Online resource management system and method based on AI large model

Abstract

The invention belongs to the technical field of online educational resource management, and particularly relates to an online resource management system and method based on an AI large model. According to the method, spatial and temporal features in an online resource scene are analyzed, a basic framework containing a time axis feature vector and a spatial resource association degree matrix can be constructed, the resource conflict probability and the conflict coverage range in a demand interval are predicted on the basis of the basic framework, and when the resource conflict probability exceeds a preset evaluation threshold value, the resource conflict coverage range is predicted. A dynamic decoupling mechanism is triggered, a resource allocation strategy corresponding to the service emergency degree and the space-time adaptation degree is generated through decoupling calculation, accurate matching of resources is achieved, in the process of executing the resource allocation strategy, resource conflict events can be monitored in real time, and when resource conflicts are found, standby resource scheduling and a cross-regional coordination mechanism are started immediately. And through continuous feedback and optimization, a closed-loop optimization mechanism can be formed step by step, and the efficiency and accuracy of resource management are continuously improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of online education resource management, and particularly relates to an online resource management system and method based on an AI large model. Background Art

[0002] With the continuous development of Internet technology, educational resources in the field of education have gradually shifted from offline to a combination of offline and online. Online education is the so-called e-learning. Due to its convenience, flexibility, and richness, online education resources enable learners to study anytime and anywhere, greatly improving the learning efficiency and effect. However, with the continuous increase of online education resources, how to efficiently manage and allocate these online education resources has become an urgent problem to be solved.

[0003] In the prior art, although there are some online education resource management systems, these often rely on traditional manual management or simple automated algorithms, and it is difficult to achieve intelligent allocation and optimization of resources. For example, in the face of a large number of diverse online education resources, it is often impossible to accurately predict resource requirements, optimize resource allocation, and effectively handle resource conflicts. Similarly, it is also difficult to perform dynamic feedback and optimization based on the resource scheduling effect, which will undoubtedly lead to problems such as low resource allocation efficiency, resource waste, and poor learner experience. Based on this, this solution provides an online resource management method based on an AI large model to solve the above problems. Summary of the Invention

[0004] The purpose of the present invention is to provide an online resource management system and method based on an AI large model, which can intelligently manage and allocate online education resources, and achieve accurate prediction of resource requirements, optimized processing of resource allocation, and effective resolution of resource conflicts.

[0005] The technical solution adopted by the present invention is specifically as follows: An online resource management method based on an AI large model, comprising: Separating the spatio-temporal features in the online resource scenario, constructing a time-axis feature vector including planned time, adjustment time, and emergency degree, and quantifying the spatial resource correlation matrix of the resource sharing ability between different regions; Performing fusion analysis on the time-axis feature vector and the spatial resource correlation matrix, predicting the resource conflict probability and conflict coverage range within the demand interval, and triggering a dynamic decoupling mechanism for resource optimization when the resource conflict probability exceeds a preset evaluation threshold; Generating a resource allocation strategy corresponding to the service emergency degree and spatio-temporal adaptability based on the decoupling calculation of time weight and spatial weight; Implement backup resource scheduling and cross - regional coordination for resource conflict events under the execution of the resource allocation strategy, synchronously generate a resource scheduling execution report, then count the scheduling success rate and scheduling delay rate of backup resource scheduling and cross - regional coordination, and conduct a resource scheduling effect evaluation to output quantitative resource scheduling effect indicators; Feed back the resource scheduling effect indicators to the construction process of the time - axis feature vector and the spatial resource correlation matrix, and synchronously update the time - axis feature vector and the spatial resource correlation matrix to form a closed - loop optimization mechanism.

[0006] In a preferred solution, the steps of separating the spatio - temporal features in the online resource scenario, constructing a time - axis feature vector including planned time, adjustment time, and emergency degree, and quantifying the spatial resource correlation matrix of the resource sharing ability between different regions include: Obtain historical resource scheduling logs, perform time - series segmentation and geographical grid division on the historical resource scheduling logs to obtain historical resource scheduling time - series data and geographical grid data; Conduct time - series correlation analysis on the time - series data and geographical grid data, extract the time features of planned time, adjustment time, and emergency degree, and fuse them into a time - axis feature vector; Obtain the computing resource transmission efficiency and transmission loss between different geographical nodes, as well as the physical distance and topological connection relationship between geographical nodes, and establish a spatial resource correlation matrix including the supply capacity, response speed, and load status between geographical nodes; Normalize the time - axis feature vector and the spatial resource correlation matrix and integrate them into a joint feature set; Among them, the time - series correlation analysis uses a dynamic window to capture the repeated pattern of the planned time, the elastic change range of the adjustment time, and the trend of the attenuation of emergency events over time.

[0007] In a preferred solution, the steps of performing fusion analysis on the time - axis feature vector and the spatial resource correlation matrix to predict the resource conflict probability and conflict coverage range within the demand interval include: Analyze the joint feature set and establish a dynamic mapping relationship between the planned time fluctuation and the spatial node transmission loss to capture the change trend of resource demand at different time points and spatial nodes; Use a dynamic time window to track the change trajectory of the elastic coefficient of the adjustment time, and combine the time - decay trend of emergency events to generate a conflict factor in the time dimension; Obtain the supply capacity gradient difference between geographical nodes and the physical distance weight between geographical nodes, where the greater the physical distance, the smaller the physical distance weight; Under the demand interval, combining the supply capacity gradient difference and the physical distance weight, calculate the conflict diffusion intensity value in the spatial dimension, and then perform weighted superposition of the conflict diffusion intensity value and the conflict factor in the time dimension to output the predicted value of the resource conflict probability and generate a predicted map of the conflict coverage range in units of geographical grids; According to the topological connection relationship between geographical nodes, perform connectivity verification on the over-limit geographical grids in the predicted conflict coverage map. After the connectivity verification passes, mark the over-limit geographical grids as potential conflict areas. Otherwise, mark the over-limit geographical grids as isolated conflict points.

[0008] In a preferred solution, the step of performing connectivity verification on the over-limit geographical grids in the predicted conflict coverage map includes: Traverse the topological connection relationship between geographical nodes, extract all physical paths that are directly and indirectly connected to the over-limit geographical grids, and mark them as paths to be verified; Based on the resource transmission efficiency and transmission loss of the current path, calculate the total resource transmission time and the total resource loss value on each path to be verified; Sort the paths to be verified in ascending order according to the total resource transmission time, and based on the sorting result, select the first several paths to be verified with the smallest total resource transmission time as the candidate path set in combination with the preset candidate path quantity threshold; Accumulate the total resource loss values of the candidate paths to obtain the total loss value of the candidate path set, and compare the total loss value with the preset loss threshold; If the total loss value is less than the loss threshold, it is determined that the over-limit geographical grid passes the connectivity verification, has the potential for resource conflict propagation, and the over-limit geographical grid is included in the potential conflict area; If the total loss value is greater than or equal to the loss threshold, it is determined that the over-limit geographical grid fails the connectivity verification, indicating that the resource conflict influence range of the corresponding over-limit geographical grid is limited, and it is marked as an isolated conflict point.

