51results about How to "Implement clustering" patented technology

The invention discloses a deep learning-based fraud transaction identification method, system and a storage medium. The method comprises the following steps: acquiring a training sample, wherein the training sample is transaction data for establishing a fraud transaction detection model; constructing a stacked RBM neural network structure; training the stacked RBM neural network structure based onthe training sample, and generating a dimension reducer; performing dimension reduction on the training sample via the dimension reducer, and clustering the binary state vector obtained by dimensionreduction so as to establish a fraud transaction detection model; obtaining transaction data to be detected, and analyzing the transaction data to be detected according to the fraud transaction detection model so as to identify fraudulent transactions. The deep learning-based fraud transaction identification method, system and the storage medium in the invention can improve the accuracy of fraudulent transaction identification, and does not need to define a similarity measurement method in advance, thereby reducing difficulty and cost, and the high tolerance to sample data is achieved.

The invention discloses a privacy protection method in multi-sensitive-attribute data release, and solves the problem of poor quality of quasi-identifier data in multi-sensitive-attribute data release. The basic thinking of the invention is as follows that: firstly, clustering is executed on data sets, the data sets of which quasi-identifiers are similar are aggregated into one aggregate, and a plurality of data aggregates are generated; secondly, a multi-dimension bucket structure is constructed on the basis of sensitive attributes, and data records are mapped into the multi-dimension bucket structure according to values of the sensitive attributes; and then on the basis of multi-dimension buckets, grouping is carried out, i.e., main sensitive attributes are selected, dimension capacity of the main sensitive attributes is calculated, L (L is greater than or equal to 2) main sensitive attributes with the maximum dimension capacity are selected, one data record is respectively selected from the L main sensitive attributes, whether the data records meet the multi-sensitive-attribute L-diversity is judged, and if not, each bucket is sequentially traversed according to the capacity from big to small until the data records meet the multi-sensitive-attribute L-diversity. The process is repeated until the data in the buckets do not meet the multi-sensitive-attribute L-diversity. Finally, all groups are subjected to anonymization processing.

The invention discloses a large-scale image online clustering system and method based on comparison learning. The system comprises an augmentation subsystem, a feature extraction subsystem, an instance level comparison head subsystem and a category level comparison head subsystem. The method comprises the following steps: S1, performing augmentation operation on an original image sample set to obtain two groups of augmentation image sets; s2, constructing a total loss function, taking two groups of augmented image sets as a training set, and training a large-scale online clustering system by adopting a gradient descent optimization method; s3, performing clustering processing on a to-be-processed image sample set by adopting the trained large-scale online clustering system, and taking a category corresponding to the maximum probability output by the category level comparison head subsystem as a clustering result of each image sample; the invention solves the problems that an existing method cannot achieve large-scale online clustering, the two stages of feature extraction and data clustering are not closely related, and error accumulation is likely to occur.

The invention discloses a local scale parameter, entropy and cosine similarity-based spectral feature selection method. A Gaussian kernel function is adopted as a similarity measurement method, defined local scale parameters which are of features and based on local standard deviation of the features are used as kernel function parameters, and problems that uniform scale parameters in calculating feature affinity matrices cannot reflect data distribution information, and local scale parameters are impacted by off-group points are overcome; entropy and cosine similarity sorting is respectively adopted to measure feature importance degrees, a suitable feature subset can be quickly selected; and technical support is provided for data analysis of diseases such as tumours, and the method has important biomedical significance.

The invention discloses an image area clustering method, an image area clustering device, an outline searching method and an outline searching device. After a difference value point image is obtained by comparing the current input image with a background image, stacking predetermined grid models on the difference value point image; calculating the energy characteristics of each edge of grids in the grid models which are stacked on the difference value point image; then, determining the edges with the larger energy characteristics as candidate edges; remaining the partial candidate edges whichcan form the closed outline; deleting the rest candidate edges; obtaining the outline of a specific area formed by pixel points with larger background difference in the current input image, correspondingly, obtaining the outline of the specific area; and clustering the pixel points of the specific area at the same time; and moreover, because the clustering mode is realized by the closed outline, the dispersed noise points can not be detected as points in the specific area by mistake.

