129 results about "Graph analytics" patented technology

Systems and methods for processing an image to determine whether segments of the image belong to an object class are disclosed. In one embodiment, the method comprises receiving digitized data representing an image, the image data comprising a plurality of pixels, segmenting the pixel data into segments at a plurality of scale levels, determining feature vectors of the segments at the plurality of scale levels, the feature vectors comprising one or more measures of visual perception of the segments, determining one or more similarities, each similarity determined by comparing two or more feature vectors, determining, for each of a first subset of the segments, a first measure of probability that the segments is a member of an object class, determining probability factors based on the determined first measures of probability and similarity factors based on the determined similarities, and performing factor graph analysis to determine a second measure of probability for each of a second subset of the segments based on the probability factors and similarity factors.

Embodiments disclosed herein provide systems and methods for a unified analytics environment. In a particular embodiment, a method provides, handling a plurality of relational functions within a relational analytics environment. The method further provides, while handling the plurality of relational functions, encountering at least one graph function that comprises a query intended for a graph analytics environment. The method further provides, in response to encountering the at least one graph function, communicating with the graph analytics environment to handle the at least one graph function.

A method and a system for analyzing dynamic graphs are disclosed. In accordance with such method and system, computations are performed at a plurality of graph vertices every time a change in the graph occurs. In order to minimize the computational load of each computation iteration, previous computation results are reused when the inputs for a computation at a given vertex are unchanged from previous computations. This approach enables real-time data mining from large dynamic graphs, without requiring users to devise their own incremental graph algorithms.

Methods, systems, apparatus, and non-transitory computer readable media are described for identifying users who are likely to have unauthorized access to secure data files in an organizational network. Various aspects may include presenting the identified users on a display for a system administrator and/or security analyst to resolve. For example, the display may include a graph data structure with users represented as nodes and connections between users represented as edges. Each connection may be a pair of users belonging to a same security group. The graph data structure display may be organized and color coded in such a manner, that a system administrator and/or security analyst may quickly and easily view the users who are most likely to have unauthorized access to secure data files. The authorized access may then be remedied or taken away.

A query on a graph database can be efficiently performed employing a combination of an abstraction program and a graph analytics appliance. The abstraction program is generated from a query request employing an abstraction program compiler residing on a computational node, and includes programming instructions for performing parallel operations on graph data. The graph analytics appliance receives or generates the abstraction program, and runs the abstraction program on data fetched from a graph database to generate filtered data that is less than the fetched data. The filtered data is returned to the computational node. The bandwidth between the graph database and the graph analytic engine can be greater than the bandwidth between the computational node and the graph analytic engine in order to utilize processing capacity of the graph analytics appliance.

One embodiment of the present invention provides a system for generating a product recommendation. During operation, the system receives graph data indicating vertices and edges of the graph. The vertices represent customers and products and the edges represent purchases. The system then receives a query of the graph to determine a product recommendation. Next, the system generates a finite-state machine (FSM) based on the query, executes the query, and determines whether a current state of the FSM is a traversal state. In response to the current state being a traversal state, the system generates a traversal FSM. The system then searches the traversal FSM for a nearest future traversal state, generates a bitmask for the future traversal state, and utilizes the generated bitmask when executing the future traversal state to generate the product recommendation.

Systems, software, and methods are described herein dynamically generate traversal code for a graph analytics environment. In one example, a method of dynamically generating traversal code for a graph analytics environment includes, in one or more processing systems, identifying a query expression for a relational analytics environment. The method further includes generating the traversal code for the graph analytics environment based on the query expression, and executing the traversal code in the graph analytics environment on a graphical database. The method also provides communicating results of the traversal code to the relational analytics environment.

The embodiment of the invention provides an anti-fraud detection method and system based on graph analysis, and the method comprises the steps: 1), building a relation graph corresponding to a user incommunication data based on the communication data which is provided by a communication operator and comprises victim communication data; 2) obtaining victim user characteristics from the relation graph, training a random forest classification model by utilizing the victim user characteristics, and then identifying whether a user to be identified is a victim result by utilizing the trained randomforest classification model; 3) shrinking other nodes except the number nodes in the relation graph into corresponding edges or nodes to obtain a target relation graph; and 4) obtaining the number node having the largest influence on other number nodes in each sub-connected graph in the target relation graph, and taking the user corresponding to the number node having the largest influence on other number nodes as a fraud criminal. By applying the embodiment of the invention, the fraud criminal can be identified.

An apparatus and method are described for accelerating graph analytics. For example, one embodiment of a processor comprises: an instruction fetch unit to fetch program code including set intersection and set union operations; a graph accelerator unit (GAU) to execute at least a first portion of the program code related to the set intersection and set union operations and generate results; and an execution unit to execute at least a second portion of the program code using the results provided from the GAU.

