36 results about "User modeling" patented technology

The invention discloses an AI virtual coach training system based on standard action matching and deviation feedback, which relates to the technical field of AI virtual coach training systems and comprises a user modeling module, an action acquisition module, a template matching module, a deviation calculation module, a feedback generation module, an interactive presentation module and a learning optimization module. The user modeling module is used for modeling a registered user by adopting a body parameter acquisition and health data analysis method to obtain a user personalized feature vector; the action acquisition module is used for capturing actions executed by a user in real time by adopting a multi-source sensor fusion method to obtain a time sequence containing key point coordinates; and the template matching module is used for comparing the time sequence of the key point coordinates with corresponding actions in a preset standard action template library by adopting an improved dynamic time warping (DTW) algorithm to obtain an optimal matching path and a corresponding minimum matching cost.

The invention relates to an automatic modeling agent system based on an MCP protocol and a large model, and the system comprises a multi-modal input interface module which is used for receiving, analyzing and standardizing modeling instructions or intention descriptions inputted by a user through different media to form structured feature representations; the semantic understanding central module is used for performing semantic analysis on the structured feature representation by using a pre-trained large language model and outputting a user modeling intention; the MCP instruction compiler module is used for converting a user modeling intention into an MCP instruction sequence following an MCP protocol; and the modeling execution engine module is used for analyzing and automatically executing the MCP instruction sequence to complete modeling of the target three-dimensional model. According to the method, three-dimensional modeling can be completed without knowing professional modeling terms and concepts by a user, and the modeling automation and intelligence level is greatly improved.

The invention relates to the technical field of network information propagation prediction, and discloses a topic propagation prediction method and system based on multi-dimensional feature fusion, and the method comprises the steps: extracting multi-dimensional features, inputting the multi-dimensional features to a decoder layer of Transform for fusion, and constructing a prediction model for topic propagation prediction; the multi-dimensional features comprise user comprehensive influence, text emotion features, time dynamic features and user interaction behaviors. According to the method, through collaborative optimization of the dynamic user modeling module, the multi-modal feature fusion module and the intelligent time sequence analysis module, the social network propagation prediction accuracy is remarkably improved, and accurate prediction of the topic propagation trend in the social network is achieved.

The invention relates to the technical field of artificial intelligence education, and discloses a personalized online education system and method based on artificial intelligence, and the system comprises a data collection and perception module which is used for collecting the dynamic behavior data and multi-mode perception data of a learner; the knowledge processing and constructing module is used for processing teaching contents to construct a structured knowledge graph; the user modeling module is used for constructing a dynamic learner portrait comprising a knowledge state model and a cognitive-emotional state model for the learner based on the collected data and the constructed knowledge graph; and the intelligent decision and intervention module is used for generating and executing personalized learning intervention based on the dynamic learner portrait. According to the method, by fusing multi-mode cognitive state perception, reinforcement learning intervention decision and a knowledge graph dynamic evolution mechanism, accurate and timely personalized intervention for learners and intelligent guidance of optimal learning strategies are realized.

Disclosed is a finite element modeling system and method for hydraulic fracturing problems based on a large language model. The method is implemented by the system. The method comprises: a user inputting a key description of a finite element model desired to be generated; a large language model of a cloud service platform generating a key parameter file; a mesh tool generating a mesh file; a parameter mesh coupling tool generating a finite element model file; a computing server executing the finite element model file, a system output unit prompting the user of completion of modeling and execution, and a display interface displaying a visualized finite element model and an execution result; the user determining whether an expectation is satisfied; if the expectation is not satisfied, the user re-inputting the description; if the expectation is satisfied, the user outputting the finite element model and the execution result.

