Key technology-based 6G typical scene knowledge graph construction method and system

Abstract

The invention relates to the technical field of electric data processing, and particularly discloses a 6G typical scene knowledge graph construction method and system based on a key technology, and the method comprises the steps: obtaining each 6G typical scene, marking the 6G typical scene as each target scene, and collecting the reasoning task data and computing resource data of the knowledge graph of each target scene; performing comprehensive analysis to obtain a length limit threshold value of the reasoning chain of each target scene; and limiting the length of each target scene reasoning chain according to the length limiting threshold of each target scene reasoning chain, and performing reasoning effect evaluation on the limited length of the reasoning chain. According to the 6G typical scene knowledge graph construction method and system based on the key technology, the actual effect of the inference chain limit threshold value can be judged, further optimization is carried out, the inference accuracy is improved, meanwhile, the system can adjust the length and complexity of the inference chain according to the evaluation result, and the reasoning accuracy is improved. Therefore, the optimal reasoning effect can be obtained under the condition of limited resources.

Description

Technical Field

[0001] The present invention relates to the field of electrical data processing technology, and specifically to a method and system for constructing a 6G typical scenario knowledge graph based on key technologies. Background Art

[0002] At present, knowledge graph construction is a very important part in the field of data analysis. With the advancement of technology and the development of 6G, providing complete and accurate knowledge graphs has become the norm. Efficient and comprehensive knowledge graph construction methods are of great significance to improving user experience and promoting technological innovation.

[0003] For example, the invention patent with announcement number CN107491555B is a method and system for constructing a knowledge graph, including: establishing the basic architecture of the knowledge graph according to general data standards; standardizing the relationships between entities in the basic architecture to obtain a standard dictionary table with standard specifications; obtaining semi-structured data related to the content in the knowledge graph; extracting entity information of key entities from the semi-structured data; fusing the entity information according to the standard dictionary table to form structured data; generating corresponding data structure pairs from the structured data, and storing the data structure pairs as a knowledge graph.

[0004] For example, the invention patent with announcement number CN115186817A is a 6G full-scenario knowledge graph and its construction method. The construction method includes: analyzing the core elements of the 6G scenario and defining the basic architecture of the 6G full-scenario ontology; improving the components of the basic architecture and determining the knowledge graph entity types; defining different types of relationships according to the knowledge graph entity types to connect different types of entities; extracting the required entities and relationships from the acquired 6G full-scenario related data and storing them in the form of triples to construct a 6G full-scenario knowledge graph.

[0005] However, in the process of implementing the technical solution of the invention in the embodiments of the present application, it was found that the above technology has at least the following technical problems: the current knowledge graph construction method pays more attention to data diversity and construction efficiency, but 6G communication technology and its typical application scenarios have higher requirements for data processing. The traditional knowledge graph construction method ignores the effect and reliability of application in typical scenarios, which makes it difficult to adapt to the needs of specific fields. Summary of the invention

[0006] In view of the shortcomings of the prior art, the present invention provides a method and system for constructing a 6G typical scenario knowledge graph based on key technologies, which can effectively solve the problems involved in the above-mentioned background technology.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: The first aspect of the present invention provides a method for constructing a 6G typical scenario knowledge graph based on key technologies, including: obtaining each 6G typical scenario, marking it as each target scenario, and collecting reasoning task data and computing resource data of each target scenario knowledge graph.

[0008] The reasoning task data of each target scenario is processed to obtain the complexity evaluation index of the reasoning task of each target scenario, the computing resource data of each target scenario is analyzed to obtain the computing resource evaluation index of each target scenario, and the reasoning chain length limit threshold of each target scenario is obtained through comprehensive analysis.

[0009] The length of the reasoning chain of each target scene is limited according to the limit threshold of the reasoning chain length of each target scene, and the reasoning effect of the limited reasoning chain length is evaluated to obtain the evaluation result and provide feedback.

[0010] As a further method, the reasoning task data of each target scene is processed to obtain the complexity evaluation index of the reasoning task of each target scene. The specific processing process is: the reasoning task data includes the number of reasoning task targets, the estimated task execution time and the estimated resource requirement number.

[0011] The number of critical reasoning task targets, critical estimated task execution time and critical estimated resource requirements are extracted from the 6G key technology database, and a comprehensive analysis is performed to obtain the complexity evaluation index of the reasoning task for each target scenario.

[0012] As a further method, the computing resource data of each target scene is analyzed to obtain a computing resource evaluation index of each target scene. The specific analysis process is as follows: the computing resource data includes CPU processing speed, memory usage and disk throughput.

[0013] The critical CPU processing speed, critical memory usage and critical disk throughput are extracted from the 6G key technology database, and the computing resource evaluation index of each target scenario is obtained through comprehensive analysis.

[0014] As a further method, the comprehensive analysis obtains the limit threshold of the reasoning chain length of each target scene. The specific analysis process is: according to the complexity evaluation index of the reasoning task of each target scene and the computing resource evaluation index of each target scene, the limit threshold of the reasoning chain length of each target scene is obtained through comprehensive analysis.

[0015] As a further method, the length of the reasoning chain of each target scene is limited according to the limit threshold of the reasoning chain length of each target scene. The specific restriction process is: extract the preset reasoning chain length of each target scene from the 6G key technology database, compare the reasoning chain length of each target scene with the corresponding limit threshold of the reasoning chain length of the target scene, if the reasoning chain length of a certain target scene is less than or equal to the limit threshold of the reasoning chain length of the target scene, then the reasoning chain length of the target scene is not restricted, if the reasoning chain length of a certain target scene is greater than the limit threshold of the reasoning chain length of the target scene, then the limit threshold of the reasoning chain length of the target scene is marked as the reasoning chain length of each target scene.