[0009] In a preferred solution, when the resource conflict probability exceeds the preset evaluation threshold, the steps of triggering the dynamic decoupling mechanism for resource optimization include: Perform multi-level resource unbinding on the potential conflict area, calculate the spatio-temporal adaptation degree weights of each sub-link by decomposing the resource dependence chain between regions; Perform local resource adjustment on the isolated conflict points and reallocate the resource loads within each geographical node; Determine the priorities of resource unbinding and local resource adjustment according to the spatio-temporal adaptation degree weights of each sub-link and the reallocated resource loads, and gradually implement resource unbinding and local resource adjustment according to the priority order; Monitor the progress of resource unbinding in real time, as well as the real-time load changes during local resource adjustment. When the resource unbinding progress reaches the preset unbinding ratio or the local resource adjustment reaches the load balancing state, trigger the resource reallocation process.

[0010] In a preferred solution, the step of generating a resource allocation strategy corresponding to the service urgency and spatio-temporal adaptability through decoupling calculation based on time weight and space weight includes: Calculate the allocation ratios of time weight and space weight according to the resource unbinding progress and the resource load after local resource adjustment; Classify the resource requirements according to the service urgency, and combine the allocation ratios of time weight and space weight to generate a resource allocation priority list under different service urgencies; Evaluate the resource allocation priority using the spatio-temporal adaptability, determine the adaptability of resources at different times and in different spaces, and obtain the matching degree score between resource supply and demand; Compare the matching degree score with a preset matching degree threshold. If the matching degree score is higher than the matching degree threshold, confirm that the resource allocation strategy is effective, and continue to perform resource unbinding and local resource adjustment; If the matching degree score is lower than or equal to the matching degree threshold, recalculate the allocation ratios of time weight and space weight until the matching degree score is higher than the matching degree threshold.

[0011] In a preferred solution, the step of statistically calculating the scheduling success rate and scheduling delay rate of standby resource scheduling and cross-regional coordination, evaluating the resource scheduling effect, and outputting a quantified resource scheduling effect index includes: Capture resource scheduling requests in real time, and record the start time and end time of resource scheduling; Calculate the actual delay time of each resource scheduling, and calculate the scheduling delay rate based on the actual delay time; Statistically calculate the number of successfully completed resource scheduling times and the number of unsuccessfully scheduled times within a preset time period, and calculate the scheduling success rate; Conduct a dual evaluation of the scheduling success rate and the scheduling delay rate: The first evaluation is an independent evaluation. Compare the scheduling success rate and the scheduling delay rate with the preset scheduling success rate threshold and scheduling delay rate threshold respectively. When the scheduling success rate does not meet the scheduling success rate threshold or the scheduling delay rate does not meet the success rate threshold, trigger the early warning mechanism, generate a scheduling exception report, and mark it as an item to be optimized; The second evaluation is a comprehensive evaluation, which weights and sums the scheduling success rate and the scheduling delay rate to obtain a comprehensive evaluation score of the resource scheduling effect. Then, the comprehensive evaluation score is compared with a preset comprehensive evaluation score threshold. When the comprehensive evaluation score is lower than the comprehensive evaluation score threshold, an early warning mechanism is triggered, a scheduling exception report is generated, and it is marked as an item to be optimized.

[0012] In a preferred solution, the step of feeding back the resource scheduling effect indicators to the construction processes of the time-axis feature vector and the spatial resource correlation matrix and synchronously updating the time-axis feature vector and the spatial resource correlation matrix includes: Collect the scheduling success rate and the scheduling delay rate in the resource scheduling indicators and map them to the burst emergency weight correction term of the time-axis feature vector and the transmission efficiency attenuation coefficient of the spatial resource correlation matrix; Collect the deviation value between the scheduling success rate and the scheduling success rate threshold and record it as the scheduling success rate deviation; Based on the scheduling success rate deviation, dynamically adjust the weight ratio of the planned time to the adjusted time in the time-axis feature vector and recalculate the time decay rate of the burst emergency; Based on the scheduling delay rate, inversely correct the physical distance weight in the spatial resource correlation matrix to improve the load distribution priority of low-latency geographical nodes.

[0013] The present invention also provides an online resource management system based on an AI large model, which uses the above-mentioned online resource management method based on an AI large model, including: A data extraction module, which is used to separate the spatio-temporal features in the online resource scenario, construct a time-axis feature vector including the planned time, the adjusted time, and the burst emergency, and quantify the spatial resource correlation matrix of the resource sharing ability between different regions; A resource optimization module, which is used to perform fusion analysis on the time-axis feature vector and the spatial resource correlation matrix, predict the resource conflict probability and the conflict coverage range within the demand interval, and when the resource conflict probability exceeds the preset evaluation threshold, trigger a dynamic decoupling mechanism for resource optimization; A resource allocation module, which is used to generate a resource allocation strategy corresponding to the service emergency degree and the spatio-temporal adaptation degree based on the decoupling calculation of the time weight and the spatial weight; A resource coordination module, which implements standby resource scheduling and cross-region coordination for resource conflict events under the execution of the resource allocation strategy, synchronously generates a resource scheduling execution report, then counts the scheduling success rate and the scheduling delay rate of the standby resource scheduling and the cross-region coordination, and conducts a resource scheduling effect evaluation to output a quantified resource scheduling effect indicator; A feedback optimization module is used to feedback the resource scheduling effect metrics to the construction processes of the time-axis feature vector and the spatial resource correlation matrix, and synchronously update the time-axis feature vector and the spatial resource correlation matrix to form a closed-loop optimization mechanism.

[0014] And, an electronic device, the electronic device includes: At least one processor; And a memory communicatively connected to the at least one processor; Wherein, the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the above-mentioned online resource management method based on the AI large model.