The invention is suitable for an information management method technology, and provides a data management analysis method and system based on credit investigation information, and the method comprises the steps: obtaining credit investigation data information; constructing edge weights according to the agent variables to form a multi-dimensional credit graph, identifying communities with similar credit behaviors according to existing data, obtaining positioning similar crowds, and judging contribution degree scores according to contribution degrees of white users; constructing a multi-dimensional credit behavior trajectory diagram according to the real-time behavior data flow and the platform data, identifying abrupt change nodes in the trajectory diagram, and judging whether the behavior change forms a signal of credit risk offset according to the abrupt change nodes; and quantifying the positive or negative influence amplitude of each behavior feature on the final score according to the final score, obtaining updated evidence material information according to the score elements, and obtaining a new score suggestion value according to the evidence material information. The key problems that in a traditional credit investigation system, white users are difficult to model, the scoring result is static and rigid, and appeal cannot be fed back are effectively solved.

The invention relates to the technical field, and discloses a personalized Chinese reading recommendation system based on semantic perception and an implementation method thereof, and the system comprises a user modeling module, a learner interest label and cognitive feature model construction unit, a semantic analysis module, a matching recommendation module and a feedback optimization module. The method has important innovative significance in the aspects of improving Chinese learning efficiency, enriching cultural connotation understanding and promoting personalized education development. The systematic design concept and technical means of the system not only can meet the individual requirements of different users, but also can continuously adapt to the change of the education environment, has wide popularization and application prospects, and provides a solid technical basis and innovative demonstration for the intelligent development of Chinese education.

The present application relates to the field of dimension marking, especially to a multi-dimensional dynamic dimension marking system for wood components. Dynamic dimension marking is a tool that can dynamically display the dimensions of a modeling object according to user operations. However, the mainstream dynamic dimension marking system can only actually realize the dynamic display of the dimensions of the modeling object when the structure of the modeling object is relatively simple. The present application comprises an information storage module for storing the parameters of the wood component and updating the parameters of the wood component when the parameters of the wood component change. The three-dimensional wood component formed by modeling is simplified into a series of point positions and functional relationships between the point positions. Through deep learning, the change of the functional relationship between the point positions is estimated according to the change of the point positions, and then the change of the shape and size of the wood component is quickly estimated, which greatly reduces the computational load and improves the generation speed of the dynamic dimension marking.

The invention belongs to the technical field of artificial intelligence and education crossing, and relates to an information security professional education system and method based on big language model agency, and the system comprises a data collection and storage module which is used for collecting and storing information security professional student data and teaching data; the user modeling and state management module is used for constructing multi-dimensional individual portraits and performing state management on the multi-dimensional individual portraits according to the characteristics of students professional for information security; the intelligent agent module is used for constructing a multi-agent large language model agent for an information security professional teaching scene; and the information security field professional knowledge representation and reasoning module is used for constructing a time-sequenced and reasonable knowledge system oriented to information security education so as to provide knowledge support for a multi-agent large language model agent. According to the method, a sequential and reasonable knowledge system oriented to information safety education is innovatively constructed, and accurate, time-efficient and reasonable knowledge support can be provided for teaching tasks through a retrieval enhancement generation mechanism.

The embodiment of the invention provides an information recommendation method and device, a storage medium and a program product. In the embodiment of the invention, the target geographic object associated with the user and the attribute data of the target geographic object in the target library are introduced, and the geographic semantic feature, the distance feature and the geographic context feature are incorporated into the user modeling process, so that the data normal form which only depends on established behavior data and static portraits is broken through; on this basis, according to attribute data and historical behavior data of the target geographic object, steady-state demand potential information and dynamic interest intention information of the user are predicted, and user demand features are described from two dimensions of long-term steady-state demand and short-term dynamic interest; furthermore, the steady-state demand potential information and the dynamic interest intention information are combined with candidate information objects in the target application scene to perform joint information recommendation, so that the recommendation result can not only fit the current interest of the user, but also keep the consistency with the long-term demand of the user, and the recommendation effect is greatly improved.