[0016] As a further method, the reasoning effect of the restricted reasoning chain length is evaluated. The specific analysis process is: deploy a number of test samples to perform reasoning tests on each target scenario, and obtain the total number of test samples and the number of correct reasoning samples for each target scenario, and divide the correct reasoning samples for each target scenario by the total number of test samples to obtain the reasoning accuracy of each target scenario.

[0017] The reasoning accuracy of each target scene before and after the reasoning chain restriction is obtained, and the effectiveness evaluation value of the reasoning chain restriction is obtained through comprehensive analysis.

[0018] As a further method, the evaluation results are obtained and feedback is provided, and the specific feedback process is: extracting the reasoning chain restriction validity evaluation threshold from the 6G key technology database, comparing the reasoning chain restriction validity evaluation value with the reasoning chain restriction validity evaluation threshold, if the reasoning chain restriction validity evaluation value is greater than or equal to the reasoning chain restriction validity evaluation threshold, retaining the reasoning chain length after restriction without performing additional operations, if the reasoning chain restriction validity evaluation value is less than the reasoning chain restriction validity evaluation threshold, adjusting the reasoning chain length until the reasoning chain restriction validity evaluation value is greater than or equal to the reasoning chain restriction validity evaluation threshold.

[0019] As a further method, the reasoning chain length is adjusted, and the specific adjustment process is: extracting the preset reasoning chain adjustment length from the 6G key technology database, extending the reasoning chain length of each target scenario in turn according to the reasoning chain adjustment length, and calculating the reasoning chain restriction validity evaluation value. When the reasoning chain restriction validity evaluation value is greater than or equal to the reasoning chain restriction validity evaluation threshold, stop the adjustment operation.

[0020] As a further method, each target scene calculates a resource evaluation index, and the specific numerical expression is:

[0021]

[0022] Among them, A irepresents the computing resource evaluation index of the i-th target scenario, e represents a natural constant, and s i Indicates the CPU processing speed of the i-th target scene, s 0 Indicates the set critical CPU processing speed, C i represents the memory usage of the i-th target scene, C 0 Indicates the set critical memory usage, L i represents the disk throughput of the i-th target scenario, L 0 represents the set critical disk throughput, ω 1 Indicates the computing resource evaluation impact factor corresponding to the set CPU processing speed, ω 2 Indicates the computing resource evaluation impact factor corresponding to the set memory usage, ω 3 It represents the computing resource evaluation impact factor corresponding to the set disk throughput, i represents each target scenario, i=1,2,3,...,m, and m represents the total number of target scenarios.

[0023] The second aspect of the present invention provides a 6G typical scenario knowledge graph construction system based on key technologies, including: a knowledge graph data acquisition module, used to obtain each 6G typical scenario, mark it as each target scene, and collect reasoning task data and computing resource data of each target scene knowledge graph.

[0024] The reasoning chain length limit threshold analysis module is used to process the reasoning task data of each target scene to obtain the complexity evaluation index of the reasoning task of each target scene, analyze the computing resource data of each target scene to obtain the computing resource evaluation index of each target scene, and comprehensively analyze to obtain the reasoning chain length limit threshold of each target scene.

[0025] The reasoning effect evaluation module is used to limit the reasoning chain length of each target scene according to the reasoning chain length limit threshold of each target scene, and evaluate the reasoning effect of the limited reasoning chain length, obtain the evaluation result and provide feedback.

[0026] The 6G key technology database is used to store data related to 6G key technologies, including the critical number of reasoning task targets, the critical expected task execution time, the critical expected resource requirements, the critical CPU processing speed, the critical memory usage rate, the critical disk throughput, the reasoning chain limit effectiveness evaluation threshold, the reasoning chain length limit threshold evaluation impact factor corresponding to the set computing resource evaluation index, the reasoning chain length limit threshold evaluation impact factor corresponding to the set reasoning task complexity evaluation index, the computing resource evaluation impact factor corresponding to the set CPU processing speed, the computing resource evaluation impact factor corresponding to the set memory usage rate, the computing resource evaluation impact factor corresponding to the set disk throughput, the reasoning task complexity evaluation impact factor corresponding to the set number of reasoning task targets, the reasoning task complexity evaluation impact factor corresponding to the set expected task execution time, the reasoning task complexity evaluation impact factor corresponding to the set expected resource requirements, and the reasoning chain length of each target scenario.

[0027] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects:

[0028] (1) The present invention provides a 6G typical scenario knowledge graph construction method and system based on key technologies, which helps to judge the actual effect of the reasoning chain limit threshold and further optimize it, thereby improving the accuracy of reasoning. At the same time, the system can adjust the length and complexity of the reasoning chain according to the evaluation results to ensure the best reasoning effect with limited resources.

[0029] (2) By evaluating the complexity of the reasoning task, the present invention can make the system more flexible and can adjust the reasoning rules, optimize the algorithm and expand the resources as needed. It also helps to reduce the risk of system overload and crash, improve the stability and reliability of the system, help quickly locate the cause of the problem and take corresponding recovery measures to ensure business continuity.

[0030] (3) By evaluating computing resources, the present invention can accurately reflect the processing capacity of the system under the current hardware and configuration, thereby more accurately setting the limit threshold of the inference chain length, which helps to balance the resource load between different tasks within the system, ensure that the system can still run stably and provide high-quality services under high load, and further improve user satisfaction and loyalty. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The present invention is further described using the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.

[0032] Figure 1 It is a schematic diagram of the method flow of the present invention.

[0033] Figure 2 It is a schematic diagram of system module connection of the present invention.

[0034] Figure 3 It is a schematic diagram of the functional relationship between the reasoning chain length limit threshold and the reasoning task complexity evaluation index of the present invention. DETAILED DESCRIPTION

[0035] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0036] Reference Figure 1 As shown, the first aspect of the present invention provides a method for constructing a 6G typical scenario knowledge graph based on key technologies, including: obtaining each 6G typical scenario, marking it as each target scenario, and collecting reasoning task data and computing resource data of each target scenario knowledge graph.

