Network problem solution verification method, apparatus and system

Abstract

Embodiments of the present application provide a network problem solution verification method, apparatus and system, which can improve the reliability of a solution corresponding to a network problem and acquired by a user. The method comprises: acquiring an identifier of a service to which a first network problem belongs; on the basis of the identifier of the first network problem, determining from a knowledge graph a first solution corresponding to the first network problem; sending a first verification instruction to a device for which a digital twin network is configured, the first verification instruction being used for instructing said device to simulate execution of the first solution corresponding to the first network problem; and receiving a first verification result from said device, the first verification result being a result obtained after said device executes the first solution corresponding to the first network problem.

Description

Method, device and system for verifying network problem solution Technical Field

[0001] The embodiments of the present application relate to the field of communications, and more specifically, to a method, device, and system for verifying solutions to network problems. Background Art

[0002] A large language model (LLM) is an artificial intelligence model trained on large datasets to perform a wide range of tasks. When LLM parameters exceed a certain scale, they can achieve unexpected capabilities, such as language comprehension, intent understanding, and the ability to remember multiple conversations.

[0003] In the mobile communications field, numerous network problems related to operation, maintenance, and optimization exist. LLMs can be used to generate solutions to these problems. However, LLMs store knowledge via network parameters, and their reasoning process is "black box," meaning it is unexplainable. This makes the solutions they generate unreliable in the mobile communications field, where reliability is paramount.

[0004] Summary of the Invention

[0005] The embodiments of the present application provide a method, device, and system for verifying a solution to a network problem, which can improve the reliability of a solution corresponding to a network problem obtained by a user.

[0006] To achieve the above objectives, this application adopts the following technical solutions:

[0007] In the first aspect, a method for verifying a solution to a network problem is provided. The method can be executed by a device configured with a knowledge graph, or by a component of a device configured with a knowledge graph, such as a processor, chip, or chip system of a device configured with a knowledge graph. It can also be implemented by a logic module or software that can realize all or part of the functions of a device configured with a knowledge graph. Taking the method as an example that can be executed by a device configured with a knowledge graph, the method includes: receiving an identifier of a business to which a first network problem belongs from a control center; determining a first solution corresponding to the first network problem from a knowledge graph based on the identifier of the first network problem; wherein the knowledge graph includes solutions to multiple network problems output by a large language model LLM, the first network problem is one of the multiple network problems, and the first solution is one of the one or more solutions corresponding to the first network problem; sending a first verification instruction to a device configured with a digital twin network, the first verification instruction is used to instruct the device configured with the digital twin network to simulate and execute the first solution corresponding to the first network problem; receiving a first verification result from the device configured with the digital twin network, the first verification result being the result obtained after the device configured with the digital twin network executes the first solution corresponding to the first network problem.

[0008] In the verification method of the network problem solution provided in the embodiment of the present application, the first network problem and the first solution are both output by LLM. LLM has a powerful ability to output network problems and corresponding solutions, so that a device configured with a knowledge graph can use the network problems and corresponding solutions output by LLM to build a more complete knowledge graph. Furthermore, a device configured with a digital twin network is used to verify the solution in the knowledge graph, and the verification result is fed back to the device configured with the knowledge graph, so that the device configured with the knowledge graph can further update its knowledge graph, thereby improving the reliability of the solution corresponding to the network problem obtained by the user.

[0009] In an embodiment of the present application, the first verification result indicates that the first network problem is resolved. The method further includes: sending first feedback information to the control center, where the first feedback information is used to indicate that the verification service corresponding to the first network problem is completed.

