mode of a bug is defined, and the fine-grained modification mode related to each bug is identified; the program elements of the fine-grained modification pattern associated with eachbug are then captured; then, the top-level modified pattern multisets of each bug are obtained, and then hierarchical clustering analysis is performed to obtain multiple top-level modified pattern multisets after clustering; and then a new modified pattern multiplex corresponding to each top-level modified pattern multiplex is obtained; then a modified pattern multiplexing diagram is obtained according to the relationship between program elements; then, the modified pattern multiset graph is segmented and optimized to obtain the modified pattern clustering; finally, the software defect repairtemplate is constructed according to the modified pattern clustering. The repair template obtained by the method of the invention has semantic characteristics, has stronger universality and versatility, and improves the efficiency and precision of defect repair.
Owner:YANGZHOU UNIV

Single-station passive positioning method and device based on extended Kalman filtering

The invention discloses a single-station passive positioning method and device based on extended Kalman filtering. The method includes the following steps: separately receiving incoming wave signals through two receiving devices, performing DFT of N-point plus Hanning window on a sample sequence of 2-way incoming wave undersampling signal, performing spectrum correction and pattern clustering on aDFT result, and obtaining incoming wave frequency estimation by applying a Chinese remainder theorem; separately measuring the receiving frequency of two receiving stations at two measurement moments, and making a difference between four frequency measurement results in pairs to obtain four frequency differences of the same station at two receiving moments and two stations at the same moment; andutilizing the four frequency differences as observed quantity, constructing a Kalman filtering model by utilizing target positions of two moments as state quantity, eliminating non-linear interference by the model, and completing positioning and tracking of a target. The device includes: an analog-to-digital converter and a DSP device.
Owner:TIANJIN UNIV

Method and apparatus for frequency estimation of undersampled signals based on pattern clustering and spectrum correction

The invention discloses an under-sampled signal frequency estimation method and device based on pattern clustering and spectrum correction, The method comprises the following steps of: obtaining DFT spectrum by DFT with N points and Hanning window for L-channel undersampled signal sample sequence, performing spectrum and phase correction on the DFT spectrum based on ratio method, obtaining corrected parameter group composed of frequency, phase and amplitude, and obtaining vector composed of the corrected parameter combination; Selecting a desired peak index from a set of vectors for the firstpath independent vector of the m-th frequency component; According to the peak index of harmonic parameters, the pattern clustering is carried out, and the remainder after clustering is obtained. Frequency residue array is constructed by using residue, and it is brought into CRT model for reconstruction, and the estimated frequency value is obtained. The device comprises an analog-to-digital converter, a DSP chip, an output driver and a display module. The invention introduces a spectrum correction algorithm to improve the precision of frequency reconstruction, and uses pattern clustering to improve the robustness of the estimator to noise.
Owner:TIANJIN UNIV

Sound source positioning method and system based on pattern clustering

The invention relates to a sound source positioning method and system based on pattern clustering. The method comprises the steps: extracting effective signals through employing the Mahalanobis distance, carrying out grouping, obtaining a positioning data set of the spatial position and energy of a sound source by employing an SRP-PHAT method according to the wave propagation speed, carrying out clustering on the positioning data set by employing a DBSCAN method, determining the clustering category corresponding to the maximum average energy as a to-be-fused clustering category, calculating the sum of the average energy and the initial energy of the to-be-fused clustering category, determining a loss function value, updating the wave propagation speed when the loss function value is not smaller than a loss threshold value, repeating the above steps, and when the loss function value is smaller than the loss threshold value, fusing all spatial position coordinates in the to-be-fused clustering category by utilizing a PCA weighted fusion method, and determining the fused spatial position coordinate as sound source position. According to the method, a speed model does not need to be established in advance, and the positioning precision of sound source positioning is improved through pattern classification and fusion.
Owner:ZHONGBEI UNIV

Adaptive building day-ahead load prediction method based on transfer learning

The invention discloses a self-adaptive building day-ahead load prediction method based on transfer learning, and relates to the technical field of building and environmental protection. The method comprises the following steps: S1, data acquisition and processing: dividing an original data set into a small data set of a target building and a big data set of a basic building group, and filling missing values of all the original data sets; s2, clustering energy consumption modes; s3, source domain data screening: screening a historical daily load curve of a load target building energy consumption mode, and respectively constructing a data migration training set and a model migration training set; s4, constructing a day-ahead load prediction model; and S5, adaptive model optimization: continuously adjusting model parameters by using Bayesian optimization to realize adaptive load prediction of the target building. According to the method, load prediction of the target building is realized through data migration and model migration methods of migration learning in combination with historical data of the building group with sufficient data.
Owner:山东国地水利土地勘察设计有限公司

A traffic flow partitioning model based on similar evolution mode clustering and dynamic time zone partitioning