[0037] The reasoning task data of each target scenario is processed to obtain the complexity evaluation index of the reasoning task of each target scenario, the computing resource data of each target scenario is analyzed to obtain the computing resource evaluation index of each target scenario, and the reasoning chain length limit threshold of each target scenario is obtained through comprehensive analysis.

[0038] The length of the reasoning chain of each target scene is limited according to the limit threshold of the reasoning chain length of each target scene, and the reasoning effect of the limited reasoning chain length is evaluated to obtain the evaluation result and provide feedback.

[0039] Specifically, the reasoning task data of each target scene is processed to obtain the complexity evaluation index of the reasoning task of each target scene. The specific processing process is: the reasoning task data includes the number of reasoning task targets, the estimated task execution time and the estimated resource requirements.

[0040] The number of critical reasoning task targets, critical estimated task execution time and critical estimated resource requirements are extracted from the 6G key technology database, and a comprehensive analysis is performed to obtain the complexity evaluation index of the reasoning task for each target scenario.

[0041] In a specific embodiment, the number of inference task targets can be directly obtained through a rule mining tool, and the estimated task execution time and the estimated resource requirements can be obtained by matching the number of inference targets and the CPU processing speed in the 6G key technology database. The estimated resource requirements refer to the sum of the number of CPU, memory and disk scheduling.

[0042] Specifically, the complexity evaluation index of each target scene reasoning task is expressed as follows:

[0043]

[0044] Among them, B i represents the complexity evaluation index of the reasoning task of the i-th target scene, e represents a natural constant, and D i represents the number of reasoning task targets for the i-th target scene, D 0 represents the number of critical reasoning task targets set, t i represents the estimated task execution time of the i-th target scenario, t 0 represents the set critical estimated task execution time, Z i represents the estimated resource requirements for the i-th target scenario, Z 0 represents the critical estimated resource demand, ρ 1 Indicates the inference task complexity evaluation impact factor corresponding to the set number of inference task targets, ρ 2 represents the inference task complexity evaluation factor corresponding to the set expected task execution time, ρ 3 It represents the inference task complexity assessment impact factor corresponding to the set estimated resource demand number, i represents each target scenario, i=1,2,3,...,m, and m represents the total number of target scenarios.

[0045] The algorithm of this embodiment combines the number of reasoning task targets, the estimated task execution time and the estimated resource requirements of each target scene, and comprehensively analyzes to obtain the complexity evaluation index of the reasoning task of each target scene. In general, the increase in the number of reasoning task targets may lead to the extension of the estimated task execution time. Because more reasoning rules mean that more possibilities need to be checked and matched in the reasoning process, thereby increasing the amount of calculation, however, in some cases, adding a small number of key rules may significantly improve the accuracy of reasoning, but will not significantly increase the execution time. On the contrary, if there is a large amount of redundancy or conflict between the rules, the reasoning efficiency may be reduced, resulting in an increase in execution time; at the same time, the increase in the number of reasoning task targets usually also means an increase in the demand for system resources, because more rules require more storage space to save, and more computing resources and memory resources are also required to execute in the reasoning process; and in most cases, the extension of the expected task execution time is usually accompanied by an increase in the estimated resource requirements, because a longer execution time often means that more computing resources and storage resources are required to complete the task, and the limitations of system resources, such as CPU speed, memory capacity, etc., will directly affect the execution time of the task. In the case of limited resources, the execution time may increase significantly, or even cause the task to be unable to complete. Comprehensive analysis can obtain a more comprehensive, more accurate and more in-depth reasoning task complexity evaluation index.

[0046] Table 1. Example of reasoning task complexity evaluation index data

[0047]

[0048] As shown in Table 1, the reasoning task complexity evaluation index is jointly determined by the number of reasoning task targets, the expected task execution time and the expected resource requirements. In a specific embodiment, i=1, the critical number of reasoning task targets is 100, the critical expected task execution time is 60min, the critical expected resource requirements are 20, the reasoning task complexity evaluation impact factor corresponding to the set number of reasoning task targets is 0.3, the reasoning task complexity evaluation impact factor corresponding to the set expected task execution time is 0.4, and the reasoning task complexity evaluation impact factor corresponding to the set expected resource requirements is 0.3. This formula takes into account three key factors, namely the number of reasoning task targets, the expected task execution time and the expected resource requirements, which can ensure that the system is configured with sufficient computing resources, storage resources and memory resources, thereby avoiding resource bottlenecks, improving the overall performance of the system, helping to select or optimize the reasoning algorithm, reducing unnecessary calculations, thereby shortening the expected task execution time, improving the system's response speed and user experience, and making the system more flexible, able to adjust the reasoning rules, optimize the algorithm and expand resources as needed, and helping to reduce the risk of system overload and crash, improve the stability and reliability of the system, help quickly locate the cause of the problem and take corresponding recovery measures to ensure business continuity. By standardizing the number of inference task targets, estimated task execution time, and estimated resource requirements for different target scenarios, we ensure that they are compared at the same level, thereby improving the fairness and comparability of the evaluation. 0 The setting helps to avoid the performance degradation of the system due to too many rules. 0 The setting can ensure that key tasks can be completed in time. 0The setting can avoid resource waste and overload. By weighting the impact of the number of inference task targets, the estimated task execution time and the estimated number of resource requirements, it reflects their relative importance in the evaluation index. The weights of different factors can be adjusted according to different needs, making the model highly adaptable. It is not difficult to see that the larger the number of inference task targets, the estimated task execution time or the estimated number of resource requirements, the larger the inference task complexity evaluation index. By evaluating the inference task complexity evaluation index, the algorithm suitable for the current task can be selected more accurately. Tasks with higher complexity may require more efficient algorithms or optimization strategies to ensure that the task can be completed within a reasonable time. By optimizing the inference task and reducing the estimated task execution time, the response speed of the system can be improved, so that users can obtain query results or complete tasks faster, thereby improving the user experience. At the same time, it is helpful for task scheduling and allocation. The system can allocate tasks to computing units or resources with corresponding processing capabilities according to the complexity of the task, thereby achieving load balancing, optimizing resource configuration and efficient utilization, and improving overall performance.