[0010] In an embodiment of the present application, the first verification result indicates that the first network problem has not been solved, and the method further includes: determining a second solution corresponding to the first network problem from the knowledge graph according to the identifier of the business to which the first network problem belongs; wherein the second solution is one of the one or more solutions corresponding to the first network problem; sending a second verification instruction to a device configured with a digital twin network, the second verification instruction is used to instruct the device configured with the digital twin network to simulate the execution of the second solution corresponding to the first network problem; receiving a second verification result from the device configured with the digital twin network, the second verification result is the result obtained after the device configured with the digital twin network executes the second solution corresponding to the first network problem. In other words, in an embodiment of the present application, when the first verification result indicates that the first network problem has not been solved, the device configured with the knowledge graph can instruct the device configured with the digital twin network to verify other solutions.

[0011] In one possible implementation, the method further includes: updating the knowledge graph according to the first verification result.

[0012] In one possible implementation, updating the knowledge graph based on the first verification result includes increasing or decreasing the weight of the first solution in the knowledge graph based on the first verification result, or updating the topology of the knowledge graph based on the first verification result. This solution allows devices equipped with a knowledge graph to update the knowledge graph based on the first verification result, thereby further improving the reliability of solutions to network problems in the knowledge graph.

[0013] In one possible implementation, the LLM is trained based on multiple problem pairs, or the LLM is trained based on multiple problem pairs and the scores of the solutions corresponding to each network problem, where the scores are obtained by inputting the solutions corresponding to each problem into a device configured with a digital twin network for simulation, and the problem pairs include network problems and their corresponding solutions in the network management dataset. In other words, in this solution, the device configured with the LLM can train the LLM model based on multiple problem pairs in the network management dataset, so that the LLM model can output solutions to multiple network problems and construct a knowledge graph; or, the device configured with the LLM can train the LLM model based on multiple problem pairs and the scores of the solutions corresponding to each problem, so that the LLM model can output solutions to multiple network problems. Since the scores corresponding to each solution are introduced in the LLM model training process, the trained LLM model can be made more reliable, and thus the solutions to multiple network problems output by the LLM model are more reliable.

[0014] In one possible implementation, the first verification result indicates that the first network problem has been resolved. The method further includes: sending first instruction information to a control center, the first instruction information being used to instruct the control center to apply the first solution to a real network; receiving second feedback information from the control center, the second feedback information being used to indicate the result of applying the first solution to the real network; and updating the knowledge graph based on the second feedback information. Because devices equipped with a knowledge graph can update the knowledge graph based on feedback from the real network, the reliability of solutions to network problems in the knowledge graph can be further improved.

[0015] In a possible implementation, the first network problem includes at least one of the following: a network alarm problem, a user complaint problem, and an optimization demand problem.

[0016] In one possible implementation, the first verification instruction includes at least one of the following: an identifier of the business to which the first network problem belongs, an identifier of the first network problem, an identifier of the first solution, an area location identifier, a control instruction, an execution time, and information collection setting information; the first verification result includes at least one of the following: an identifier of the business to which the first network problem belongs, an identifier of the first network problem, an identifier of the first solution, an area location identifier, and network status information.

[0017] In a second aspect, a method for verifying a solution to a network problem is provided. The method can be executed by a device configured with a digital twin network, or by a component of the device configured with a digital twin network, such as a processor, chip, or chip system of the device configured with a digital twin network, or by a logic module or software that can realize all or part of the functions of the device configured with a digital twin network. Taking the method as an example that the method can be executed by a device configured with a digital twin network, the method includes: receiving a first verification instruction from a device configured with a knowledge graph, the first verification instruction being used to instruct the device configured with the digital twin network to simulate and execute a first solution corresponding to a first network problem, wherein the first solution corresponding to the first network problem is determined by the device configured with the knowledge graph based on an identifier of a business to which the first network problem belongs, from the knowledge graph, the knowledge graph including solutions to multiple network problems output by a large language model (LLM), the first network problem being one of the multiple network problems, and the first solution being one of the one or more solutions corresponding to the first network problem; and sending a first verification result to the device configured with the knowledge graph, the first verification result being a result obtained after the device configured with the digital twin network executes the first solution corresponding to the first network problem.

