15 results about "Demand profile" patented technology

The invention provides a self-adaptive distribution system and method for GPU server load prediction, and the method comprises the steps: obtaining a monitoring data stream of a GPU server in real time, and constructing a monitoring sequence matrix according to the monitoring data stream; performing load mutation extraction on each sensitive mutation index in the monitoring sequence matrix to obtain a load pulse factor of each sensitive mutation index, constructing a multi-scale feature set, and generating a context representation vector sequence according to the multi-scale feature set; carrying out load prediction by using the context representation vector sequence to obtain a GPU load prediction sequence, and mapping the GPU load prediction sequence into a resource demand profile; and generating a resource migration cost matrix according to the current resource state information and the resource demand profile, and generating a resource allocation optimization strategy based on the resource migration cost matrix. According to the technical scheme provided by the invention, the abrupt change mode of the load can be accurately captured and represented, and the allocation decision is optimized by coupling the resource migration cost, so that the overall stability of self-adaptive resource allocation is realized.

Systems and methods for dynamically adjusting an application demand profile during slice fallback are provided. In some aspects, an attribute quality table for an application of a UE may be received. The attribute quality table determines a demand profile for the application on a 5G SA network. A network slice may be allocated based on the demand profile. When the UE falls back from the 5G SA network to a 5G NSA network or an LTE network, a negotiation utilizing the attribute quality table, current RF conditions, and the QoS capabilities of the 5G NSA network or the EPC of the LTE network determines an optimal combination of network resources that provides the highest quality user experience. In some aspects, the application adjusts, in real-time, the demand profile that best meets the available resources of the 5G NSA network or the EPC of the LTE network.

This invention relates to the fields of smart healthcare and medical information technology, and in particular to a humanistic personalized service recommendation system oriented towards the medical needs of the elderly. To address the problems of existing triage systems having a single dimension of demand profiles and statically coupled interactions, this system includes a data acquisition layer, a profile construction layer, a recommendation decision layer, an interaction adaptation layer, and a service presentation layer. It extracts medical needs and humanistic status features and uses a dual-channel attention mechanism to weightedly fuse them to generate a multi-dimensional demand profile, which is then matched with a service candidate library to generate personalized recommendations. Simultaneously, it calculates the interaction difficulty score based on real-time implicit behavioral signals, reverses the profile, and forward dynamically adjusts the presentation complexity level of the interface. This invention achieves closed-loop adaptation of recommendation strategies and interaction complexity, effectively reducing the operational threshold for elderly users.

Methods, systems, and devices for satellite operations are described. A system for satellite communications may include a payload, a power system, and a thermal management component. The payload may be configured to provide a service with varying levels of capacity based on a demand profile. The payload may consume electrical energy at a peak rate when a level of demand indicated by the demand profile is above a threshold and at a lower, off-peak rate when a level of demand indicated by the demand profile is below a threshold. The peak rate may exceed a rate at which electrical energy is generated by the power system. The thermal management component may process excess thermal energy generated by the payload when the payload operates at the peak rate. Processing the excess thermal energy may include storing thermal energy while the payload operates at the peak rate.

The invention provides a user interaction method and device, electronic equipment and a computer readable storage medium, and the method comprises the steps: obtaining user demand information, and constructing a user demand portrait based on demand types contained in the user demand information and demand details corresponding to each demand type; according to the user demand portrait, screening out at least one target demand task matched with the user demand portrait from a plurality of preset demand tasks, and pushing the target demand task to an associated account of the user; and monitoring the completion progress of the target demand task, and issuing the excitation resource corresponding to the target demand task to the associated account of the user according to the completion progress. Through the method, the accuracy of task pushing is improved, so that the user experience is improved.

Providing computer simulations of a hydrogen refilling station (HRS) includes obtaining an HRS architecture file including architecture parameters and a demand profile file including demand parameters. When at least one variable parameter is present in the architecture or demand parameters, each of the variable parameters having a specified range of values and step size, a plurality of model instances is generated for each unique combination of values for the at least one variable parameter. Each of the plurality of model instances is provided to at least one processing device and, at each of the at least one processing device, the model instance is executed and an executed model instance and model results are returned. Upon completion of execution of plurality of model instances, at least a portion of the model results are compiled in an output file.

This invention relates to a knowledge graph-based intelligent customer service question-and-answer precision recommendation and processing method, belonging to the field of electronic digital data technology. The method includes: constructing a conversation prediction knowledge graph based on the historical conversation counts from multiple channels; predicting the future conversation counts from multiple channels based on the conversation prediction knowledge graph and the current conversation counts from multiple channels; setting the number of robot agents and human agents based on the future conversation counts from multiple channels; accessing conversations from multiple channels, using robot agents to perform semantic understanding of user questions in the conversation, constructing a user demand profile, and determining whether to assign a human agent; if so, assigning the optimal human agent based on the status data of the human agent; if not, generating answers based on the user demand profile, conducting a dialogue, and generating recommended questions based on a question association graph. The question association graph records the relationships between multiple questions, which has the advantage of improving the quality and efficiency of customer service.