[0049] It should be understood that in this embodiment, 1 , 2 and ρ 3 They are the inference task complexity assessment influencing factors corresponding to the number of inference task targets, estimated task execution time and estimated resource requirements preset in the 6G key technology database, respectively, and represent the numerical values ​​of the degree of influence of the number of inference task targets, estimated task execution time and estimated resource requirements on the inference task complexity assessment index of each target scenario. They can be directly obtained from the 6G key technology database when used. For example, the CPU processing speed and the inference task complexity assessment influencing factors corresponding to the number of inference task targets, estimated task execution time and estimated resource requirements preset in the 6G key technology database form a mapping set, and the real-time CPU processing speed is input into the mapping set to obtain the inference task complexity assessment influencing factors corresponding to the number of inference task targets, estimated task execution time and estimated resource requirements. The mapping relationship can be a one-to-many relationship. In this example, its value range is between 0 and 1.

[0050] It should be explained that the complexity evaluation index of the reasoning task of each target scenario in this embodiment is a quantitative indicator obtained by analyzing the number of reasoning task targets, the expected task execution time and the expected resource requirements of each target scenario, and is used to quantify the complexity of the reasoning task of the target scenario.

[0051] Furthermore, the computing resource data of each target scene is analyzed to obtain a computing resource evaluation index of each target scene. The specific analysis process is as follows: the computing resource data includes CPU processing speed, memory usage and disk throughput.

[0052] The critical CPU processing speed, critical memory usage and critical disk throughput are extracted from the 6G key technology database, and the computing resource evaluation index of each target scenario is obtained through comprehensive analysis.

[0053] In a specific embodiment, the CPU processing speed can be directly obtained through a graphical interface tool. The memory usage rate represents the proportion of the memory currently used by the system to the total memory, which can be monitored through a graphical interface tool. The disk throughput refers to the amount of data processed by the disk per unit time, which can be directly obtained through a performance monitoring tool.

[0054] Specifically, each target scenario calculates the resource evaluation index, and the specific numerical expression is:

[0055]

[0056] Among them, A i represents the computing resource evaluation index of the i-th target scenario, e represents a natural constant, and s i Indicates the CPU processing speed of the i-th target scene, s 0 Indicates the set critical CPU processing speed, C i represents the memory usage of the i-th target scene, C 0 Indicates the set critical memory usage, L i represents the disk throughput of the i-th target scenario, L 0 represents the set critical disk throughput, ω 1 Indicates the computing resource evaluation impact factor corresponding to the set CPU processing speed, ω 2 Indicates the computing resource evaluation impact factor corresponding to the set memory usage, ω 3 It represents the computing resource evaluation impact factor corresponding to the set disk throughput, i represents each target scenario, i=1,2,3,...,m, and m represents the total number of target scenarios.

[0057] The algorithm of this embodiment combines the CPU processing speed, memory usage and disk throughput of each target scenario, and comprehensively analyzes to obtain the computing resource evaluation index of each target scenario. When the CPU processing speed is faster, it can process data in the memory faster, thereby reducing the residence time of data in the memory, which may further reduce the memory usage. At the same time, efficient CPU processing can reduce frequent access to the disk, because more data processing can be completed in the memory, thereby reducing the dependence on disk I / O, and high memory usage may cause the CPU to spend more time on memory access and data exchange, thereby reducing the effective processing speed of the CPU. At the same time, when the memory usage is too high and the system cannot allocate enough memory space for new data, it may write data to the disk, which will increase the write burden of the disk, and the improvement of disk throughput can speed up the reading and writing speed of data and reduce the waiting time of the CPU. Comprehensive analysis can obtain a more comprehensive, accurate and in-depth computing resource evaluation index.

[0058] It should be explained that three key factors, namely CPU processing speed, memory usage and disk throughput, are considered in this embodiment. The system can process query and reasoning tasks faster, provide faster response speed, ensure that resources are reasonably allocated and utilized, avoid resource waste, and support more complex query requirements, improve user experience, and more easily identify system performance bottlenecks, so as to carry out targeted optimization. By standardizing the CPU processing speed, memory usage and disk throughput of different target scenarios, it is ensured that they are compared at the same level, which improves the fairness and comparability of the evaluation. By weighting the impact of CPU processing speed, memory usage and disk throughput, it reflects their relative importance in the evaluation index, and the weights of different factors can be adjusted according to different needs, so that the model has good adaptability. It is not difficult to see that when the CPU processing speed is larger or the memory usage is smaller or the disk throughput is larger, the computing resource evaluation index is larger. By evaluating the computing resource evaluation index, we can accurately reflect the processing capacity of the system under the current hardware and configuration, so as to more accurately set the limit threshold of the inference chain length, avoid resource overload caused by too high a threshold or limit system performance due to too low a threshold, and help balance the resource load between different tasks within the system to ensure that critical tasks receive sufficient resource support. At the same time, it avoids non-critical tasks occupying too many resources and causing the overall performance of the system to decline, ensuring that the system can still run stably and provide high-quality services under high load, which can further improve user satisfaction and loyalty.