[0018] In the verification method of the network problem solution provided in the embodiment of the present application, the first network problem and the first solution are both output by LLM. LLM has a powerful ability to output network problems and corresponding solutions, so that a device configured with a knowledge graph can use the network problems and corresponding solutions output by LLM to build a more complete knowledge graph. Furthermore, a device configured with a digital twin network is used to verify the solution in the knowledge graph, and the verification result is fed back to the device configured with the knowledge graph, so that the device configured with the knowledge graph can further update its knowledge graph, thereby improving the reliability of the solution corresponding to the network problem obtained by the user.

[0019] In an embodiment of the present application, the first verification result indicates that the first network problem has not been solved, and the method further includes: receiving a second verification instruction from a device configured with a knowledge graph, the second verification instruction being used to instruct the device configured with a digital twin network to simulate and execute the second solution corresponding to the first network problem, wherein the second solution corresponding to the first network problem is determined from the knowledge graph by the device configured with the knowledge graph based on the identifier of the business to which the first network problem belongs, and the second solution is one of the one or more solutions corresponding to the first network problem; sending a second verification result to the device configured with the knowledge graph, the second verification result being the result obtained after the device configured with the digital twin network executes the second solution corresponding to the first network problem. In other words, in an embodiment of the present application, when the first verification result indicates that the first network problem has not been solved, the device configured with the digital twin network can verify other solutions according to the instructions of the device configured with the knowledge graph.

[0020] In an embodiment of the present application, the LLM is obtained by training based on multiple problem pairs, or the LLM is obtained by training based on multiple problem pairs and the scores of the solutions corresponding to each network problem, and the scores are obtained after the solutions corresponding to each problem are input into a device configured with a digital twin network for simulation, and the multiple problem pairs include network problems and their corresponding solutions in the network management data set. In other words, in this solution, the device configured with the LLM can train the LLM model based on the multiple problem pairs in the network management data set, so that the LLM model can output solutions to multiple network problems and construct a knowledge graph; or, the device configured with the LLM can train the LLM model based on the multiple problem pairs and the scores of the solutions corresponding to each problem, so that the LLM model can output solutions to multiple network problems. Since the scores corresponding to each solution are introduced in the LLM model training process, the trained LLM model can be made more reliable, and thus the solutions to multiple network problems output by the LLM model are more reliable.

[0021] In the embodiment of the present application, the first network problem includes at least one of the following: a network alarm problem, a user complaint problem, and an optimization demand problem.

[0022] In an embodiment of the present application, the first verification instruction includes at least one of the following: an identifier of the business to which the first network problem belongs, an identifier of the first network problem, an identifier of the first solution, an area location identifier, a control instruction, an execution time, and information collection setting information; the first verification result includes at least one of the following: an identifier of the business to which the first network problem belongs, an identifier of the first network problem, an identifier of the first solution, an area location identifier, and network status information.

[0023] On the third aspect, a model training method is provided, which can be executed by a device configured with an LLM, or by a component of a device configured with an LLM, such as a processor, chip, or chip system of a device configured with an LLM, or by a logic module or software that can realize all or part of the functions of a device configured with an LLM. Taking the method that can be executed by a device configured with an LLM as an example, the method includes: inputting a solution corresponding to each network problem in a plurality of problem pairs to a device configured with a digital twin network, where the problem pairs include network problems and their corresponding solutions in a network management data set; receiving a score for the solution corresponding to each network problem from the device configured with a digital twin network, wherein the score is obtained by the device configured with the digital twin network verifying the solution corresponding to each problem and evaluating it according to the verification results; based on the scores of the plurality of problem pairs and the solutions corresponding to each network problem, performing model training to obtain a large language model LLM.