According to the method, a traffic flow time sequence partitioning model based on similar evolution mode clustering and dynamic time zone partitioning is provided, the dynamic time-space characteristics of traffic flow changing along with time are tried to be excavated for the first time, and the challenge of traffic flow time non-stationarity in short-time traffic flow prediction is solved. The invention specifically comprises the following steps: firstly, automatically identifying road sections with similar traffic flow evolution modes in a road network by using an affinity propagation clustering algorithm (APC); and secondly, for the intra-day evolution difference of the traffic flow, performing dynamic time zone division on the traffic flow in the similar evolution mode by using a curvature K-Means algorithm, and mining the space-time state characteristics of the road network traffic flow in a deeper level; after similar mode identification and automatic time zone division, performing modeling on traffic flows in different time zones in different modes, and quantifying the state information of the traffic flows, so that the prediction precision of the model is more accurate; and finally, verifying the validity of the provided model by using a real data set.
Owner:SICHUAN UNIV

Risk identification method and related device

The invention discloses a risk identification method, and the method comprises the steps of performing data statistics processing on acquired commodity data of a plurality of organizations according to commodity types and industry standards to obtain purchase and sale commodity matrixes of all the organizations; clustering the purchase and sale commodity matrixes of all organizations to obtain a purchase and sale mode clustering result; and determining organizations deviating from a preset proportion as risk organizations from the purchase and sale mode clustering result. According to the method, the corresponding purchasing and selling modes are determined by clustering the counted purchasing and selling commodity matrixes, and then the risk organization deviating from the purchasing andselling modes is determined on the basis of the normal purchasing and selling modes, so the accuracy of risk identification is improved. The invention further discloses a risk identification device, aserver and a computer readable storage medium, which have the above beneficial effects.
Owner:SERVYOU SOFTWARE GRP

Traffic mode clustering model training method, mode recognition method and storage medium

The invention discloses a traffic mode clustering model training method, a mode recognition method and a storage medium, and relates to the technical field of intelligent traffic. The method providedby the embodiment of the invention comprises the following steps: adding a clustering layer to an embedded layer of a convolutional auto-encoder network to obtain a composite clustering model; and training the label-free traffic track data by adopting a composite clustering model to obtain a traffic mode clustering model. According to the method, data mining and algorithm design are carried out onthe label-free data, so that the label data requirement can be effectively relieved, and the application cost is saved; by adding the clustering layer to the embedded layer of the convolutional auto-encoder network, better clustering performance can be obtained; and the traffic mode clustering model is obtained by training the composite clustering model through the label-free traffic trajectory data, and the traffic mode can be recognized under the label-free unsupervised learning condition.
Owner:SOUTH UNIVERSITY OF SCIENCE AND TECHNOLOGY OF CHINA

Data consistency evaluation method based on data pattern clustering

The invention discloses a data consistency evaluation method based on data mode clustering. The method is applied to the big data analysis processing field, and solve the problem that in the prior art, consistency evaluation of multi-mode coexistence fields cannot be achieved. The method comprises the following steps: firstly, carrying out pattern clustering on to-be-evaluated fields read from a database according to a determined pattern clustering algorithm, then, determining a standard pattern in clustered patterns, and finally, carrying out pattern matching on values of the to-be-evaluatedfields by adopting the standard pattern to obtain dirty data. The method is especially suitable for application scenarios in which data engineers do not know about services and which modes exist reasonably are difficult to determine.
Owner:UESTC COMSYS INFORMATION

Crystal powder diffraction pattern clustering method and device and storage medium

The invention relates to a crystal powder diffraction pattern clustering method and device and a storage medium. The method comprises the following steps: acquiring a crystal XRPD map to be clustered; obtaining a clustering parameter; removing background lines in the crystal XRPD atlas to be clustered to acquire a crystal XRPD atlas with the background lines removed; calculating a similarity matrix of the crystal XRPD maps of which the background lines are removed, wherein elements of the similarity matrix comprise similarity values of every two crystal XRPD maps in the crystal XRPD maps of which the background lines are removed; and according to the similarity matrix and the clustering parameter, carrying out clustering on the crystal XRPD atlas of which the background lines are removed to obtain a clustering result. According to the technical scheme, the accuracy of diffraction pattern clustering can be improved.
Owner:SHENZHEN JINGTAI TECH CO LTD

Method and device for clustering and extracting entity-relationship patterns

The invention provides a method and device for clustering and extracting entity relationship modes. The method for clustering the entity relationship modes comprises the steps that original sentences are preprocessed so that entity words expressing entities in the original sentences are identified; the entity relationship among the entity words in the preprocessed sentences is determined according to the entity words, relational words in the relational word bodies, and special occurring sequences of the entity words and the relational words in the preprocessed sentences, and the preprocessed sentences are split into secondary sentences according to the determined entity relationship; the entity relationship modes of the split secondary sentences are extracted, wherein the entity relationship modes of the secondary sentences are expressed through relationship tuples composed of the entity words and contexts among the entity words; a first similarity among the entity relationship modes of the extracted secondary sentences is calculated; the entity relationship modes of the secondary sentences are clustered into entity relationship mode types according to the calculated first similarity among the entity relationship modes of the secondary sentences.
Owner:FUJITSU LTD