[0059] It should be understood that in this embodiment, 1is the computing resource evaluation impact factor corresponding to the CPU processing speed preset in the 6G key technology database, indicating the numerical value of the influence degree of the CPU processing speed on the computing resource evaluation index of each target scenario. When used, the computing resource evaluation impact factor corresponding to the CPU processing speed can be directly obtained from the 6G key technology database, and the corresponding relationship can be a preset mapping relationship. For example, the CPU processing speed and the computing resource evaluation impact factor corresponding to the CPU processing speed preset in the 6G key technology database form a mapping set, and the real-time CPU processing speed is input into the mapping set to obtain the computing resource evaluation impact factor corresponding to the CPU processing speed. The mapping relationship can be one-to-one or many-to-one. In this example, its value range is between 0 and 1; ω 2 The computing resource evaluation impact factor corresponding to the memory usage rate preset in the 6G key technology database represents the numerical value of the influence degree of the memory usage rate on the computing resource evaluation index of each target scenario. When used, the computing resource evaluation impact factor corresponding to the memory usage rate can be directly obtained from the 6G key technology database. The corresponding relationship can be a preset mapping relationship. For example, the memory usage rate and the computing resource evaluation impact factor corresponding to the memory usage rate preset in the 6G key technology database form a mapping set, and the real-time memory usage rate is input into the mapping set to obtain the computing resource evaluation impact factor corresponding to the memory usage rate. The mapping relationship can be one-to-one or many-to-one. In this example, its value range is between 0 and 1; ω 3 It is the computing resource assessment impact factor corresponding to the disk throughput preset in the 6G key technology database, which represents the numerical value of the impact degree of the disk throughput on the computing resource assessment index of each target scenario. When used, the computing resource assessment impact factor corresponding to the disk throughput can be directly obtained from the 6G key technology database, and the corresponding relationship can be a preset mapping relationship. For example, the disk throughput and the computing resource assessment impact factor corresponding to the disk throughput preset in the 6G key technology database form a mapping set, and the real-time disk throughput is input into the mapping set to obtain the computing resource assessment impact factor corresponding to the disk throughput. The mapping relationship can be one-to-one or many-to-one. In this example, its value range is between 0 and 1.

[0060] It should be explained that the computing resource evaluation index of each target scene in this embodiment is a quantitative indicator obtained by analyzing the CPU processing speed, memory usage and disk throughput of each target scene, and is used to quantify the computing resources of each target scene.

[0061] It should be explained that, in this embodiment, the CPU processing speed determines the amount of calculation that can be processed per unit time. A faster CPU processing speed can support the processing of more reasoning task targets at the same time, thereby increasing the number of processable targets. At the same time, under the same task load, the faster the CPU processing speed, the shorter the execution time of a single reasoning task. Therefore, increasing the CPU processing speed can significantly reduce the expected task execution time. And under the premise of achieving the same performance target, a faster CPU processing speed may mean that fewer CPU resources are required. A higher memory usage may limit the number of reasoning task targets that can be processed simultaneously, because the system needs to allocate sufficient memory space for each task. Memory usage High memory usage may lead to frequent memory swapping, thereby increasing I / O waiting time and affecting the execution efficiency of reasoning tasks. At the same time, high memory usage may prompt the demand for increased memory resources to ensure that the system can run stably and handle more reasoning tasks. Disk throughput mainly affects the efficiency of data read and write operations. In reasoning tasks, if a large amount of data is involved, higher disk throughput can support the simultaneous processing of more reasoning task targets. At the same time, high disk throughput means fast data read and write operations, which can reduce the time for data loading and processing, thereby shortening the expected task execution time. In some cases, in order to improve disk throughput, it may be necessary to increase disk resources.

[0062] Specifically, a comprehensive analysis is performed to obtain the limit threshold of the reasoning chain length of each target scenario. The specific analysis process is: based on the reasoning task complexity evaluation index of each target scenario and the computing resource evaluation index of each target scenario, a comprehensive analysis is performed to obtain the limit threshold of the reasoning chain length of each target scenario.

[0063] It needs to be explained that, in this embodiment, the larger the computing resource evaluation index is, the more resources the system or platform has to support complex reasoning processes, which means that even if the length of the reasoning chain increases, the system can maintain a high processing speed and stability and will not be affected by insufficient resources. Increasing the reasoning chain length limit threshold can be regarded as a response and match of the system to high-precision requirements. The larger the reasoning task complexity evaluation index is, the complexity of the reasoning task often determines the depth that needs to be explored in the reasoning process. For complex reasoning tasks, more reasoning steps may be required to reveal hidden relationships or solve complex problems. Therefore, when the reasoning task complexity evaluation index increases, in order to fully cope with this complexity, the reasoning chain length limit threshold also needs to be increased accordingly.

[0064] Specifically, the inference chain length limit threshold of each target scene is expressed as:

[0065] D i =cosh[θ 1 *A i +θ2 *B i ];

[0066] Among them, D i represents the length limit threshold of the inference chain of the i-th target scene, A i represents the computing resource evaluation index of the i-th target scenario, B i represents the complexity evaluation index of the reasoning task of the i-th target scene, θ 1 Indicates the evaluation impact factor of the inference chain length limit threshold corresponding to the set computing resource evaluation index, θ 2 It represents the inference chain length limit threshold evaluation impact factor corresponding to the set reasoning task complexity evaluation index, i represents each target scene, i=1,2,3,...,m, m represents the total number of target scenes.

[0067] like Figure 3 As shown, in a specific embodiment, i=1, θ 1 =0.3,θ 2 =0.5. When A i = 0.1, the functional relationship between the inference chain length limit threshold and the inference task complexity evaluation index is shown in curve a; when A i = 0.5, the functional relationship between the inference chain length limit threshold and the inference task complexity evaluation index is shown in curve b; when A i When =1, the functional relationship between the inference chain length limit threshold and the inference task complexity evaluation index is shown in curve c.

[0068] The algorithm of this embodiment combines the reasoning task complexity evaluation index and the computing resource evaluation index of each target scenario, and comprehensively analyzes to obtain the reasoning chain length limit threshold of each target scenario. The complexity of the reasoning task is positively correlated with the demand for computing resources, that is, the more complex the reasoning task, the more computing resources are required. For reasoning tasks with higher complexity, more computing resources can be allocated to ensure their efficient execution; and for reasoning tasks with lower complexity, the allocation of computing resources can be appropriately reduced to reduce costs. Comprehensive analysis can obtain a more comprehensive, accurate and in-depth reasoning chain length limit threshold.