[0024] The model training method provided in the embodiment of the present application is that a device configured with LLM inputs the solution corresponding to each network problem in multiple problem pairs to a device configured with a digital twin network, so that the device configured with the digital twin network simulates the execution of the solution corresponding to each network problem, and obtains a score for the solution corresponding to each network problem, thereby realizing an automatic scoring mechanism and avoiding errors caused by scoring by operation and maintenance experts. Furthermore, the device configured with LLM uses multiple problem pairs and the solution corresponding to each network problem, and trains the LLM model with the highest score as the target score. Since the scoring of the device configured with the digital twin network is introduced in the LLM model training process, the trained LLM model can be made more reliable.

[0025] In one possible implementation, the LLM includes a network policy model, which is trained based on multiple question pairs.

[0026] In one possible implementation, the LLM includes a reward model trained based on multiple problem pairs and the scores of the solutions corresponding to each network problem in the problem pairs. The scores of the solutions corresponding to each network problem used to train the reward model are output by devices equipped with digital twin networks. This implements an automatic scoring mechanism, avoids errors caused by scoring by operations and maintenance experts, and increases the reliability of the reward model.

[0027] In a fourth aspect, a scoring method is provided, which can be executed by a device configured with a digital twin network, or by a component of a device configured with a digital twin network, such as a processor, chip, or chip system of a device configured with a digital twin network, or by a logic module or software that can realize all or part of the functions of a device configured with a digital twin network. Taking the method as an example in which the method can be executed by a device configured with a digital twin network, the method includes: receiving a solution corresponding to each network problem in a plurality of problem pairs from a device configured with an LLM, the plurality of problem pairs including network problems and their corresponding solutions in a network management data set; and sending a score of the solution corresponding to each network problem to the device configured with the LLM, wherein the score is obtained after the device configured with the digital twin network verifies the solution corresponding to each problem and evaluates it based on the verification results.

[0028] The scoring method provided in the embodiment of the present application is that a device configured with LLM inputs the solution corresponding to each network problem in multiple problem pairs to a device configured with a digital twin network, so that the device configured with the digital twin network simulates and executes the solution corresponding to each network problem, and obtains the score of the solution corresponding to each network problem, thereby realizing an automatic scoring mechanism and avoiding errors caused by scoring by operation and maintenance experts. Furthermore, the device configured with LLM uses multiple problem pairs and the solution corresponding to each network problem, and trains the LLM model with the highest score as the target score. Since the scoring of the device configured with the digital twin network is introduced in the LLM model training process, the trained LLM model can be made more reliable.

[0029] In a fifth aspect, a communication device is provided for implementing the various methods described above. The communication device may be the device configured with a knowledge graph in the first aspect, or a device included in the device configured with a knowledge graph, such as a chip; or, the communication device may be the device configured with a digital twin network in the second aspect, or a device in the device configured with a digital twin network, such as a chip. Alternatively, the communication device may be the device configured with an LLM in the third aspect, or a device included in the device configured with an LLM, such as a chip; or, the communication device may be the device configured with a digital twin network in the fourth aspect, or a device in the device configured with a digital twin network, such as a chip.

[0030] The communication device includes modules, units, or means corresponding to the above-mentioned method, which can be implemented by hardware, software, or hardware executing corresponding software implementation. The hardware or software includes one or more modules or units corresponding to the above-mentioned functions.

[0031] In some possible designs, the communication device may include a processing module and a communication module. The communication module may include an output module (or a sending module) and an input module (or a receiving module), respectively configured to implement the output (or sending) and input (or receiving) functions of any of the above aspects and any possible designs thereof. The processing module may be configured to implement the processing functions of any of the above aspects and any possible designs thereof.

[0032] Optionally, the communication device further includes a storage module for storing program instructions and data.

[0033] In a sixth aspect, a communication device is provided, comprising: at least one processor configured to execute a computer program or instruction, or configured to cause the communication device to execute the method described in any of the above aspects through a logic circuit. The communication device may be the device configured with a knowledge graph in the first aspect, or a device included in the device configured with a knowledge graph, such as a chip; or the communication device may be the device configured with a digital twin network in the second or fourth aspect, or a device included in the device configured with a digital twin network, such as a chip; or the communication device may be the device configured with an LLM in the third aspect, or a device included in the device configured with an LLM, such as a chip.