[0015] The technical effects achieved by the present invention are as follows: By analyzing the spatio-temporal features in the online resource scenario, and based on this, a time-axis feature vector and a spatial resource correlation matrix are constructed. The two together constitute the basic framework for resource management and allocation. Based on this, the resource conflict probability and conflict coverage range within the demand interval can be predicted. When the resource conflict probability exceeds the preset evaluation threshold, a dynamic decoupling mechanism will be triggered, and a resource allocation strategy corresponding to the service urgency and spatio-temporal adaptability will be generated through decoupling calculation, which not only considers the urgent demand for resources, but also takes into account the adaptability of resources in time and space, so as to achieve precise matching of resources. During the execution of the resource allocation strategy, resource conflict events will be monitored in real time. When a resource conflict is found, the standby resource scheduling and cross-region coordination mechanism will be immediately started to ensure the stable supply of resources. At the same time, a resource scheduling execution report will be synchronously generated, recording the scheduling success rate and scheduling delay rate of the standby resource scheduling and cross-region coordination, providing data support for the evaluation of the resource scheduling effect. Through continuous feedback and optimization, a closed-loop optimization mechanism can be gradually formed, continuously improving the efficiency and accuracy of resource management. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a schematic flowchart of the method of the present invention; Figure 2 is a schematic diagram of the system module of the present invention; Figure 3 is a schematic diagram of the structure of the electronic device of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0017] In order to make the above objects, features and advantages of the present invention more obvious and understandable, the following detailed description of the specific embodiments of the present invention will be made with reference to the accompanying drawings of the specification.

[0018] In the following description, many specific details are set forth in order to provide a thorough understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0019] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation manner of the present invention. The "in a preferred embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it an individual or alternative embodiment that is mutually exclusive with other embodiments.

[0020] Please refer to Figure 1 As shown, the present invention provides an online resource management method based on an AI large model, including: S1. Separate the spatio-temporal features in the online resource scenario, construct a time-axis feature vector including planned time, adjustment time, and sudden urgency, and quantify the spatial resource correlation matrix of the resource sharing ability between different regions; In the step S1, when it is necessary to effectively manage and optimize online education resources, it is first necessary to identify and separate the spatio-temporal features in the online resource scenario. The planned time refers to the preset resource usage or allocation time point, the adjustment time refers to the adjustment of the planned time according to actual needs, and the sudden urgency is the assessment based on the suddenness and urgency of resource requirements. By constructing the time-axis feature vector, it is possible to understand the resource demand situation and urgency at different time periods, providing a basis for the effective scheduling of resources. At the same time, quantifying the spatial resource correlation matrix of the resource sharing ability between different regions can clarify the complementarity and dependence of resources between regions, thus providing decision-making support for cross-regional resource scheduling. When constructing the spatial resource correlation matrix, the considered factors include but are not limited to geographical location, network bandwidth, resource type, and quality, etc., to ensure that the matrix can truly reflect the resource sharing ability between regions. Among them, the steps of separating the spatio-temporal features in the online resource scenario, constructing a time-axis feature vector including planned time, adjustment time, and sudden urgency, and quantifying the spatial resource correlation matrix of the resource sharing ability between different regions include: Obtain the historical resource scheduling log, and perform time series segmentation and geographical grid division on the historical resource scheduling log to obtain historical resource scheduling time series data and geographical grid data; Perform time series correlation analysis on the time series data and geographical grid data, extract the time features of planned time, adjustment time, and sudden urgency, and fuse them into a time-axis feature vector; Obtain the computing resource transmission efficiency and transmission loss between different geographical nodes, as well as the physical distance and topological connection relationship between each geographical node, and establish a spatial resource correlation matrix including the supply capacity, response speed, and load status between geographical nodes; Normalize the time-axis feature vector and the spatial resource correlation matrix, and integrate them into a joint feature set; Among them, time series correlation analysis uses a dynamic window to capture the repeated patterns of the planned time, the elastic change range of the adjustment time, and the trend of sudden emergency events decaying over time; Specifically, when separating the spatio-temporal features in the online resource scenario, first perform time series segmentation and geographical grid division on the historical resource scheduling logs. Time series segmentation can identify the resource demand patterns in different time periods, while geographical grid division can identify the resource distribution and flow characteristics in the geographical space, so as to determine the spatio-temporal rules of resource scheduling. Then, perform time series correlation analysis on the time series data and geographical grid data, so as to extract the time features of the planned time, adjustment time, and sudden emergency degree, and fuse the time features into a time-axis feature vector. When establishing the spatial resource correlation matrix, consider the computing resource transmission efficiency and transmission loss between different geographical nodes, as well as the physical distance and topological connection relationship between each geographical node. The two jointly determine the sharing efficiency and scheduling feasibility of resources between different geographical nodes, and based on this, a spatial resource correlation matrix that can comprehensively reflect the resource sharing ability between regions. After that, normalize the time-axis feature vector and the spatial resource correlation matrix to eliminate the dimensional differences between different features, make the data comparable, and facilitate subsequent resource scheduling decisions. Integrating them into a joint feature set provides a unified data basis for subsequent resource scheduling and effect evaluation. Through the joint feature set, the dual factors of time and space can be comprehensively considered to achieve efficient resource scheduling. It should be noted that time series correlation analysis uses the dynamic time warping algorithm and the sliding window technique to capture the periodic patterns of the planned time, the flexible change range of the adjustment time, and the trend of sudden emergency events gradually weakening or strengthening over time, so as to ensure the accuracy and practicality of the time-axis feature vector.