One embodiment of the present invention provides a system for generating a product recommendation by translating transaction data to graph representation for input to a graph analytics application. During operation, the system generates a transaction table to store transaction data, a customer table to store customer data, and a product table to store products data. The system generates a table containing topology and edge identifier information and a table containing edge attribute information. Next, the system generates headers that include data describing the customer table and/or the product table and/or the table containing edge attribute information. The system then generates files containing the one or more headers and data from the tables, in which the data describes a graph with edges representing transactions and vertices representing customers or products. Subsequently, the system submits the one or more files as input to the graph analytics application to generate a product recommendation.

The invention relates to the technical field of process control, in particular to a design and implementation method of a PSC7-based continuous process control strategy with a recovery process. Aimingat the defects that a traditional continuous process system wastes materials and is low in yield, multiple angles of safety, production optimization, energy conservation and the like are comprehensively considered, system characteristics are analyzed by using an SDG graph, an improved cascade-ratio control strategy with a recovery process and control ideas of eight parts of the whole system are formulated, and a complete set of control schemes of hardware configuration, CFC, parallel SFC and the like are completely designed based on PCS7. An advanced multifunctional process control experimental device SMPT-1000 is adopted for simulation of the designed strategy, the operation result shows that the design and implementation method can achieve full-automatic stable operation, the material recycling rate and the product yield are remarkably improved, and safety, practicability and robustness are good. The configuration based on WinCC design can monitor important indexes such as an operating state, concentration, final yield and recovery process flow of the system in real time, and the important indexes can be visually displayed.

The embodiment of the invention provides a method for monitoring application performance, related equipment and a storage medium. The method is applied to an application resource management system. the method comprises: deploying probes for monitoring points in a data layer, and providing a probe deployment strategy based on a rough set theory; setting an acquisition strategy of the probe based onthe probe deployment strategy, receiving an acquisition request in a data layer, and acquiring multiple performance data of the application system performance through the probe according to the acquisition request; performing business logic processing on the multiple performance data in the service layer to obtain multiple performance data after business processing, and inputting the multiple performance data into the application layer; in an application layer, generating a plurality of performance indexes according to different scenes and different business requirements of a user and the plurality of performance data processed by the service layer, generating a plurality of views according to the plurality of performance indexes, displaying the plurality of views, and setting applicationsetting function icons on the plurality of views; and analyzing the performance of the application resource management system on an application layer based on multiple views. According to the scheme,the information system operation efficiency and the monitoring quality can be improved.

The invention relates to the technical field of cloud services and data attack analysis, in particular to a data attack analysis method applied to cloud services and a server. Due to the fact that in the attack intention analysis process, service event positioning processing can analyze a random and changeable service environment, according to at least one first sensitive interaction event, the first target service session set in the cloud service log triggering the data attack analysis condition is determined, and the first multi-modal interaction event corresponding to the first target service session set is determined, that is, the intention analysis result in the attack intention analysis process is obtained, according to the first multi-modal interaction event, attack intention positioning processing is started, so that a first target data attack intention corresponding to the first target service session set in the service scene can be obtained, so that the intention analysis of the service session sets in different states in the cloud service log triggering the data attack analysis condition can be efficiently and accurately realized.

The invention relates to the technical field of intelligent decision making, and discloses an intelligent question answering method based on natural language processing, which comprises the following steps: acquiring a user question, and extracting semantic information of the user question; analyzing the semantic information through a pre-trained intention analysis algorithm, obtaining intention information corresponding to the user question, wherein training data of the intention analysis algorithm comprises feedback information, obtained in the training process, of the user for a training result; utilizing at least two modes to obtain at least two query results corresponding to the user question, wherein the at least two modes comprise a standard knowledge base and a knowledge graph corresponding to the intention information; and determining a standard question answer of the user question according to the at least two query results. The invention further relates to a block chain technology, and the intention analysis algorithm can be stored in a block chain node. The invention further provides an intelligent question answering device and equipment based on natural language processing and a computer readable storage medium. The question answering accuracy in the question answering process can be improved.

The invention relates to the technical field of software testing, in particular to a stability testing method and system based on control arrangement. The method comprises the steps of obtaining an interactive control of an initial page; analyzing the function of each obtaining control to obtain a skipping control; skipping to other pages through the skipping control, and arranging all obtained interactive controls into a tree structure; after a tree structure is obtained, analyzing the tree diagram by proper traversal algorithms; comparing traversal analysis results to select a traversal algorithm with short traversal time; and performing a traversal test by using the selected traversal algorithm. The method and the system have the advantages of comprehensive test, good integrity, easiness in jumping to the next page in traversal work and high working efficiency, the traversal algorithm is selectively executed, the test workload and the test time are reduced, the test efficiency is improved, and the test cost is reduced. The technical defects that an existing stability test is incomplete in test, not prone to page turning and large in test workload are overcome, and the software test requirement is met.