[0034] In some possible designs, the communication device further includes a memory for storing computer instructions and / or configuration files of logic circuits. Optionally, the memory is integrated with the processor, or the memory is independent of the processor.

[0035] In one possible design, the communication device further includes a communication interface for inputting and / or outputting signals.

[0036] In some possible designs, the communication interface is an interface circuit for reading and writing computer instructions. For example, the interface circuit is used to receive computer execution instructions (computer execution instructions are stored in a memory, may be read directly from the memory, or may pass through other devices) and transmit them to the processor.

[0037] In some possible designs, the communication interface is used to communicate with modules outside the communication device.

[0038] In some possible designs, the communication device may be a chip system. When the communication device is a chip system, the chip system may include a chip or may include a chip and other discrete devices.

[0039] In a seventh aspect, a communication device is provided, comprising: a logic circuit and an interface circuit; the interface circuit is used to input information and / or output information; the logic circuit is used to execute the method described in any of the above aspects, processing the input information and / or generating output information. The communication device can be the device configured with a knowledge graph in the first aspect, or a device included in the device configured with a knowledge graph, such as a chip; or the communication device can be the device configured with a digital twin network in the second or fourth aspect, or a device included in the device configured with a digital twin network, such as a chip; or the communication device can be the device configured with an LLM in the third aspect, or a device included in the device configured with an LLM, such as a chip.

[0040] It can be understood that when the communication device provided in any one of the fifth to seventh aspects is a chip, the above-mentioned sending action / function can be understood as output information, and the above-mentioned receiving action / function can be understood as input information.

[0041] In an eighth aspect, a computer-readable storage medium is provided, in which a computer program or instruction is stored. When the computer program or instruction is executed by a processor, the method described in any one of the above aspects is executed.

[0042] In a ninth aspect, a computer program product is provided, which, when executed by a processor, enables the method described in any one of the above aspects to be executed.

[0043] In a tenth aspect, a communication device is provided, which includes a module / unit for executing the method described in any one of the first to fourth aspects above.

[0044] Among them, the technical effects brought about by any design method in the above-mentioned fifth to tenth aspects can refer to the technical effects brought about by different design methods in the above-mentioned first, second, third or fourth aspects, and will not be repeated here.

[0045] On the eleventh aspect, a communication system is provided, including: a device configured with a knowledge graph and a device configured with a digital twin network; the device configured with the knowledge graph is used to receive an identifier of a business to which a first network problem belongs from a control center, and determine a first solution corresponding to the first network problem from the knowledge graph according to the identifier of the business to which the first network problem belongs; wherein the knowledge graph includes solutions to multiple network problems output by a large language model LLM, the first network problem is one of the multiple network problems, and the first solution is one of the one or more solutions corresponding to the first network problem; the device configured with the knowledge graph is also used to send a A device configured with a digital twin network sends a first verification instruction, wherein the first verification instruction is used to instruct the device configured with the digital twin network to simulate the execution of a first solution corresponding to a first network problem; the device configured with the digital twin network is used to receive the first verification instruction from a device configured with a knowledge graph, and after simulating the execution of the first solution according to the first verification instruction, send a first verification result to the device configured with the knowledge graph, wherein the first verification result is the result obtained after the device configured with the digital twin network executes the first solution corresponding to the first network problem; the device configured with the knowledge graph is also used to receive the first verification result from the device configured with the digital twin network.

[0046] In the communication system provided in the embodiment of the present application, the first verification result indicates that the first network problem is resolved; the device configured with the knowledge graph is also used to send first feedback information to the control center, and the first feedback information is used to indicate that the verification service corresponding to the first network problem is completed.