[0021] S2. Perform fusion analysis on the time-axis feature vector and the spatial resource correlation matrix, predict the resource conflict probability and conflict coverage range within the demand interval, and when the resource conflict probability exceeds the preset evaluation threshold, trigger a dynamic decoupling mechanism for resource optimization; In the step S2, after the time-axis feature vector and the spatial resource correlation matrix are output, the time-axis feature vector and the spatial resource correlation matrix are integrated together, so as to predict the resource conflict probability within the demand interval and the scope that the conflict may affect. When the predicted conflict probability exceeds the preset threshold, the dynamic decoupling mechanism will be automatically triggered to instantaneously optimize the resources to avoid or reduce the occurrence of conflicts. Among them, the steps of performing fusion analysis on the time-axis feature vector and the spatial resource correlation matrix to predict the resource conflict probability and the conflict coverage within the demand interval include: Analyze the joint feature set, and establish a dynamic mapping relationship between the planned time fluctuation and the spatial node transmission loss to capture the changing trends of resource demands at different time points and spatial nodes; Adopt a dynamic time window to track and adjust the change trajectory of the time elasticity coefficient, and generate a conflict factor in the time dimension in combination with the time decay trend of sudden emergency events; Obtain the supply capacity gradient difference between geographical nodes and the physical distance weight between geographical nodes. Among them, the greater the physical distance, the smaller the physical distance weight; Under the demand interval, combine the supply capacity gradient difference and the physical distance weight to calculate the conflict diffusion intensity value in the spatial dimension, and then perform weighted superposition of the conflict diffusion intensity value and the conflict factor in the time dimension to output the predicted value of the resource conflict probability and generate a predicted map of the conflict coverage range in units of geographical grids; According to the topological connection relationship between geographical nodes, perform connectivity verification on the over-limit geographical grids in the predicted map of the conflict coverage range. After the connectivity verification is passed, mark the over-limit geographical grids as potential conflict areas. Otherwise, mark the over-limit geographical grids as isolated conflict points; Specifically, when performing fusion analysis on the time-axis feature vector and the spatial resource correlation matrix, it is first necessary to analyze the joint feature set. The joint feature set contains the normalized processing results of the time-axis feature vector and the spatial resource correlation matrix, reflecting the time characteristics and spatial correlation characteristics of resource demands. By establishing a dynamic mapping relationship between the planned time fluctuation and the spatial node transmission loss, the changing trends of resource demands at different time points and spatial nodes can be captured, providing a corresponding basis for subsequent resource scheduling. Then, use the dynamic time window technology to track and adjust the change trajectory of the time elasticity coefficient. The elasticity coefficient reflects the flexibility and adaptability in the resource scheduling process. Combining the time decay trend of sudden emergency events, a conflict factor in the time dimension can be generated ( , where represents the conflict factor in the time dimension, represents the time elasticity coefficient for adjustment, represents the time decay factor of sudden emergency events. Among them, , represents the time decay rate, represents the time span), the conflict factor is used to quantify the degree of conflict of resource requirements over time, while obtaining the supply capacity gradient difference and physical distance weight between geographical nodes. The supply capacity gradient difference reflects the difference in resource supply capacity between different geographical nodes, and the physical distance weight takes into account the impact of geographical distance on the efficiency of resource sharing. The greater the physical distance, the greater the difficulty and loss of resource sharing usually are. Under the demand interval, by combining the supply capacity gradient difference and the physical distance weight, the conflict diffusion intensity value in the spatial dimension can be calculated (conflict diffusion intensity value = supply capacity gradient difference × physical distance weight). Then, the conflict diffusion intensity value is weighted and superimposed with the conflict factor in the time dimension, and thus the predicted value of the resource conflict probability can be output. At the same time, according to the actual connection situation between geographical nodes, a conflict coverage prediction map is generated with geographical grids as the unit. The conflict coverage prediction map can visually display the geographical areas that may be affected by resource conflicts, so as to quickly identify potential conflict areas. Finally, according to the topological connection relationship between geographical nodes, the connectivity verification of the over-limit geographical grids in the conflict coverage prediction map is carried out. The connectivity verification is used to determine whether the over-limit geographical grids form a connected conflict area. If a connected conflict area is formed, the over-limit geographical grids are marked as potential conflict areas, indicating that the resource conflicts in these areas are relatively serious and need to be focused on and scheduled preferentially. Otherwise, the over-limit geographical grids are marked as isolated conflict points, indicating that the resource conflicts in the corresponding areas are relatively independent and have a limited influence range.

[0022] Secondly, the steps for carrying out the connectivity verification of the over-limit geographical grids in the conflict coverage prediction map include: Traverse the topological connection relationship between geographical nodes, and extract all physical paths that are directly and indirectly connected to the over-limit geographical grids, and mark them as paths to be verified; Based on the resource transmission efficiency and transmission loss of the current path, calculate the total resource transmission time and the total resource loss value on each path to be verified; Sort the paths to be verified in ascending order according to the total resource transmission time, and based on the sorting result, select the first several paths to be verified with the smallest total resource transmission time as the candidate path set in combination with the preset candidate path quantity threshold; Accumulate the total resource loss values of the candidate paths to obtain the total loss value of the candidate path set, and compare the total loss value with the preset loss threshold; If the total loss value is less than the loss threshold, it is determined that the over-limit geographical grids pass the connectivity verification, have the potential resource conflict propagation ability, and the over-limit geographical grids are included in the potential conflict area; If the total loss value is greater than or equal to the loss threshold, it is determined that the over-limit geographic grid fails the connectivity check, indicating that the resource conflict influence range of the corresponding over-limit geographic grid is limited, and it is marked as an isolated conflict point; In this embodiment, when performing the connectivity check of the over-limit geographic grid, first traverse the topological connection relationship between geographic nodes to find all physical paths associated with the over-limit geographic grid and record them as paths to be checked. The paths to be checked include both direct connections and indirect connections, jointly constituting potential channels for resource transmission. Then, based on the resource transmission efficiency and transmission loss of each path to be checked, calculate the total transmission time and total resource loss value of the resource on the path to be checked. Then, sort all paths to be checked according to the total transmission time of the resource to select the path to be checked with the highest transmission efficiency. Combining the preset candidate path quantity threshold, select the first several paths with the smallest time consumption from the sorting result as the candidate path set. After that, accumulate the total resource loss values in the candidate path set to obtain the total loss value, and compare it with the preset loss threshold to evaluate the overall loss situation of the candidate path set during resource transmission to determine whether it meets the requirements of resource scheduling. If the total loss value is less than the loss threshold, it indicates that the candidate path set performs well in terms of both resource transmission efficiency and loss. Therefore, it is determined that the over-limit geographic grid passes the connectivity check, has the potential for resource conflict propagation, and is included in the potential conflict area. Conversely, if the total loss value is greater than or equal to the loss threshold, it indicates that the loss of the candidate path set during resource transmission is too large and does not meet the requirements of resource scheduling. Therefore, it is determined that the over-limit geographic grid fails the connectivity check, indicating that its resource conflict influence range is limited, and it is marked as an isolated conflict point to reduce unnecessary resource scheduling and conflict handling costs.

[0023] Next, when the resource conflict probability exceeds the preset evaluation threshold, trigger the steps of the dynamic decoupling mechanism for resource optimization, including: Perform multi-level resource unbinding on the potential conflict area, calculate the spatio-temporal adaptability weights of each sub-link by decomposing the resource dependence chain between regions; Perform local resource adjustment on the isolated conflict point and reallocate the resource load within each geographic node; Determine the priorities of resource unbinding and local resource adjustment according to the spatio-temporal adaptability weights of each sub-link and the reallocated resource load, and gradually implement resource unbinding and local resource adjustment according to the priority order; Real-time monitor the progress of resource unbinding and the real-time load changes of local resource adjustment, and trigger the resource reallocation process when the resource unbinding progress reaches the preset unbinding ratio or the local resource adjustment reaches the load balancing state; Specifically, after the decoupling mechanism is triggered, a multi-level resource decoupling strategy is first implemented for potential conflict areas. The multi-level resource decoupling strategy lies in decomposing the resource dependency chain between regions, identifying each sub-link, and calculating its adaptation weight in terms of time and space. The spatio-temporal adaptation weight is used to measure the flow efficiency and effect of resources on different sub-links. For isolated conflict points, a local resource adjustment strategy is adopted. The local resource adjustment strategy focuses on reallocating the resource load within each geographical node to alleviate local resource conflicts and optimize resource allocation. Through local resource adjustment, the resource load within each geographical node becomes more balanced, reducing the possibility of resource conflicts. When determining the priorities of resource decoupling and local resource adjustment, it is necessary to comprehensively consider the spatio-temporal adaptation weights of each sub-link and the resource load after reallocation, determine the priority order, and gradually implement resource decoupling and local resource adjustment according to this priority order to ensure the orderliness and effectiveness of resource optimization. During the implementation of resource decoupling and local resource adjustment, it is also necessary to monitor the progress of resource decoupling and the real-time load changes of local resource adjustment in real time, so as to be able to timely discover and solve problems that occur during resource optimization and ensure the smooth progress of resource optimization. And when the resource decoupling progress reaches the preset decoupling ratio or the local resource adjustment reaches the load balance state, a resource reallocation process is triggered to achieve further optimized allocation of resources.