The invention relates to the technical field of large model reasoning resource dynamic allocation, in particular to a large model reasoning resource dynamic allocation method and system, and the method comprises the following steps: carrying out the stage division of a reasoning task, and building a resource demand portrait for each execution stage; according to the resource demand portrait of the first execution stage of the reasoning task, computing resources are allocated from a resource pool, and the reasoning task is started; in the reasoning task execution process, the resource consumption condition of the reasoning task is monitored, and stage conversion of the reasoning task is judged; according to the judgment result of the stage conversion and the resource demand portrait of the next execution stage of the reasoning task, the computing resources occupied by the reasoning task are adjusted, and the adjustment includes returning the computing resources which are not needed any more to a resource pool and applying for additional computing resources from the resource pool; and when the reasoning task is completed, returning all computing resources occupied by the reasoning task to the resource pool. The resource utilization rate and the task execution efficiency are improved.

The invention discloses an engineering bidding knowledge dynamic pushing method and system based on student behavior analysis, and belongs to the technical field of online education. Through full-dimensional behavior data acquisition and accurate demand feature mining, multi-dimensional behavior data of student knowledge demands are realized, a stereoscopic and accurate demand portrait is generated, current weak links and preferences of students are clarified, potential demands can be pre-judged, and it is ensured that pushed contents are highly matched with the student demands; based on demand feature core elements and multi-dimensional matching conditions, an intelligent matching engine is built to carry out bidirectional comparison and comprehensive evaluation, secondary verification and adaptation index screening are combined, target knowledge quality is ensured, pushing modes and opportunities are selected in combination with learning habits of students, core knowledge points and reinforcing contents are preferentially pushed, and pushing frequency is dynamically controlled. According to the method, blindness and interference of traditional pushing are avoided, the knowledge viewing rate and utilization efficiency are improved, and pushing service better meets individual requirements of trainees through optimization and continuous iteration.

The invention relates to the technical field of information analysis, and discloses a recruitment post information analysis and user intelligent distribution method and system. The method comprises the following steps: acquiring job-seeking voice data of a job seeker, and identifying job-seeking information elements in the job-seeking voice data; and performing demand feature analysis according to the job application information elements to obtain a job application demand file, and matching the job application demand file with an enterprise post library to obtain a job seeker post allocation result. According to the method, the job seeker can replace traditional character input through natural voice expression, the resume creation time is shortened, the job hunting threshold is reduced, and then the employment satisfaction degree of the job seeker is improved.

This invention discloses an intelligent allocation method for server GPU computing power for large-scale model training, relating to the field of computer resource management technology. It includes: S1: Determining the dependencies between stages in the large-scale model development workflow based on a directed acyclic graph (DAG), and obtaining the corresponding resource demand profile matrix; S2: Setting up a dynamic priority resource reservation pool based on the resource demand profile matrix, and generating a corresponding resource supply curve; S3: Determining the short-term predicted demand for each task based on the actual resource consumption and dynamic resource demand curve for each task on the critical path of the large-scale model development workflow, and comparing the short-term predicted demand with the resource supply curve to dynamically adjust the dynamic priority resource reservation pool. This invention can adapt to the fluctuations and uncertainties during training task execution, ensuring the progress of the critical path while flexibly handling excess resource demands.

Providing computer simulations of a hydrogen refilling station (HRS) includes obtaining an HRS architecture file including architecture parameters and a demand profile file including demand parameters. When at least one variable parameter is present in the architecture or demand parameters, each of the variable parameters having a specified range of values and step size, a plurality of model instances is generated for each unique combination of values for the at least one variable parameter. Each of the plurality of model instances is provided to at least one processing device and, at each of the at least one processing device, the model instance is executed and an executed model instance and model results are returned. Upon completion of execution of plurality of model instances, at least a portion of the model results are compiled in an output file.

This invention relates to the field of intelligent resource allocation technology, specifically an intelligent allocation system for legal benefits resources. The system includes: a user demand input module, a resource database module, an intelligent matching engine module, an allocation execution module, a dynamic optimization module, and a platform interaction interface. Its allocation method includes: collecting user demands for legal benefits and constructing demand profiles; constructing and updating legal benefit resource profiles, etc. The beneficial effects are: this invention achieves accurate, efficient, and automatic allocation of legal benefit service resources through intelligent means, effectively solving the problems of resource mismatch and low efficiency in traditional legal and other industry services. Its core purpose is to help parties involved (enterprises, families, individuals, various industries, and organizations) avoid the "profit maximization trap," obtain appropriate legal benefits, improve service quality and user satisfaction, achieve the goal of resolving disputes, and ultimately achieve a win-win situation.

An inventory machine learning process can be applied iteratively. The inventory machine learning process can be applied by a computing device. The inventory machine learning process can include determining similarity scores between demand profiles for individual parts and group centers for similarity groups of parts. The inventory machine learning process can include assigning individual parts into the similarity groups using similarity scores. The inventory machine learning process can include calculating representative demand profiles for each similarity group and determining whether the representative demand profile meets a threshold. When the threshold is met, inventory parameters can be determined for each similarity group based on the inventory machine learning process.

This invention relates to the field of visual element recommendation, and more particularly to a method and system for personalized visual communication element recommendation based on internet big data. The method includes the following steps: acquiring multi-dimensional internet behavior data of users and mining user interaction behavior on each platform to obtain a multi-platform user interaction content data pool; inferring user interaction intent from the multi-platform user interaction content data pool and mining personalized interaction interest preferences to obtain user interaction interest preferences; mining deep interaction needs based on user interaction interest preferences and performing adaptive profile evolution to construct a personalized user demand profile; and parsing visual elements and performing multimodal visual mining based on the multi-platform user interaction content data pool to obtain multimodal visual element information for each homepage. This invention improves the accuracy and quality of visual element recommendation through precise analysis of user aesthetic needs.