Method, system and device for online classification and evaluation of slurry quality in desulfurization system

The invention discloses a method, system and device for online classification and evaluation of slurry quality in a desulfurization system. The method includes: collecting historical operation data of relevant parameters of the desulfurization system and corresponding slurry quality evaluation labels, obtaining an original data sample D, and performing data analysis. Cleaning; Steady-state screening with steady-state judgment conditions and standardized preprocessing; Dimensionality reduction processing for dimensionally standardized samples BD using local preservation projection LPP algorithm, and agglomerative k-means clustering method for dimensionality-reduced samples JD conducts pattern clustering and recognition, analyzes the clustering results, and obtains the slurry quality classification and evaluation library N; obtains a new sample S of relevant parameters of the desulfurization system, adds it to the steady-state operation data sample set SD for iterative calculation, and obtains a dimensionally standardized sample BDS and the sample JDS after dimensionality reduction, and pattern clustering of the sample JDS, compared with the typical sample labels of N in the classification evaluation library, to obtain the evaluation category of the new sample S.
Owner:DATANG ENVIRONMENT IND GRP +1

Online classification evaluation method, system and device for slurry quality of desulfurization system

The invention discloses an online classification evaluation method, system and device for the slurry quality of a desulfurization system. The method comprises the steps: collecting historical operation data of related parameters of a desulfurization system and a corresponding slurry quality evaluation label, obtaining an original data sample D, and carrying out the data cleaning; carrying out steady-state screening according to steady-state judgment conditions, and carrying out standardization pretreatment; carrying out dimension reduction processing on the dimension-standardized sample BD by adopting a local preserving projection (LPP) algorithm, carrying out pattern clustering and identification on the dimension-reduced sample JD by adopting a cohesion k-means clustering method, and analyzing a clustering result to obtain a slurry quality classification evaluation library N; and obtaining a new sample S of related parameters of the desulfurization system, adding the new sample S into the steady-state operation data sample set SD for iterative calculation to obtain a dimension standardized sample BDS and a dimension-reduced sample JDS, performing mode clustering on the sample JDS, and comparing the sample JDS with a typical sample label of N in the classification evaluation library to obtain an evaluation category of the new sample S.
Owner:DATANG ENVIRONMENT IND GRP +1

Mode clustering method of battery alarm characteristic data and accident characteristic identification technology

The invention relates to the technical field of batteries, in particular to a mode clustering method of battery alarm characteristic data and an accident characteristic recognition technology. The method comprises the following steps: S1, collecting operation data of a battery before and after alarm in the operation of an accident vehicle and a normal vehicle; S2, performing dimension reduction processing on the operation data to obtain mode features; S3, performing clustering analysis on the mode features after dimension reduction to obtain classification features of the operation data; S4, analyzing the statistical difference between the accident vehicle and the normal vehicle according to the classification features; S5, judging whether the vehicle is an accident vehicle or not by taking the statistical difference as a standard. The method has the advantages that compared with the prior art, the judgment standard in the scheme is not single and fuzzy, the mode characteristics, the classification characteristics and the statistical difference are obtained in sequence by analyzing the operation data of the battery, the accident vehicle can be accurately identified, and the technical problem that the accident vehicle is difficult to accurately identify in the prior art is solved.
Owner:CHINA AUTOMOTIVE ENG RES INST

A parallel network flow feature detection method and system based on pattern clustering

The invention discloses a parallel network flow feature detection method based on pattern clustering, comprising the steps of: selecting a pattern matching algorithm set, selecting a pattern length segmentation point, and dividing the pattern set into a short pattern set and a long pattern set; Divide the long pattern set into multiple long pattern subsets, determine the number of processing units of the short pattern set and the long pattern subset; copy the number of copies of the text to be detected as the total number of copies of the long and short pattern subsets, and divide each copy to be tested The text is input into the processing unit corresponding to each pattern subset; combined with the processing results of each pattern subset, it is judged whether there is an attack pattern in the text to be detected; the above steps are repeated to continuously detect the data stream transmitted by the fast network stream to be detected. This method can be applied to various levels of performance requirements and pattern sets of various scales, and has extremely far-reaching effects on systems such as high-performance content detection, intrusion detection, virus protection, unified threat management, and network information monitoring.
Owner:EASYWAY