[0069] It should be explained that in this embodiment, two key factors, namely the reasoning task complexity evaluation index and the computing resource evaluation index, are considered, which can set the reasoning chain length limit threshold more accurately, avoiding the problem of being too high or too low that may be caused by setting the threshold based on only a single factor. The reasoning chain length limit threshold can also be adjusted in real time or regularly to adapt to the needs of different scenarios, while avoiding the occurrence of system overload, and can also reduce the failure of reasoning tasks or performance degradation caused by insufficient resources, which helps to evaluate the system's potential to support more complex reasoning tasks and improve the overall performance and competitiveness of the system. By weighting the influence of the reasoning task complexity evaluation index and the computing resource evaluation index, it reflects their relative importance in the evaluation index, and the weights of different factors can be adjusted according to different needs, so that the model has good adaptability. It is not difficult to see that the larger the reasoning task complexity evaluation index or the smaller the computing resource evaluation index, the larger the reasoning chain length limit threshold. By evaluating the inference chain length limit threshold, we can avoid performance bottlenecks caused by resource exhaustion, reduce unnecessary computing overhead, improve the overall performance of the system, help reduce system processing time, and increase response speed, so that users can get results and feedback faster, thereby improving user experience. It can also be adjusted and optimized according to actual needs to meet the needs of different scenarios.

[0070] It should be understood that in this embodiment, θ 1 is the inference chain length limit threshold assessment impact factor corresponding to the computing resource assessment index preset in the 6G key technology database, indicating the numerical value of the degree of influence of the computing resource assessment index on the inference chain length limit threshold of each target scenario. When used, the inference chain length limit threshold assessment impact factor corresponding to the computing resource assessment index can be directly obtained from the 6G key technology database, and the corresponding relationship can be a preset mapping relationship. For example, the computing resource assessment index and the inference chain length limit threshold assessment impact factor corresponding to the computing resource assessment index preset in the 6G key technology database form a mapping set, and the real-time computing resource assessment index is input into the mapping set to obtain the inference chain length limit threshold assessment impact factor corresponding to the computing resource assessment index. The mapping relationship can be one-to-one or many-to-one. In this example, its value range is between 0 and 1; θ 2It is the inference chain length limit threshold evaluation influence factor corresponding to the inference task complexity evaluation index preset in the 6G key technology database, which represents the numerical value of the influence degree of the inference task complexity evaluation index on the inference chain length limit threshold of each target scenario. When used, the inference chain length limit threshold evaluation influence factor corresponding to the inference task complexity evaluation index can be directly obtained from the 6G key technology database, and the corresponding relationship can be a preset mapping relationship. For example, the inference task complexity evaluation index and the inference chain length limit threshold evaluation influence factor corresponding to the inference task complexity evaluation index preset in the 6G key technology database form a mapping set, and the real-time inference task complexity evaluation index is input into the mapping set to obtain the inference chain length limit threshold evaluation influence factor corresponding to the inference task complexity evaluation index. The mapping relationship can be one-to-one or many-to-one. In this example, its value range is between 0 and 1.

[0071] It should be explained that the threshold value for limiting the length of the reasoning chain of each target scene in this embodiment is a quantitative indicator obtained by analyzing the complexity evaluation index of the reasoning task of each target scene and the computing resource evaluation index of each target scene, and is used to quantify the maximum limit on the length of the reasoning chain of each target scene.

[0072] Specifically, the reasoning chain length of each target scene is restricted according to the limit threshold of the reasoning chain length of each target scene. The specific restriction process is: extract the preset reasoning chain length of each target scene from the 6G key technology database, and compare the reasoning chain length of each target scene with the corresponding reasoning chain length limit threshold of the target scene. If the reasoning chain length of a target scene is less than or equal to the reasoning chain length limit threshold of the target scene, the reasoning chain length of the target scene is not restricted. If the reasoning chain length of a target scene is greater than the reasoning chain length limit threshold of the target scene, the reasoning chain length limit threshold of the target scene is marked as the reasoning chain length of each target scene.

[0073] In a specific embodiment, the reasoning effect is evaluated on the limited reasoning chain length. The specific analysis process is: deploy a number of test samples to perform reasoning tests on each target scene, and obtain the total number of test samples and the number of correct reasoning samples for each target scene, and divide the correct reasoning samples for each target scene by the total number of test samples to obtain the reasoning accuracy of each target scene.

[0074] The reasoning accuracy of each target scene before and after the reasoning chain restriction is obtained, and the effectiveness evaluation value of the reasoning chain restriction is obtained through comprehensive analysis.

[0075] Specifically, the reasoning chain restricts the validity evaluation value, and the specific numerical expression is:

[0076]

[0077] Where C represents the validity evaluation value of the reasoning chain restriction, y i represents the reasoning accuracy of the i-th target scene under the restricted reasoning chain length, represents the inference accuracy of the i-th target scene without restriction on the inference chain length, i represents each target scene, i=1,2,3,...,m, and m represents the total number of target scenes.

[0078] It should be explained that in this embodiment, two key factors are comprehensively considered, namely, the reasoning accuracy of each target scene under the limited reasoning chain length and the reasoning accuracy of each target scene under the unlimited reasoning chain length, which helps to improve the accuracy and efficiency of reasoning, ensure that the system can operate stably and give accurate reasoning results when facing complex and changeable target scenes, enhance system stability and reliability, and improve users' trust and satisfaction with the system. It is not difficult to see that the greater the reasoning accuracy of each target scene under the limited reasoning chain length or the smaller the reasoning accuracy of each target scene under the unlimited reasoning chain length, the greater the reasoning chain limit validity evaluation value. By evaluating the reasoning chain limit validity evaluation value, it is helpful to judge the actual effect of the reasoning chain limit threshold and further optimize it. It can identify those erroneous reasonings that may be caused by too long or too short reasoning chains and reduce the occurrence of such errors, thereby improving the accuracy of reasoning. At the same time, the system can adjust the length and complexity of the reasoning chain according to the evaluation results to ensure the best reasoning effect under limited resources.