The invention relates to an image processing method and device, electronic equipment and a storage medium, and the method comprises the steps: carrying out the feature extraction of a to-be-processedimage, and obtaining a first feature of the to-be-processed image; according to the first feature and class center features of a plurality of reference image classes in a feature library, determiningan image class of the to-be-processed image; and under the condition that the image category of the to-be-processed image is a first category in the plurality of reference image categories, updating acategory center feature of the first category according to the first feature and a plurality of pieces of feature information of the first category in the feature library. According to the embodimentof the invention, the image retrieval speed and accuracy can be improved.

The invention discloses a fishing web page clustering method and device. The method comprises the steps of receiving any fishing website; acquiring the domain name of the fishing website; acquiring the domain name type corresponding to the domain name from a preset domain name list; according to the domain name type, realizing fishing web page clustering. The fishing web page clustering method and device can realize fishing web page clustering after acquiring the domain name type corresponding to fishing websites, so that two defects generated by the clustering method in the prior art when a fishing criminal uses a secondary domain name of a second-level domain are overcome. Consequently, the false alarm rate and the missing reporting rate of fishing web pages are reduced, the detection ratio of the fishing webpages is improved, and the broadcast of fishing web pages is completely stopped from the source.

The embodiment of the invention provides a public opinion hotspot tracking and prediction method based on community division and keywords information entropy calculation. The method includes the steps of constructing a similar relation network according to users concerned public opinion hotspots, analyzing the network characteristics, conducting community division for the similar relation network, analyzing the node characteristics in a community, obtaining the Hub nodes of each community, focusing and obtaining Hub node network homepage on a real-time basis, and conducting word information entropy calculating and analyzing for the obtained real-time text data, to realize tracking and prediction of public opinion hotspots. The embodiment of the invention further provides a public opinion hotspot tracking and prediction apparatus. Mass redundant social network data processing can be reduced, accurate public opinion hotspot tracking and timely prediction of subsequent hot spots can be realized to provide support for real-time public opinion early warning and deciding.

The invention provides a batch processing method for identifying defects of an X-ray directional instrument, belonging to the field of single crystal material processing. A method for detect defects in off-line batch single crystal is proposed, the eigenvector of the pendulum curve of single crystal is designed, so that the characteristics of the curve are abstracted, combined particle swarm optimization algorithm, and introduces the effective radius based on the density function, the traditional FCM algorithm is improved, the robustness of the lifting algorithm to the initialization of clustering centers is improved, falling into local optima is avoided, and interfering data is well filtered out, so as to realize clustering of batch data, only the clustering center eigenvector needs to bedetected. The defect types of the curves to be tested can be obtained according to the membership degree relation, wherein, the improved fuzzy transfer closure clustering algorithm proposed by the invention is included, the fuzzy similarity matrix is defined, the accuracy of the similarity calculation is ensured, and the invention provides a new idea and a realization mode for the crystal detection technology.

The embodiment of the invention provides a live broadcast room recommendation method and a related device, which are used for quickly recommending a live broadcast room to a user. The method comprisesthe steps: obtaining a user watching sequence of a live broadcast platform in a preset time period; determining a transfer weight between live broadcast rooms in the live broadcast platform accordingto the user watching sequence; determining a first directed graph of the live broadcast platform according to the transfer weight between the live broadcast rooms in the live broadcast platform; cutting the first directed graph based on a preset out-degree and/or a preset in-degree to obtain a second directed graph; calculating the transition probability of each node in the second directed graph;calculating the target arrival probability of each node in the second directed graph according to the transition probability of each node in the second directed graph; determining live broadcast roomclustering of the live broadcast platform in a preset time period based on the target arrival probability of each node in the second directed graph; and recommending a live broadcast room to the target user according to a set of live broadcast rooms watched by the target user in the preset time period and the live broadcast room clustering of the live broadcast platform.