The invention discloses an AI-based multi-modal digital content intelligent recommendation method and system, relates to the technical field of artificial intelligence and education recommendation, and is used for solving the problems of inaccurate matching of multi-modal learning resources and user intentions, false feature interference and unclear semantic links in a weak network scene. Firstly, intention analysis and user modeling are carried out on user multi-modal interaction behaviors through a large model, and user feature vectors with unified dimensions and stable calibers are formed; and the resource side extracts semantic and time sequence features, and introduces preview elf chart recognition in a weak network scene. In order to suppress false features, a minimum evidence account book and a three-point co-location check chain are provided, and accounting management is carried out on evidence lease; and main channel writing is released only when the real frame backbone is consistent with the subtitle key points. Then combining modal gating fusion, knowledge network anchoring and path construction, and finally implementing hierarchical reordering according to the reference semantic vector and the incidence matrix to generate a traceable, verifiable and robust personalized recommendation result.

The invention belongs to the technical field of computer collaborative modeling, and discloses a system modeling language online collaborative modeling method and system in a network environment. The method comprises the steps that user modeling operation is packaged into an editing event through a browser side and sent to a server side, the server side conducts concurrency control through a multi-granularity locking mechanism, a system modeling language graph serves as the main locking granularity, and model elements serve as the auxiliary locking granularity; the browser end maintains a command execution queue, and multi-end state synchronization is achieved through a version field; the system supports a design intention interaction mechanism, allows a non-lock holder to request operation through an intention event, and allows a server to forward the operation to a lock holder and execute corresponding operation according to feedback, thereby realizing state consistency, concurrency control and intention communication in a collaborative modeling process.

The invention relates to a method and a system for automatically generating a SysML v2 visual analysis model of system engineering, belongs to the technical field of program analysis, and solves the problem that a traditional system engineering document cannot be effectively and automatically modeled in the prior art. The method comprises the following steps: converting an input initial system engineering design document into a document structured representation model; extracting related document contents from the document structured representation model according to user modeling requirements, extracting related knowledge entries from a pre-configured domain knowledge base, and extracting related case entries from a pre-configured modeling case library; and inputting the extracted document content, knowledge entries and case entries as context information into a large language model, and generating a SysML v2 visual analysis model of the system engineering by using the large language model according to the modeling requirement of a user. Based on the document structured representation model, the field knowledge base and the modeling case base, automatic analysis and modeling of the system engineering document are achieved.

This application relates to the field of healthcare informatics technology, providing a drug recommendation method and related equipment based on drug representation and dynamic user modeling. User features are generated based on the acquired user's historical health records and current health status. Diagnostic features and procedural features are sequentially input into a GRU network and a Transformer network, respectively, to generate user representations through dynamic modeling. Drug features are input into a pre-constructed graph attention network to construct a heterogeneous graph between drug attributes and molecular motifs. Drug representations are generated by message propagation and stacking on this heterogeneous graph. User and drug representations are input into a pre-constructed feedforward neural network, outputting fused features between the drug and the user. The fused features are used to generate probabilities through an activation function, and recommendation information is generated based on the target drugs corresponding to these probabilities. This method can accurately match user health needs and provide personalized and effective drug recommendations.

The present disclosure provides an interactive and parametric modeling automation testing method for CAD, the method comprising: step S10, starting the script recording mode, capturing all user modeling operation instructions; step S20, serializing the user modeling operation instructions, and associating the mouse screen coordinates recorded in the mouse operation instructions with the corresponding three-dimensional world coordinates, and writing them into a script file; step S30, starting the script playback mode and loading the script file; step S40, parsing the script file, and sequentially deserializing and generating user modeling operation instructions. For the mouse operation instructions, the three-dimensional world coordinates in the script file are used as a reference, combined with the viewport size, camera transformation matrix and working plane parameters in the current test running environment, the reconstructed mouse screen coordinates are calculated, the mouse operation instructions are updated according to the reconstructed screen coordinates, and the updated user modeling operation instructions are executed. The method disclosed herein supports all user operation types to ensure the comprehensiveness and accuracy of the test.