[0079] In a specific embodiment, the reasoning chain restriction effectiveness evaluation value is a quantitative indicator obtained by analyzing the reasoning accuracy of each target scenario under the restricted reasoning chain length and the reasoning accuracy of each target scenario under the unrestricted reasoning chain length, and is used to quantify the effectiveness of the reasoning chain restriction.

[0080] Furthermore, the evaluation results are obtained and feedback is given. The specific feedback process is as follows: extracting the reasoning chain restriction validity evaluation threshold from the 6G key technology database, comparing the reasoning chain restriction validity evaluation value with the reasoning chain restriction validity evaluation threshold; if the reasoning chain restriction validity evaluation value is greater than or equal to the reasoning chain restriction validity evaluation threshold, retaining the restricted reasoning chain length without performing additional operations; if the reasoning chain restriction validity evaluation value is less than the reasoning chain restriction validity evaluation threshold, adjusting the reasoning chain length until the reasoning chain restriction validity evaluation value is greater than or equal to the reasoning chain restriction validity evaluation threshold.

[0081] Specifically, the reasoning chain length is adjusted. The specific adjustment process is: extract the preset reasoning chain adjustment length from the 6G key technology database, extend the reasoning chain length of each target scenario in turn according to the reasoning chain adjustment length, and calculate the reasoning chain restriction validity evaluation value. When the reasoning chain restriction validity evaluation value is greater than or equal to the reasoning chain restriction validity evaluation threshold, stop the adjustment operation.

[0082] It should be explained that in the complex reasoning process in this embodiment, there may be certain errors in each step of the reasoning. When the reasoning chain is long, these errors will continue to accumulate, and eventually lead to a decrease in the accuracy of the overall reasoning result. By increasing the limit threshold of the reasoning chain, the number or complexity of the reasoning steps can be limited, thereby reducing the accumulation of errors and improving the accuracy of the reasoning results.

[0083] Reference Figure 2 As shown, the second aspect of the present invention provides a 6G typical scenario knowledge graph construction system based on key technologies, including: a knowledge graph data acquisition module, used to obtain each 6G typical scenario, mark it as each target scene, and collect reasoning task data and computing resource data of each target scene knowledge graph.

[0084] The reasoning chain length limit threshold analysis module is used to process the reasoning task data of each target scene to obtain the complexity evaluation index of the reasoning task of each target scene, analyze the computing resource data of each target scene to obtain the computing resource evaluation index of each target scene, and comprehensively analyze to obtain the reasoning chain length limit threshold of each target scene.

[0085] The reasoning effect evaluation module is used to limit the reasoning chain length of each target scene according to the reasoning chain length limit threshold of each target scene, and evaluate the reasoning effect of the limited reasoning chain length, obtain the evaluation result and provide feedback.

[0086] The 6G key technology database is used to store data related to 6G key technologies, including the critical number of reasoning task targets, the critical expected task execution time, the critical expected resource requirements, the critical CPU processing speed, the critical memory usage rate, the critical disk throughput, the reasoning chain restriction validity evaluation threshold, the reasoning chain length limit threshold evaluation impact factor corresponding to the set computing resource evaluation index, the reasoning chain length limit threshold evaluation impact factor corresponding to the set reasoning task complexity evaluation index, the computing resource evaluation impact factor corresponding to the set CPU processing speed, the computing resource evaluation impact factor corresponding to the set memory usage rate, the computing resource evaluation impact factor corresponding to the set disk throughput, the reasoning task complexity evaluation impact factor corresponding to the set number of reasoning task targets, the reasoning task complexity evaluation impact factor corresponding to the set expected task execution time, the reasoning task complexity evaluation impact factor corresponding to the set expected resource requirements, and the reasoning chain length of each target scenario. The data in the 6G key technology database are obtained by collecting data from multiple 6G key technology experiments.

[0087] The above contents are merely examples and explanations of the structure of the present invention. The technicians in this technical field may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the structure of the invention or exceed the scope defined by the present invention, they should all fall within the protection scope of the present invention.

Claims

1. A method for constructing a 6G typical scenario knowledge graph based on key technologies, characterized in that: include: Obtain each 6G typical scenario, mark it as each target scenario, and collect the reasoning task data and computing resource data of the knowledge graph of each target scenario; The reasoning task data of each target scene is processed to obtain the complexity evaluation index of the reasoning task of each target scene, the computing resource data of each target scene is analyzed to obtain the computing resource evaluation index of each target scene, and the reasoning chain length limit threshold of each target scene is obtained through comprehensive analysis; The length of the reasoning chain of each target scene is limited according to the limit threshold of the reasoning chain length of each target scene, and the reasoning effect of the limited reasoning chain length is evaluated to obtain the evaluation result and provide feedback.

2. The method for constructing a 6G typical scenario knowledge graph based on key technologies according to claim 1 is characterized in that: The reasoning task data of each target scene is processed to obtain the complexity evaluation index of the reasoning task of each target scene. The specific processing process is: The reasoning task data includes the number of reasoning task targets, the estimated task execution time and the estimated resource requirements; The number of critical reasoning task targets, critical estimated task execution time and critical estimated resource requirements are extracted from the 6G key technology database, and a comprehensive analysis is performed to obtain the complexity evaluation index of the reasoning task for each target scenario.

3. The method for constructing a 6G typical scenario knowledge graph based on key technologies according to claim 1 is characterized in that: The computing resource data of each target scenario is analyzed to obtain the computing resource evaluation index of each target scenario. The specific analysis process is as follows: The computing resource data includes CPU processing speed, memory usage and disk throughput; The critical CPU processing speed, critical memory usage and critical disk throughput are extracted from the 6G key technology database, and the computing resource evaluation index of each target scenario is obtained through comprehensive analysis.