[0024] S3. Generate a resource allocation strategy corresponding to the business urgency and spatio-temporal adaptation degree through decoupling calculation based on time weight and space weight; In the step S3, during the resource optimization process, a decoupling calculation strategy based on time weight and space weight is used to determine the optimal resource allocation for business urgency and spatio-temporal adaptation degree, so as to ensure that the priority of critical services is guaranteed while maximizing the spatio-temporal utilization rate of resources. Among them, the step of generating a resource allocation strategy corresponding to the business urgency and spatio-temporal adaptation degree through decoupling calculation based on time weight and space weight includes: Calculate the allocation ratio of time weight and space weight according to the resource decoupling progress and the resource load after local resource adjustment; Classify the resource requirements according to the business urgency, and combine the allocation ratio of time weight and space weight to generate a resource allocation priority list under different business urgencies; Evaluate the resource allocation priority using spatio-temporal adaptation degree, determine the adaptation degree of resources at different times and spaces, and obtain the matching degree score between resource supply and demand; Compare the matching degree score with a preset matching degree threshold. If the matching degree score is higher than the matching degree threshold, confirm that the resource allocation strategy is effective, and continue to execute resource decoupling and local resource adjustment; If the matching degree score is lower than or equal to the matching degree threshold, recalculate the allocation ratios of the time weight and the space weight until the matching degree score is higher than the matching degree threshold; Specifically, during the resource optimization process, first calculate the allocation ratios of the time weight and the space weight according to the resource unbinding progress and the resource load situation after local resource adjustment (time weight = resource unbinding progress / (resource unbinding progress + load balancing index after local resource adjustment), space weight = load balancing index after local resource adjustment / (resource unbinding progress + load balancing index after local resource adjustment)). The time weight reflects the urgency of resource requirements in different time periods, while the space weight reflects the demand distribution of resources in different geographical regions. Then, classify the resource requirements according to the business urgency. Different businesses have different demand priorities for resources due to differences in aspects such as importance, urgency, and resource demand. Therefore, in this embodiment, combine the allocation ratios of the time weight and the space weight to generate a resource allocation priority list under different business urgencies to ensure that critical businesses can be preferentially guaranteed during the resource allocation process. After obtaining the resource allocation priority list, evaluate the resource allocation priority using the spatio-temporal fitness degree. The spatio-temporal fitness degree reflects the adaptation degree of resources in different times and spaces and can determine the matching degree score between resource supply and demand (matching degree score = (resource supply quantity - resource demand quantity) / resource demand quantity × 100%). The higher the score, the better the matching degree between resource supply and demand). By comparing the matching degree score with the preset matching degree threshold, the effectiveness of the resource allocation strategy can be judged. If the matching degree score is higher than the matching degree threshold, it indicates that the resource allocation strategy is reasonable and effective, and resource unbinding and local resource adjustment can be continued. If the matching degree score is lower than or equal to the matching degree threshold, it is necessary to recalculate the allocation ratios of the time weight and the space weight and adjust the resource allocation strategy until the matching degree score is higher than the matching degree threshold. Through this resource allocation strategy, not only can the priority of critical businesses be guaranteed, but also the spatio-temporal utilization rate of resources can be improved to meet the changing business needs.

[0025] S4. Implement standby resource scheduling and cross-region coordination for resource conflict events under the execution of the resource allocation strategy, and synchronously generate a resource scheduling execution report. Then, count the scheduling success rate and scheduling delay rate of standby resource scheduling and cross-region coordination, and conduct a resource scheduling effect evaluation to output quantitative resource scheduling effect indicators; In step S4, after the resource allocation policy is executed, standby resource scheduling and cross-region coordination are implemented to handle possible conflict events. Meanwhile, a detailed resource scheduling execution report is generated. The resource scheduling execution report records the success rate and delay rate of scheduling, providing a quantitative basis for evaluating the resource scheduling effect. Among them, the steps of counting the success rate and scheduling delay rate of standby resource scheduling and cross-region coordination, evaluating the resource scheduling effect, and outputting the quantitative resource scheduling effect indicators include: Capture resource scheduling requests in real time and record the start time and end time of resource scheduling; Calculate the actual delay time of each resource scheduling and calculate the scheduling delay rate based on the actual delay time; Count the number of successfully completed resource schedulings and the number of failed schedulings within a preset time period, and calculate the scheduling success rate; Conduct a dual evaluation of the scheduling success rate and the scheduling delay rate: The first evaluation is an independent evaluation. The scheduling success rate and the scheduling delay rate are respectively compared with the preset scheduling success rate threshold and the scheduling delay rate threshold. When the scheduling success rate does not meet the scheduling success rate threshold or the scheduling delay rate does not meet the success rate threshold, a warning mechanism is triggered, a scheduling exception report is generated, and it is marked as an item to be optimized; The second evaluation is a comprehensive evaluation. The scheduling success rate and the scheduling delay rate are weighted and summed to obtain a comprehensive evaluation score of the resource scheduling effect. The comprehensive evaluation score is compared with the preset comprehensive evaluation score threshold. When the comprehensive evaluation score is lower than the comprehensive evaluation score threshold, a warning mechanism is triggered, a scheduling exception report is generated, and it is marked as an item to be optimized; Specifically, during the execution of the resource allocation strategy, it is necessary to capture resource scheduling requests in real time to ensure quick response and handling of resource requirements. At the same time, the start time and end time of resource scheduling will be recorded, which not only helps calculate the scheduling delay rate but also provides basic data for subsequent evaluation of scheduling effects. By statistically analyzing the actual delay time of each resource scheduling, the scheduling delay rate can be obtained to reflect the efficiency of resource scheduling. Meanwhile, by counting the number of successfully completed resource scheduling times and the number of failed scheduling times, the scheduling success rate can be calculated to reflect the reliability of resource scheduling. To comprehensively evaluate the resource scheduling effect, a dual evaluation mechanism is adopted. The first evaluation is an independent evaluation, where the scheduling success rate and the scheduling delay rate are respectively compared with the preset scheduling success rate threshold and scheduling delay rate threshold. If the scheduling success rate is lower than the preset scheduling success rate threshold or the scheduling delay rate is higher than the preset scheduling delay threshold, the warning mechanism will be triggered, and a scheduling exception report will be generated and marked as an item to be optimized for recording, so as to promptly discover and solve problems in the resource scheduling process. The second evaluation is a comprehensive evaluation. Specifically, the scheduling success rate and the scheduling delay rate are weighted and summed to obtain a comprehensive evaluation score of the resource scheduling effect. The comprehensive evaluation score can more comprehensively reflect the overall effect of resource scheduling. If the comprehensive evaluation score is lower than the preset comprehensive evaluation score threshold, the warning mechanism will also be triggered, a scheduling exception report will be generated, and it will be marked as an item to be optimized.