The invention discloses an environment sensing system based on a laser radar. The environment sensing system comprises the laser radar, an automatic driving automobile calibration module, a ground point cloud extraction module, a point cloud data down-sampling module and a laser radar data point cloud clustering module. According to the environment sensing system based on the laser radar, the calibration between the laser radar data and the automatic driving automobile can be completed; a stable rotation translation matrix can be used, the algorithm difficulty is low, the development and maintenance cost is low, laser radar data down-sampling can be achieved, the algorithm has a very long history by using the stable algorithm, the algorithm reliability is fully proved by various experiments, and the development and maintenance cost is low.

The invention relates to a multi-view clustering machine learning method with missing kernel. The method fuses filling and clustering, fills missing kernel under the guidance of clustering, and clusters with the filled kernel. The method comprises the following steps: 1) obtaining a target data sample, mapping the target data sample to a multi-kernel space; 2) establishing a missing multi-core k-means clustering optimization objective function; 3) adopting a three-step alternating method to solve that missing multi-core k-means clustering optimization objective function to realize clustering.Compared with the prior art, the invention considers the joint optimization of filling and clustering, and has the advantages of good clustering effect and the like.

According to the method for identifying the three-dimensional fluorescence spectrogram of the soluble organic matter, database construction and comparison are carried out at the same time during sample comparison of a fluorescence spectrum database of the soluble organic matter, extended functions such as updating are automatically perfected, and the identification of a tested sample is improved.Necessary data processing is carried out on fluorescence spectrum data of a detected sample, similarity calculation and matching can be carried out on the fluorescence spectrum data and samples in a reference comparison database, and therefore rapid and accurate recognition and judgment are obtained. And the method further includes identifying peak position coordinates and peak intensity information of the sample, establishing a probabilistic neural network, and realizing clustering classification and discrimination of the sample in combination with a Bayes theory; calculating cosine coefficients according to the fluorescence data matrix of the tested sample to obtain a maximum matching degree and a comprehensive similarity coefficient index, and forming spectrogram identification information; performing Cosine similarity coefficient calculation on fluorescence spectrum data obtained through parallel factor analysis and fluorescence spectrum data of all components, and therefore further classification and information identification of samples are achieved.

Embodiments of the invention provide an information processing method and device. The method comprises the following steps of: obtaining first information of each object; obtaining a plurality of numerical values according to the first information, and generating first vectors of the objects according to the plurality of numerical values, wherein each numerical value is used for describing the first information from different dimensionalities; obtaining distances between the first vectors in pairs; and clustering the first vectors according to the distances so as to complete information processing. The important step of data mining is data clustering; according to the method, the first information (such as activity description information) of each object (such as a merchant) is firstly obtained; generating a first vector for each object according to numerical values under various dimensionalities in the first information of each object, and the first information is clustered on the basis so as to realize the clustering of the objects, so that the data mining efficiency on network information platform can be greatly improved and the information processing time is saved.

The invention relates to an adaptive multi-mean two-step clustering method, and the method comprises the steps: carrying out the preliminary clustering of input data through employing a multi-mean clustering algorithm based on a chaotic quantum particle swarm in the first step, and carrying out the further clustering on the basis of a clustering result in the first step through employing an adaptive hierarchical clustering algorithm in the second step, and obtaining a final clustering result; the method can be used for cluster distribution data clustering, can also be used for non-cluster data clustering, and has the advantages of being high in operation speed, low in complexity, wide in application range and small in influence of abnormal values. The method can be used as a basic technology for data processing, and can be used for data processing work in the fields of system modeling, pattern recognition, machine learning, data mining and the like.