4. The method for constructing a 6G typical scenario knowledge graph based on key technologies according to claim 3 is characterized in that: The comprehensive analysis obtains the limit threshold of the inference chain length of each target scene. The specific analysis process is as follows: According to the complexity evaluation index of the reasoning task of each target scenario and the computing resource evaluation index of each target scenario, a comprehensive analysis is performed to obtain the limit threshold of the reasoning chain length of each target scenario.

5. The method for constructing a 6G typical scenario knowledge graph based on key technologies according to claim 4 is characterized in that: The length of the inference chain of each target scene is limited according to the length limit threshold of the inference chain of each target scene. The specific limiting process is: The preset reasoning chain lengths of each target scenario are extracted from the 6G key technology database, and the reasoning chain lengths of each target scenario are compared with the corresponding reasoning chain length limit threshold of the target scenario. If the reasoning chain length of a target scenario is less than or equal to the reasoning chain length limit threshold of the target scenario, the reasoning chain length of the target scenario is not restricted. If the reasoning chain length of a target scenario is greater than the reasoning chain length limit threshold of the target scenario, the reasoning chain length limit threshold of the target scenario is marked as the reasoning chain length of each target scenario.

6. The method for constructing a 6G typical scenario knowledge graph based on key technologies according to claim 1 is characterized in that: The reasoning effect evaluation of the limited reasoning chain length is performed, and the specific analysis process is as follows: Deploy test samples to perform reasoning tests on each target scenario, and obtain the total number of test samples and the number of correct reasoning samples for each target scenario. Divide the correct reasoning samples for each target scenario by the total number of test samples to obtain the reasoning accuracy of each target scenario. The reasoning accuracy of each target scene before and after the reasoning chain restriction is obtained, and the effectiveness evaluation value of the reasoning chain restriction is obtained through comprehensive analysis.

7. The method for constructing a 6G typical scenario knowledge graph based on key technologies according to claim 6 is characterized in that: The evaluation results are obtained and feedback is given, and the specific feedback process is as follows: The reasoning chain restriction validity evaluation threshold is extracted from the 6G key technology database, and the reasoning chain restriction validity evaluation value is compared with the reasoning chain restriction validity evaluation threshold. If the reasoning chain restriction validity evaluation value is greater than or equal to the reasoning chain restriction validity evaluation threshold, the reasoning chain length after restriction is retained without additional operation. If the reasoning chain restriction validity evaluation value is less than the reasoning chain restriction validity evaluation threshold, the reasoning chain length is adjusted until the reasoning chain restriction validity evaluation value is greater than or equal to the reasoning chain restriction validity evaluation threshold.

8. The method for constructing a 6G typical scenario knowledge graph based on key technologies according to claim 7 is characterized in that: The specific adjustment process of adjusting the length of the inference chain is as follows: The preset reasoning chain adjustment length is extracted from the 6G key technology database, the reasoning chain length of each target scenario is extended in turn according to the reasoning chain adjustment length, and the reasoning chain restriction validity evaluation value is calculated. When the reasoning chain restriction validity evaluation value is greater than or equal to the reasoning chain restriction validity evaluation threshold, the adjustment operation is stopped.

9. The method for constructing a 6G typical scenario knowledge graph based on key technologies according to claim 3 is characterized in that: The resource evaluation index of each target scenario is calculated, and the specific numerical expression is: Among them, A i represents the computing resource evaluation index of the i-th target scenario, e represents a natural constant, and s i represents the CPU processing speed of the i-th target scene, s0 represents the set critical CPU processing speed, C i represents the memory usage of the i-th target scene, C0 represents the set critical memory usage, L i represents the disk throughput of the i-th target scenario, L0 represents the set critical disk throughput, ω1 represents the computing resource evaluation impact factor corresponding to the set CPU processing speed, ω2 represents the computing resource evaluation impact factor corresponding to the set memory usage, ω3 represents the computing resource evaluation impact factor corresponding to the set disk throughput, i represents each target scenario, i=1,2,3,...,m, and m represents the total number of target scenarios.

10. A system using a method for constructing a 6G typical scenario knowledge graph based on key technologies as described in any one of claims 1 to 9, characterized in that: include: The knowledge graph data collection module is used to obtain each typical 6G scenario, mark it as each target scenario, and collect the reasoning task data and computing resource data of the knowledge graph of each target scenario; The reasoning chain length limit threshold analysis module is used to process the reasoning task data of each target scene to obtain the complexity evaluation index of the reasoning task of each target scene, analyze the computing resource data of each target scene to obtain the computing resource evaluation index of each target scene, and comprehensively analyze to obtain the reasoning chain length limit threshold of each target scene; The reasoning effect evaluation module is used to limit the length of the reasoning chain of each target scene according to the limit threshold of the reasoning chain length of each target scene, and to evaluate the reasoning effect of the limited reasoning chain length, obtain the evaluation result and provide feedback; The 6G key technology database is used to store data related to 6G key technologies, including the critical number of reasoning task targets, the critical expected task execution time, the critical expected resource requirements, the critical CPU processing speed, the critical memory usage rate, the critical disk throughput, the reasoning chain limit effectiveness evaluation threshold, the reasoning chain length limit threshold evaluation impact factor corresponding to the set computing resource evaluation index, the reasoning chain length limit threshold evaluation impact factor corresponding to the set reasoning task complexity evaluation index, the computing resource evaluation impact factor corresponding to the set CPU processing speed, the computing resource evaluation impact factor corresponding to the set memory usage rate, the computing resource evaluation impact factor corresponding to the set disk throughput, the reasoning task complexity evaluation impact factor corresponding to the set number of reasoning task targets, the reasoning task complexity evaluation impact factor corresponding to the set expected task execution time, the reasoning task complexity evaluation impact factor corresponding to the set expected resource requirements, and the reasoning chain length of each target scenario.

Images