A system and method for presenting laboratory data to a user are presented. The system comprises a perception component for continuously gathering in-situ context data regarding a laboratory and the user, a user modeling component for receiving the in-situ context data from the perception component to create a user specific model for each user, a laboratory device awareness component for monitoring the status, performance, alarms, and / or maintenance of the laboratory devices, a notification component for receiving the in-situ context data from the perception component and the laboratory device status data from the laboratory device awareness component and for processing and determining which data from the in-situ context data and the laboratory device status data are to be presented to the user, and a presentation component for presenting the processed data from the notification component. The presented data comprises both public and private notification of the data to the user.

The invention discloses a fairness-enhanced graph neural network social recommendation method, which comprises the following steps of: firstly, designing a fine-grained preference modeling mechanism for mining potential interests of low-activity users and relieving the problem of incomplete user representation caused by insufficient interactive data so as to reduce modeling deviation of a user side; secondly, in order to solve the problem of insufficient information propagation caused by uneven relationship distribution in the social network, a similarity-based social propagation mechanism is provided, and an original social graph structure is reconstructed, so that information can be efficiently and directionally propagated among users, and the information acquisition capability of low-activity users is improved. And finally, the interest similarity between the users is combined with social structure features, the expression ability of user representation is enhanced, and the coverage effect of the social relationship is improved. According to the method, the association between the social recommendation task and the fairness target is considered, the model can learn user preferences more comprehensively through joint optimization of the user modeling and propagation strategy, and the prejudice problem of the user side is effectively relieved through structural modification and information enhancement. According to the method, the recommendation accuracy and the user fairness are remarkably improved.

The invention belongs to the technical field of vision training, and particularly relates to a children personalized vision training system and method based on multi-parameter self-adaption, and the method comprises the following steps: S1, obtaining clinical baseline data of a user, the information at least comprising basic physiological parameters of the user and an initial visual ability evaluation target; guiding a user to complete a standardized baseline visual evaluation task, and synchronously acquiring multi-modal dynamic response data of the user; and S2, inputting the clinical baseline data and the multi-modal dynamic response data into a fusion analysis model constructed based on an information bottleneck theory. According to the invention, personalized, dynamic, safe and controllable visual function rehabilitation training in a real sense can be realized through multi-modal data fusion analysis, user modeling based on an information bottleneck theory, a dual-time scale adaptive decision engine and a man-machine cooperation calibration mechanism; and the scientificity and effectiveness of the training scheme are guaranteed through a continuous learning mechanism.

The invention provides a CAD (computer-aided design)-oriented interactive and parametric modeling automatic testing method, which comprises the following steps of: S10, starting a script recording mode, and capturing all user modeling operation instructions; step S20, serializing a user modeling operation instruction, associating a mouse screen coordinate recorded in a mouse operation instruction with a corresponding three-dimensional world coordinate, and writing into a script file; step S30, starting a script playback mode and loading a script file; and step S40, analyzing the script file, sequentially deserializing to generate a user modeling operation instruction, for the mouse operation instruction, calculating to obtain a reconstructed mouse screen coordinate by taking the three-dimensional world coordinate in the script file as a reference and combining the viewport size, the camera transformation matrix and the working plane parameter in the current test operation environment, and performing the operation of the user modeling operation instruction. And updating a mouse operation instruction according to the reconstructed screen coordinate, and executing the updated user modeling operation instruction. According to the method disclosed by the invention, full user operation types are supported, and the comprehensiveness and accuracy of the test are ensured.

This invention discloses a dynamic user profile update method based on multimodal data, relating to the field of user modeling technology. The method includes: acquiring the collaborative analysis software of each software in the user's device; obtaining the association weight based on the association filtering value of the collaborative analysis software; acquiring status update words and updating the user's profile based on the collaborative analysis software of the updated reference software and the association weight of the collaborative analysis software. This invention addresses the problem in existing dynamic user profile update methods that lack analysis of multimodal data generated by users based on user software usage data. This results in the inability to effectively acquire dynamic profile data related to the user's immediate interests and needs, leading to inaccurate user profiles and a loss of dynamism after data processing.