[0026] S5. Feed back the resource scheduling effect indicators to the construction process of the time-axis feature vector and the spatial resource correlation matrix, and synchronously update the time-axis feature vector and the spatial resource correlation matrix to form a closed-loop optimization mechanism.

[0027] In step S5, the evaluation result of the resource scheduling effect will be directly fed back to the initial stage of resource management, that is, the construction process of the time-axis feature vector and the spatial resource correlation matrix. According to the evaluation result, corresponding parameter adjustments and model updates will be carried out to form a self-learning and optimizing closed-loop mechanism to ensure the continuous improvement and adaptability of resource management. Among them, the steps of feeding back the resource scheduling effect indicators to the construction process of the time-axis feature vector and the spatial resource correlation matrix and synchronously updating the time-axis feature vector and the spatial resource correlation matrix include: Collect the scheduling success rate and the scheduling delay rate in the resource scheduling indicators and map them to the correction term of the burst emergency degree weight of the time-axis feature vector and the attenuation coefficient of the transmission efficiency of the spatial resource correlation matrix; Collect the deviation value between the scheduling success rate and the scheduling success rate threshold and record it as the scheduling success rate deviation; Based on the scheduling success rate deviation, dynamically adjust the weight ratio of the planned time to the adjusted time in the time-axis feature vector and recalculate the time attenuation rate of the burst emergency degree; Based on the scheduling delay rate, the physical distance weight in the spatial resource correlation matrix is corrected inversely to enhance the load allocation priority of low-latency geographical nodes.

[0028] In this embodiment, after the evaluation result feedback of the resource scheduling effect, the time-axis feature vector and the spatial resource correlation matrix will be further optimized adaptively. First, the scheduling success rate and the scheduling delay rate in the resource scheduling are collected and mapped to the burst emergency weight correction term of the time-axis feature vector and the transmission efficiency attenuation coefficient of the spatial resource correlation matrix. At the same time, the deviation value between the scheduling success rate and the scheduling success rate threshold, that is, the scheduling success rate deviation, will be calculated to determine the gap between the current scheduling effect and the expected target. Based on the scheduling success rate deviation, the weight ratio of the planned time to the adjusted time in the time-axis feature vector will be dynamically adjusted (specifically, the weight of the planned time can be increased to reduce the overreaction to the urgency of resource scheduling, or the weight of the adjusted time can be increased to respond more quickly to the change of resource demand) to flexibly cope with the change of resource demand. Then, the time decay rate of the burst emergency will be recalculated to ensure that the resources can be allocated in a timely and effective manner according to the priority. In addition, based on the scheduling delay rate, the physical distance weight in the spatial resource correlation matrix will be corrected inversely, that is, the load allocation priority of low-latency geographical nodes will be enhanced to reduce the delay of resource transmission, improve the efficiency of resource scheduling, and ensure the stable operation of key services.

[0029] Please refer to Figure 2 , an online resource management system based on the AI large model, using the above-mentioned online resource management method based on the AI large model, including: A data extraction module for separating the spatio-temporal features in the online resource scenario, constructing a time-axis feature vector including planned time, adjusted time, and burst emergency, and quantifying the spatial resource correlation matrix of the resource sharing ability between different regions; A resource optimization module for performing fusion analysis on the time-axis feature vector and the spatial resource correlation matrix, predicting the resource conflict probability and the conflict coverage range within the demand interval, and triggering a dynamic decoupling mechanism for resource optimization when the resource conflict probability exceeds the preset evaluation threshold; A resource allocation module for generating a resource allocation strategy corresponding to the service emergency degree and the spatio-temporal adaptability based on the decoupling calculation of the time weight and the spatial weight; A resource coordination module for implementing backup resource scheduling and cross-region coordination for resource conflict events under the execution of the resource allocation strategy, synchronously generating a resource scheduling execution report, then counting the scheduling success rate and the scheduling delay rate of the backup resource scheduling and the cross-region coordination, and performing resource scheduling effect evaluation to output a quantified resource scheduling effect index; A feedback optimization module is used to feedback the resource scheduling effect indicators to the construction processes of the time-axis feature vector and the spatial resource correlation matrix, and synchronously update the time-axis feature vector and the spatial resource correlation matrix to form a closed-loop optimization mechanism.

[0030] Among the above, the data extraction module is responsible for separating key spatio-temporal features from the online resource scenario. The spatio-temporal features include the planned time, adjustment time, and sudden urgency, which together constitute the time-axis feature vector. The spatio-temporal feature vector can dynamically reflect the changing trend and urgency of resource demands. At the same time, it will also quantify the resource sharing capabilities between different regions, construct the spatial resource correlation matrix, and reveal the complementarity and dependency relationships of resources between regions, providing basic data for subsequent resource optimization and coordination. The resource optimization module will conduct in-depth fusion analysis on the time-axis feature vector and the spatial resource correlation matrix, and predict the possible resource conflict probability and the coverage of conflicts within the demand interval. When the predicted resource conflict probability exceeds the preset evaluation threshold, it will immediately trigger the dynamic decoupling mechanism to quickly identify and separate the conflicting resources. The resource allocation module will generate a resource allocation strategy corresponding to the business urgency and spatio-temporal adaptability based on the decoupling calculation of time weights and spatial weights to ensure the reasonable allocation of resources in terms of time and space and meet the urgent needs of the business. The resource coordination module will promptly respond to and handle resource conflict events occurring during the execution of the resource allocation strategy. By implementing backup resource scheduling and cross-region coordination, it ensures the effective supply of resources and the continuous operation of the business. During the resource coordination process, the system will synchronously generate a resource scheduling execution report, recording the execution situations of backup resource scheduling and cross-region coordination, including the types, quantities, times of the scheduled resources, and the scheduling results, etc., for subsequent statistics and analysis. After the statistics are completed, it will count the scheduling success rate and scheduling delay rate of backup resource scheduling and cross-region coordination, and based on this, evaluate the resource scheduling effect, quantify the evaluation result into specific resource scheduling effect indicators to reflect the efficiency and quality of resource scheduling. The feedback optimization module will feedback the quantified resource scheduling effect indicators as reference bases to the construction processes of the time-axis feature vector and the spatial resource correlation matrix. By continuously adjusting and optimizing the weights of the planned time, adjustment time, and sudden urgency in the time-axis feature vector, as well as the physical distance weight and transmission efficiency attenuation coefficient in the spatial resource correlation matrix, the entire resource management system can more accurately predict the changes in resource demands, effectively avoid resource conflicts, improve the efficiency and accuracy of resource scheduling, and form a closed-loop mechanism for continuously improving and optimizing resource management.