Method and device for extracting POI (Point of Interest) data from webpages

The invention discloses a method and a device for extracting POI (Point of Interest) data from webpages, and relates to the technical field of information point extraction. The method comprises the following steps: acquiring a plurality of webpages comprising POI data; performing address mode clustering on the webpages according to the URL (Uniform Resource Locator) address of each webpage; sequencing a plurality of address modes based on the quantity of webpages corresponding to each address mode to obtain the sequencing result of each address mode; selecting N address modes with largest webpage quantities; extracting POI data comprised in the webpages corresponding to the N address modes respectively. Through the method and the device, mass webpages of POI data can be determined from hundred-billion-scale webpages more rapidly, and the POI data are extracted from the webpages more accurately.
Owner:BEIJING QIHOO TECH CO LTD

A Fingerprint Quality Evaluation Method Based on Visual Cognition Machine Learning of Line Quality Experts

ActiveCN109003259BDynamic big data analysisReal-time big data analysisImage enhancementImage analysisImaging qualityNetwork model
The invention relates to a fingerprint quality evaluation method based on ridge quality expert visual cognition machine learning. Including: expert cognition and quality marking of the image quality level of stamping lines in the "on-site reconstruction area of ​​fingerprint line leftover positions", and "expert individual quality evaluation stability analysis" and "expert quality evaluation" for quality marking data. Pattern cluster analysis" and get the priority of each expert's quality marking data; cut the expert's quality marking data into pieces, and use them for image quality evaluation neural network model training according to the priority. Construct and train the neural network model until it evaluates the quality of local blocks and reaches the set accuracy threshold. Using the local block quality evaluation data made by the neural network model, the global comprehensive quality evaluation of the imprinted fingerprint image is calculated. The invention takes into account the bicuspid requirements of "multi-genre fingerprint comparison algorithm" and "expert fingerprint identification" on fingerprint quality, and is widely applicable to image quality evaluation of heterogeneous fingerprints of various specifications.
Owner:张威 +1

A software defect repair template extraction method based on cluster analysis

The invention discloses a method for extracting software defect repair templates based on cluster analysis, which belongs to the field of software maintenance. The steps are as follows: firstly define the fine-grained modification mode of the bug, and identify the fine-grained modification mode related to each bug; The program elements of fine-grained modification patterns related to each bug are captured; then the top-level modification pattern multiset of each bug is obtained, and then hierarchical clustering analysis is performed to obtain multiple top-level modification pattern multisets after clustering; after that, each A new modification mode multiset corresponding to a top-level modification mode multiset; then obtain the modification mode multiset map according to the relationship between program elements; then segment and optimize the modification mode multiset map to obtain modification mode clustering; finally according to the modification Pattern clustering constructs templates for software defect repair. The repair template obtained by the method of the invention has semantic features, and has stronger universality and versatility, thereby improving the efficiency and precision of defect repair.
Owner:YANGZHOU UNIV

Electronic component near-field scanning electromagnetic pattern clustering analysis method and system

PendingCN114861782AConvenient electromagnetic radiation analysisElectromagnetic radiation analysis process intelligenceCharacter and pattern recognitionComputational physicsElectronic component
The invention relates to a near-field scanning electromagnetic pattern clustering analysis method for an electronic component, and the method comprises the steps: carrying out the near-field scanning of a tested device, carrying out the post-processing of the data obtained through scanning, obtaining an electromagnetic image, and carrying out the clustering analysis of the obtained electromagnetic pattern through an improved K-Means clustering method. According to the improved K-Means clustering method, firstly, a K value is determined through an elbow method, and then clustering analysis is carried out on the obtained electromagnetic pattern through the K-Means clustering method. The invention further relates to an electronic component near-field scanning system for implementing the method. According to the method, the improved K-Means algorithm is adopted, an initial clustering value K does not need to be set in advance, and electromagnetic pattern clustering can be carried out more intelligently.
Owner:SOUTH CHINA NORMAL UNIVERSITY

Space division multiple access transmission method of statistical characteristic mode

A Space Division Multiple Access transmission method based on a statistical characteristic pattern including configuring multiple receiving and transmitting antennae at a base station side into one or more antenna arrays; comparing a diagonalization performance among the unitary matrixes for a correlation matrix of subscriber channels, with a long-time statistical correlation matrix of the subscriber channels at the base station side; calculating an optimal diagonalized correlation matrix, and then performing characteristic pattern clustering for the subscribers in a cell with the space resource divisions of the subscribers; grouping the subscribers that belong to a same characteristic pattern cluster into space division subscriber groups; carrying out SDMA transmission, by the subscribers in a same space division subscriber group, with their space resource divisions; dynamically performing subscriber characteristic pattern clustering and space division subscriber grouping described above to generate updated characteristic pattern clusters for the subscribers and updated space division subscriber groups.
Owner:HUAWEI TECH CO LTD
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