The invention relates to a recommendation system (100) and method (200) where personalized recommendation clusters are computed by querying streaming data for flow-based recommendation systems used in the machinery and power systems sector, and these recommendations are presented to the user.

According to the message pushing method based on user behavior analysis provided by the invention, the fused user portrait is generated by obtaining the multi-source heterogeneous data stream and carrying out space-time alignment and cross-modal feature fusion, so that the comprehensiveness and accuracy of the user portrait are ensured. And performing real-time user intention reasoning through an edge end lightweight time sequence prediction model based on the fused user portrait, and combining with a federated learning framework to aggregate encryption model parameters to update a global recommendation model so as to generate a target global recommendation model. And based on the target global recommendation model and the target push candidate set, generating a target push strategy by integrating causal reinforcement learning of anti-factual reasoning. And performing attribution analysis based on user behavior feedback data corresponding to the target push strategy, and optimizing system model parameters to form closed-loop iteration. By constructing a complete technical link, the technical problems of difficulty in multi-modal data collaboration, complexity in real-time computing power scheduling and insufficient privacy security protection caused by unidirectional user modeling of an existing push system are solved.

The invention discloses a Modelica intelligent modeling optimization method based on AI-Agent. The Modelica intelligent modeling optimization method is used for modeling an electrical-thermal coupling system of a high-dynamic and high-reliability energy system and an energy battery. The method comprises the following steps: acquiring device parameters, an equivalent circuit and an electro-thermal coupling dynamic equation set of a target energy system or a battery unit to be researched, and inputting the data and user modeling requirements into AI-Agent; the AI-Agent triggers a physical knowledge retrieval module to retrieve related physical knowledge from a knowledge base, then calls a mathematical equation processing module to process an equation set of the physical knowledge, outputs an optimized electric-thermal coupling equation set, and generates an initial Modelica model of the target based on the optimized equation set; and carrying out model code verification and error correction, and outputting a final optimized model. The model obtained through the method is more accurate and higher in robustness, the model development period is greatly shortened, and the model simulation speed is greatly increased.

An electric power network electric quantity balance deduction method based on a graph neural network comprises the steps that clean energy power generation equipment is modeled as a power generation node, a user is modeled as a power utilization node, a transformer substation is modeled as a transmission node, and a topological graph structure comprising a node set and an edge set is constructed; training a clean energy generating capacity prediction model and a user electricity consumption prediction model based on the topological graph structure, and generating a generating capacity prediction value of a power generation node and an electricity consumption prediction value of an electricity consumption node; injecting the generating capacity predicted value and the power consumption predicted value into the topological graph structure, training a power network power balance evaluation model, and constructing a power dispatching strategy generation model based on the minimum power dispatching cost; and determining the power grid unbalance degree according to the output of the power network electric quantity balance evaluation model, and when the unbalance degree exceeds a threshold value, triggering the power dispatching strategy generation model to output an optimization scheme and adjusting power grid operation parameters. The method can output the power dispatching strategy according to the characteristics of the power grid to meet the power utilization guarantee of users and reduce the power loss.

The invention belongs to the technical field of news recommendation, and provides a personalized news recommendation method based on causal intervention. According to the method, firstly, features of news are constructed through news titles, categories and news click behaviors; secondly, combining popularity features with a candidate perception recommendation model, and learning a news recommendation model containing popularity deviation; then, a causal intervention technology is introduced, false correlation between news features and popularity is cut off through a do-dry budget, and interference of popularity deviation on user interest modeling is effectively restrained; and finally, an inference mechanism based on intervention is designed, so that the adverse effect of news popularity deviation on user modeling is weakened. According to the method, the correlation of the user news is reordered by utilizing a causal intervention technology, so that the accuracy of news recommendation is improved, and technical support is provided for a personalized news recommendation system.