[0031] Please refer to Figure 3 , an electronic device, the electronic device includes: At least one processor; and a memory communicatively connected to at least one processor; wherein the memory stores a computer program executable by at least one processor, and the computer program is executed by at least one processor to enable the at least one processor to execute the above-mentioned online resource management method based on the AI large model.

[0032] The processor of the above electronic device can be a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), etc., or multiple processors can be used in combination to improve data processing capabilities and operating efficiency. The memory can be a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), or a portable read-only memory (CD-ROM), etc., for storing computer programs and data to ensure the normal operation of the electronic device. In addition, the electronic device can also include an arithmetic unit, an input device, an output device, and a network interface, etc. The arithmetic unit can be an arithmetic logic unit (ALU) for performing various arithmetic and logical operations; the input device can include a keyboard, a mouse, a touch screen, etc., for receiving user operation instructions and data input; the output device can include a display, a printer, etc., for presenting processing results and outputting data; the network interface is used to connect to an external network to achieve remote transmission and sharing of data.

[0033] It should be noted that in this article, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article or method comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article or method. Without further limitation, an element defined by the phrase "comprising a..." does not exclude the presence of additional identical elements in the process, apparatus, article or method comprising that element.

[0034] The above are only the preferred embodiments of the present invention. It should be pointed out that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention. Structures, devices, and operation methods not specifically described and explained in the present invention are implemented according to conventional means in the art without special description and limitation.

Claims

1. An online resource management method based on an AI large model, characterized in that: Including: Separate the spatio-temporal features in the online resource scenario, construct a time-axis feature vector including planned time, adjustment time, and sudden emergency degree, and quantify the spatial resource correlation matrix of the resource sharing ability between different regions; Conduct a fusion analysis on the time-axis feature vector and the spatial resource correlation matrix to predict the resource conflict probability and conflict coverage range within the demand interval, and trigger a dynamic decoupling mechanism for resource optimization when the resource conflict probability exceeds the preset evaluation threshold; Generate a resource allocation strategy corresponding to the service emergency degree and spatio-temporal adaptability through decoupling calculation based on time weight and spatial weight; Implement spare resource scheduling and cross-region coordination for resource conflict events under the execution of the resource allocation strategy, and synchronously generate a resource scheduling execution report. Then, statistically calculate the scheduling success rate and scheduling delay rate of spare resource scheduling and cross-region coordination, and conduct a resource scheduling effect evaluation to output quantified resource scheduling effect indicators; Feed back the resource scheduling effect indicators to the construction process of the time-axis feature vector and the spatial resource correlation matrix, and synchronously update the time-axis feature vector and the spatial resource correlation matrix to form a closed-loop optimization mechanism.

2. An online resource management method based on an AI large model according to claim 1, characterized in that: The step of separating the spatio-temporal features in the online resource scenario, constructing a time-axis feature vector including planned time, adjustment time, and sudden emergency degree, and quantifying the spatial resource correlation matrix of the resource sharing ability between different regions includes: Obtain historical resource scheduling logs, and perform time series segmentation and geographical grid division on the historical resource scheduling logs to obtain historical resource scheduling time series data and geographical grid data; Conduct a time series correlation analysis on the time series data and the geographical grid data, extract the time features of planned time, adjustment time, and sudden emergency degree, and fuse them into a time-axis feature vector; Obtain the computing resource transmission efficiency and transmission loss between different geographical nodes, as well as the physical distance and topological connection relationship between geographical nodes, and establish a spatial resource correlation matrix including the supply capacity, response speed, and load status between geographical nodes; Normalize the time-axis feature vector and the spatial resource correlation matrix, and integrate them into a joint feature set; Among them, the time series correlation analysis uses a dynamic window to capture the repeated pattern of the planned time, the elastic change range of the adjustment time, and the trend of the attenuation of sudden emergency events over time.

3. An online resource management method based on an AI large model according to claim 2, characterized in that: The step of conducting a fusion analysis on the time-axis feature vector and the spatial resource correlation matrix to predict the resource conflict probability and conflict coverage range within the demand interval includes: Analyze the joint feature set, and establish a dynamic mapping relationship between the planned time fluctuation and the spatial node transmission loss to capture the change trend of resource demand at different time points and spatial nodes; Adopt a dynamic time window to track the change trajectory of the elastic coefficient of the adjustment time, and generate a conflict factor in the time dimension in combination with the time attenuation trend of sudden emergency events; Obtain the supply capacity gradient difference between geographical nodes and the physical distance weight between geographical nodes. Among them, the greater the physical distance, the smaller the physical distance weight; Under the demand interval, combining the supply capacity gradient difference and the physical distance weight, calculate the conflict diffusion intensity value in the spatial dimension, and then perform weighted superposition of the conflict diffusion intensity value and the conflict factor in the time dimension to output the predicted value of the resource conflict probability and generate a predicted map of the conflict coverage area in units of geographical grids; According to the topological connection relationship between geographical nodes, perform connectivity verification on the over-limit geographical grids in the predicted map of the conflict coverage area. After the connectivity verification passes, mark the over-limit geographical grids as potential conflict areas. Otherwise, mark the over-limit geographical grids as isolated conflict points.