The invention discloses an event processing method and apparatus. The method includes detecting error events; generating error description information corresponding to the error events when the error events are detected; and uploading the error events including the error description information to a server to enable the server to cluster the error events based on the error description information. Therefore, the server can rapidly position the error events based on the error description information, an effective solution can be provided, and user experience is improved.

The invention discloses a cooperative monitoring system, device and method. According to the embodiment of the invention, one of at least two monitoring devices is configured as a host, and the communication connection between the host and rear-end equipment is built; other monitoring devices in the at least two monitoring devices are configured as slaves, and the slaves establish communication connection with the host to form topological connection cascaded to the back-end equipment through the host; under the topological network structure, a picture library is configured in the host, monitoring images in the monitoring data of the host and/or the slaves are intelligently analyzed based on the picture library, and the host transmits an intelligent analysis result to the back-end equipment; and the host also updates the picture library by utilizing the monitoring images in the monitoring data. Therefore, the embodiment of the invention realizes clustering of monitoring information of other monitoring devices by taking one monitoring device as a host.

The invention discloses an unsupervised pile collection method in a complex system, which comprises the steps that: discrete characteristic variables and class variables are determined according to the original information of a complex system sample; the relevancy between two characteristic variables is calculated; 'the crowd of friends and relatives' of each characteristic variable is determined; the unsupervised pile collection is carried out to the characteristic variables according to the self-organization of the pile collection to obtain the combination of the characteristic variables; each pile is back substituted into original data to obtain the sensitivity; the degree of the sensitivity is judged; verification is carried out to the unsupervised pile collection method by utilizing the class variables of the system to obtain the optimum combination of the characteristic variables. The method solves the problem that the traditional relevancy can not distinguish positive correlation and negative correlation, has self-organization, needs no human intervention, has high running speed, and is suitable for a large amount of data, even mass data. Furthermore, the method can realize clustering and the appearance of certain variables in certain different classes, can carry out the verification to the unsupervised pile collection so as to find out the optimal pile, and has wide application value in the fields such as ecological differentiation and clinical medical data analysis, etc.

The embodiment of the invention provides an information processing method and device. The method comprises the following steps that: obtaining various categories and the classification rule of first information, wherein each category of classification corresponds to one code; according to the classification rule, carrying out classification on the first information of each object; according to the code corresponding to the category to which the object relates, generating the total code of the object; and clustering all total codes to finish information processing. An important step of data mining is data clustering; in order to cluster objects on a network information platform, the embodiment of the invention carries out processing to obtain a plurality of codes of each object, and the plurality of codes of each object are integrated so as to obtain the total code of each object. Therefore, clustering is implemented for numerous total codes so as to realize object clustering to greatly improve data mining efficiency on the network information platform and save information processing time.

The invention provides a text processing method and device, equipment and a storage medium, and relates to the field of big data and text processing. According to the specific implementation scheme, a to-be-processed text is compared with all processed texts in a classification result table, at least one processed text similar to the to-be-processed text is determined according to a comparison result, and the determined processed text serves as a similar text of the to-be-processed text; wherein the classification result table comprises a plurality of processed texts and classification centers corresponding to the processed texts; and a classification result of the to-be-processed text is determined according to the relevancy between the to-be-processed text and each similar text. According to the technical scheme, clustering processing of the to-be-processed text can be achieved, the calculation amount in the text processing process is reduced, and the text processing efficiency is improved.

The invention provides an environment monitoring system with a good monitoring effect. The system comprises an environment monitoring terminal, storage equipment, processing equipment and cloud equipment, wherein the environment monitoring terminal is used for monitoring environment information in a predetermined area, the storage equipment is used for storing monitoring data, the processing equipment is used for clustering the stored data and acquiring a data clustering result, and the cloud equipment is used for storing the data clustering result in the cloud. The invention has the beneficial effects that the environment monitoring system is provided, collection and clustering of data and assessment and cloud storage of the clustering result are achieved, the assessment result is fed back to the clustering unit, and thus the clustering unit can be improved.