The invention provides a large model enhancement recommendation method and system based on multi-view nonlinear modeling, and relates to the technical field of computer personalized recommendation, and the method comprises the steps: constructing a recommendation project library; obtaining a first historical interaction record of browsing the recommended project library by at least one user; constructing a multi-head Fourier recommendation model, wherein the multi-head Fourier recommendation model comprises a user history mining module, a Fourier-kernel adaptive LLM encoder and a user modeling module; performing end-to-end optimization training on the multi-head Fourier recommendation model to obtain a trained multi-head Fourier recommendation model; and obtaining a historical interaction record of a target user, calculating the click probability of each item in the recommended item library clicked by the user based on the trained multi-head Fourier recommendation model, and judging a user click result. By constructing the Fourier-kernel adaptive LLM encoder, the limitation that a nonlinear relationship in a high-dimensional space is neglected in an existing method is solved, and the expression ability and recommendation precision of high-dimensional sparse features are remarkably improved.

This application provides a modular simulation modeling system, method, electronic device, and storage medium, comprising: a model development device for responding to user modeling requirements and generating a model matching the modeling requirements; a component modeling device for responding to user component modeling operations, determining a target component general template corresponding to the component modeling operation from a component general template, and generating a target component based on the target component general template and the model; an entity assembly device for responding to user entity modeling operations, determining a target entity general template corresponding to the entity modeling operation from an entity general template, and generating a target entity based on the target entity general template and the target component; and a distributed deployment device for deploying different models on servers with different computing power according to the model computing power requirements of multiple target models. This application can improve the flexibility and convenience of modeling in different application scenarios and can improve the resource utilization rate of simulation after modeling.

The present invention relates to an automatic modeling agent system based on the MCP protocol and a large model. The system comprises: a multimodal input interface module for receiving, parsing, and standardizing modeling instructions or intention descriptions input by users through different media to form structured feature representations; a semantic understanding hub module for using a pre-trained large language model to perform semantic analysis on the structured feature representations and output the user's modeling intentions; an MCP instruction compiler module for converting the user's modeling intentions into MCP instruction sequences that comply with the MCP protocol; and a modeling execution engine module for parsing and automatically executing the MCP instruction sequences to complete the modeling of the target three-dimensional model. This system allows users to complete three-dimensional modeling without having to understand specialized modeling terminology and concepts, significantly improving the level of modeling automation and intelligence.

The invention discloses a user modeling method and system of a universal user encoder based on Mindpoint, and relates to the technical field of recommendation systems. By integrating the multi-modal information contained in the interaction sequence of the user, comprehensive understanding of the user is realized, and the user encoder can generate rich user representations so as to capture interests and preferences of the user more accurately. By means of fusion of multi-modal information, the performance of the recommendation system in personalized recommendation tasks is improved. And a new Hard-Negative negative sampling strategy is adopted, so that the learning ability of a user encoder in distinguishing preferential and non-preferential items of the user is enhanced, and the discrimination ability of the user encoder is improved. According to the method, the historical interaction information and the attribute information of the user are utilized to form richer and multi-dimensional user representation, so that the recommendation accuracy can be improved, and the migration capability of a user encoder can also be improved. Various different data enhancement is performed on the interaction sequence, so that various features among the articles in the interaction sequence are extracted.

A system and method for machine learning platform management includes a unified architecture for developing and deploying machine learning models at scale. The platform integrates feature generation, model training, and inference services through a centralized interface. The system processes source data through a feature platform to generate training datasets and real-time features. A multi-stage training pipeline enables automated model experimentation through configurable workflows combining core frameworks and user modeling code. The platform implements specialized inference services optimized for high-throughput ranking and recommendation use cases, with distributed feature stores and local caching for efficient feature serving. A comprehensive monitoring system tracks model performance, feature distributions, and prediction quality through automated anomaly detection. The platform enables rapid experimentation while maintaining production reliability through automated deployment orchestration, optimized inference engines, and continuous feedback loops for model improvement.