4. An online resource management method based on an AI large model according to claim 3, characterized in that: The steps of performing connectivity verification on the over-limit geographical grids in the predicted map of the conflict coverage area include: Traverse the topological connection relationship between geographical nodes, extract all physical paths that are directly and indirectly connected to the over-limit geographical grids, and mark them as paths to be verified; Based on the resource transmission efficiency and transmission loss of the current path, calculate the total resource transmission time and the total resource loss value on each path to be verified; Sort the paths to be verified in ascending order according to the total resource transmission time, and based on the sorting result, select the first several paths to be verified with the smallest total resource transmission time as the candidate path set in combination with the preset candidate path quantity threshold; Accumulate the total resource loss values of the candidate paths to obtain the total loss value of the candidate path set, and compare the total loss value with the preset loss threshold; If the total loss value is less than the loss threshold, it is determined that the over-limit geographical grid passes the connectivity verification, has the potential for resource conflict propagation, and the over-limit geographical grid is included in the potential conflict area; If the total loss value is greater than or equal to the loss threshold, it is determined that the over-limit geographical grid fails the connectivity verification, indicating that the resource conflict influence range of the corresponding over-limit geographical grid is limited, and it is marked as an isolated conflict point.

5. An online resource management method based on an AI large model according to claim 3, characterized in that: When the resource conflict probability exceeds the preset evaluation threshold, the steps of triggering the dynamic decoupling mechanism for resource optimization include: Adopt multi-level resource unbinding for the potential conflict area, calculate the spatio-temporal adaptability weights of each sub-link by decomposing the resource dependence chain between regions; Adopt local resource adjustment for the isolated conflict points to reallocate the resource loads within each geographical node; Determine the priorities of resource unbinding and local resource adjustment according to the spatio-temporal adaptability weights of each sub-link and the reallocated resource loads, and gradually implement resource unbinding and local resource adjustment according to the priority order; Real-time monitor the progress of resource unbinding and the real-time load changes of local resource adjustment, and trigger the resource reallocation process when the resource unbinding progress reaches the preset unbinding ratio or the local resource adjustment reaches the load balancing state.

6. An online resource management method based on an AI large model according to claim 5, characterized in that: The steps of generating a resource allocation strategy corresponding to the service urgency and spatio-temporal adaptability based on the decoupling calculation of time weight and space weight include: Calculate the allocation ratios of the time weight and the space weight according to the resource unbinding progress and the resource loads after local resource adjustment; Classify the resource requirements according to the service urgency, and generate a resource allocation priority list under different service urgencies in combination with the allocation ratios of the time weight and the space weight; Evaluate the resource allocation priority using the spatio-temporal fitness degree to determine the fitness degree of resources at different times and in different spaces, and obtain the matching degree score between resource supply and demand; Compare the matching degree score with a preset matching degree threshold. If the matching degree score is higher than the matching degree threshold, confirm that the resource allocation strategy is effective, and continue to execute resource unbinding and local resource adjustment; If the matching degree score is lower than or equal to the matching degree threshold, recalculate the allocation ratio of the time weight and the space weight until the matching degree score is higher than the matching degree threshold.

7. The online resource management method based on the AI large model according to claim 1, wherein: The steps of statistically calculating the scheduling success rate and scheduling delay rate of the standby resource scheduling and cross-regional coordination, evaluating the resource scheduling effect, and outputting the quantified resource scheduling effect indicators include: Capture resource scheduling requests in real time, and record the start time and end time of resource scheduling; Calculate the actual delay time of each resource scheduling, and calculate the scheduling delay rate based on the actual delay time; Statistically count the number of successfully completed resource schedulings and the number of failed schedulings within a preset time period, and calculate the scheduling success rate; Conduct a dual evaluation of the scheduling success rate and the scheduling delay rate: The first evaluation is an independent evaluation. Compare the scheduling success rate and the scheduling delay rate with the preset scheduling success rate threshold and scheduling delay rate threshold respectively. When the scheduling success rate does not meet the scheduling success rate threshold or the scheduling delay rate does not meet the success rate threshold, trigger an early warning mechanism, generate a scheduling exception report, and mark it as an item to be optimized; The second evaluation is a comprehensive evaluation. Perform a weighted sum of the scheduling success rate and the scheduling delay rate to obtain the comprehensive evaluation score of the resource scheduling effect. Compare the comprehensive evaluation score with the preset comprehensive evaluation score threshold. When the comprehensive evaluation score is lower than the comprehensive evaluation score threshold, trigger an early warning mechanism, generate a scheduling exception report, and mark it as an item to be optimized.

8. An online resource management method based on an AI large model according to claim 7, characterized in that: The steps of feeding back the resource scheduling effect indicators to the construction process of the time axis feature vector and the spatial resource correlation matrix, and synchronously updating the time axis feature vector and the spatial resource correlation matrix include: Collect the scheduling success rate and the scheduling delay rate in the resource scheduling indicators, and map them to the sudden emergency degree weight correction term of the time axis feature vector and the transmission efficiency attenuation coefficient of the spatial resource correlation matrix; Collect the deviation value between the scheduling success rate and the scheduling success rate threshold, and record it as the scheduling success rate deviation; Based on the scheduling success rate deviation, dynamically adjust the weight ratio of the planned time and the adjusted time in the time axis feature vector, and recalculate the time decay rate of the sudden emergency degree; Based on the scheduling delay rate, perform reverse correction on the physical distance weight in the spatial resource correlation matrix to improve the load allocation priority of low-latency geographical nodes.

9. An online resource management system based on an AI large model, characterized in that: Use the online resource management method based on the AI large model described in any one of claims 1 to 8, including: A data extraction module for separating the spatio-temporal features in the online resource scenario, constructing a time axis feature vector including planned time, adjusted time, and sudden emergency degree, and a spatial resource correlation matrix quantifying the resource sharing ability between different regions; A resource optimization module, which is used to perform fusion analysis on the time-axis feature vector and the spatial resource correlation matrix, predict the resource conflict probability and the conflict coverage range within the demand interval, and trigger a dynamic decoupling mechanism for resource optimization when the resource conflict probability exceeds a preset evaluation threshold; A resource allocation module, which is used to generate a resource allocation strategy corresponding to the business urgency and the spatio-temporal adaptability based on the decoupling calculation of the time weight and the spatial weight; A resource coordination module, which implements backup resource scheduling and cross-region coordination for resource conflict events under the execution of the resource allocation strategy, synchronously generates a resource scheduling execution report, then statistically calculates the scheduling success rate and the scheduling delay rate of the backup resource scheduling and the cross-region coordination, and conducts a resource scheduling effect evaluation to output a quantitative resource scheduling effect index; A feedback optimization module, which is used to feedback the resource scheduling effect index to the construction process of the time-axis feature vector and the spatial resource correlation matrix, and synchronously update the time-axis feature vector and the spatial resource correlation matrix to form a closed-loop optimization mechanism.

10. An electronic device, characterized in that: The electronic device includes: At least one processor; And a memory communicatively connected to the at least one processor; Wherein, the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the online resource management method based on the AI large model according to any one of claims 1 to 8.

Images