The invention belongs to the technical field of new energy wind power generation power prediction, and relates to an ultra-short-term wind power prediction method based on an adaptive deep residual network, which comprises the following steps: S1, collecting historical data of a wind power plant; s2, screening data by adopting a Pearson correlation coefficient method, and processing abnormal databy adopting a box type diagram analysis method; s3, normalizing the variable data; s4, establishing a deep residual error network wind power prediction model; s5, optimizing the prediction model by adopting an adaptive AdaDelta optimization algorithm to obtain an optimal parameter; s6, inputting the data into a prediction model for training; s7, enabling the output variable to be a wind power prediction value at the to-be-predicted moment; s8, performing reverse normalization processing on the prediction value to obtain a final wind power prediction result; and S9, establishing an evaluation index system, and evaluating the accuracy of the wind power prediction result. According to the method, the problem of network degradation caused by network deepening in the traditional deep neural network model training process is solved, and the model prediction precision is improved.

The embodiment of the invention provides a user behavior big data analysis method and system based on a mobile internet. The method comprises the steps of obtaining a user operation behavior segment generated by a target user in a preset time period in a mobile internet scene, sequentially taking each user operation behavior in the user operation behavior segment as a target user operation behavior needing to be subjected to user behavior intention analysis, and then performing behavior intention analysis on the target user operation behavior according to a pre-generated user operation behavior sequence library comprising a plurality of user operation behavior sequences, and finally, acquiring the current behavior portrait of the target user through analysis according to the user behavior intention corresponding to each user operation behavior, and pushing corresponding mobile internet information to the target user according to the current behavior portrait. Thus, the user operation behavior intention of the target user based on the user operation behavior segmentation can be obtained through accurate analysis, and a good application effect can be obtained when the method is used for subsequent applications such as mobile internet information pushing.

The invention is suitable for the technical field of information, and provides a GIS-based refuse landfill site selection method, which comprises the steps of obtaining geographic data of a target area; establishing a geographic database and storing the geographic data into the geographic database; making geographic thematic maps corresponding to various geologies according to the geographic database; analyzing the geographic thematic map to obtain at least one candidate region in the target region; and constructing a site selection model to judge the site selection suitability of the candidate region, and when the candidate region meets the site selection suitability and conforms to the site selection model, judging that the candidate region is a selected region. The site selection methoddepends on a geographic information system platform, can utilize the spatial analysis function of the geographic information system and the intuition of result output to quickly and accurately selectthe refuse landfill with less influence on the surrounding environment, and is low in cost, short in time, high in evaluation accuracy, time-saving and labor-saving compared with the traditional manual on-site investigation.

The invention relates to an electric power monitoring system based on a big data fusion model. The system is divided into three layers: a field equipment layer, a communication management layer and a system management layer. The system adopts a CoxPH algorithm based on deep learning to perform characteristic value extraction on received data, and can complete various practical functions such as switch state monitoring, analog quantity acquisition, comprehensive protection monitoring, remote control, alarm management, curve bar graph analysis, report generation, statistics printing, user management, accident recall and the like. The power utilization reliability is enhanced, the power utilization management efficiency is improved, the power supply quality is improved, and the operation and maintenance cost is reduced.

The invention relates to an automobile stabilizer bar multidisciplinary optimization design method based on fatigue and tilting performance. The automobile stabilizer bar multidisciplinary optimization design method comprises the following steps: step 1, performing problem identification and P graph analysis; step 2, building a stabilizer bar fatigue simulation analysis model, conducting fatigue simulation analysis on the stabilizer bar, building a front suspension tilting simulation analysis model, and conducting tilting simulation analysis on the front suspension; step 3, carrying out parametric modeling of design variables; step 4, carrying out DOE sampling calculation of design variables; step 5, extracting DOE sample points and calculation results, constructing a response surface approximation model I based on fatigue performance, and constructing a response surface approximation model II based on roll performance; and step 6, based on the two approximate models in the step 5, performing optimization design on the stabilizer bar by taking the tilting performance as a constraint condition and taking the minimum fatigue damage as an optimization target. The stabilizer bar structure capable of simultaneously meeting fatigue and tilting performances is designed more efficiently by comprehensively applying CAE simulation and MDO technologies.

The invention discloses a method for reducing GPU (Graphics Processing Unit) memory occupation in a deep neural network training process. The method is suitable for a deep neural network (DNN) framework based on a CPU-GPU heterogeneous computing platform, provides basic-general-extreme memory optimization options, realizes zero-overhead memory allocation during operation through static computing graph analysis, selects a proper operator re-computing strategy through heuristic search, and improves the memory optimization efficiency. And meanwhile, the extra memory overhead of the Cudnn library convolution algorithm is analyzed and limited. A user can realize memory occupation optimization in a neural network training process only through optimization level control of one parameter.

A method includes maintaining, by a controller device managing a plurality of network devices, a data structure, and receiving, by the controller device, an indication of a status intent. The method further includes generating, by the controller device, a logical packet based on the stateful intent, and determining, by the controller device, a dependency intent of the data structure based on the logical packet. The method also includes determining that the dependent intent conflicts with the stateful intent. The method further includes, in response to determining that the dependency intent conflicts with the stateful intent, determining, by the controller device, a solution to resolve the conflict between the dependency intent and the stateful intent based on the dependency intent and the stateful intent, and outputting, by the controller device, an indication of the solution.