[0047] In the communication system provided in an embodiment of the present application, the first verification result indicates that the first network problem has not been resolved; the device configured with the knowledge graph is also used to determine the second solution corresponding to the first network problem from the knowledge graph based on the identifier of the first network problem; wherein the second solution is one of the one or more solutions corresponding to the first network problem; the device configured with the knowledge graph is also used to send a second verification instruction to the device configured with the digital twin network, wherein the second verification instruction is used to instruct the device configured with the digital twin network to simulate the execution of the second solution corresponding to the first network problem; the device configured with the digital twin network is also used to receive the second verification instruction from the device configured with the knowledge graph, and after simulating the execution of the second solution according to the second verification instruction, send the second verification result to the device configured with the knowledge graph; the device configured with the knowledge graph is also used to receive the second verification result from the device configured with the digital twin network.

[0048] In the communication system provided in the embodiment of the present application, the device configured with the knowledge graph is also used to update the knowledge graph according to the first verification result.

[0049] In the communication system provided in an embodiment of the present application, the device configured with the knowledge graph is also used to update the knowledge graph according to the first verification result, including: the device configured with the knowledge graph is also used to increase or decrease the weight corresponding to the first solution in the knowledge graph according to the first verification result, or, update the topological structure of the knowledge graph according to the first verification result.

[0050] In the communication system provided in the embodiment of the present application, the LLM is obtained by training based on multiple problem pairs, or the LLM is obtained by training based on multiple problem pairs and the scoring of the solution corresponding to each problem, and the score is obtained after inputting the solution corresponding to each problem into a device configured with a digital twin network for simulation, and the multiple problem pairs include network problems in the network management data set and their corresponding solutions.

[0051] In the communication system provided in the embodiment of the present application, the first verification result indicates that the first network problem is solved; the device configured with the knowledge graph is also used to send first indication information to the control center, and the first indication information is used to instruct the control center to apply the first solution to the real network; the device configured with the knowledge graph is also used to receive second feedback information from the control center, and update the knowledge graph according to the second feedback information, wherein the second feedback information is used to indicate the result after the first solution is applied to the real network.

[0052] In the communication system provided by the embodiment of the present application, the first network problem includes at least one of the following: a network alarm problem, a user complaint problem, and an optimization demand problem.

[0053] In the communication system provided in an embodiment of the present application, the first verification instruction includes at least one of the following: an identifier of the service to which the first network problem belongs, an identifier of the first network problem, an identifier of the first solution, an area location identifier, a control instruction, an execution time, and information collection setting information; the first verification result includes at least one of the following: an identifier of the service to which the first network problem belongs, an identifier of the first network problem, an identifier of the first solution, an area location identifier, and network status information.

[0054] Among them, the technical effects brought about by any design method in the above-mentioned eleventh aspect can refer to the technical effects brought about by different design methods in the above-mentioned first aspect or second aspect, and will not be repeated here.

[0055] In the twelfth aspect, a communication system is provided, which includes a device configured with LLM for executing the method described in the third aspect and a device configured with a digital twin network for executing the method described in the fourth aspect.

[0056] Among them, the technical effects brought about by any design method in the above-mentioned twelfth aspect can refer to the technical effects brought about by different design methods in the above-mentioned third aspect or fourth aspect, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] FIG1 is a schematic diagram of injecting knowledge of a knowledge graph into an LLM according to an embodiment of the present application;

[0058] FIG2 is a schematic diagram of a communication system provided in an embodiment of the present application;

[0059] FIG3 is a schematic diagram of device interaction in a communication system according to an embodiment of the present application;

[0060] FIG4 is a schematic diagram of a communication device provided in an embodiment of the present application;

[0061] FIG5 is a schematic diagram of a model training scheme provided in an embodiment of the present application;

[0062] FIG6 is a schematic diagram of a specific process of model training provided in an embodiment of the present application;

[0063] FIG7 is a schematic diagram of a method for verifying a network problem solution provided in an embodiment of the present application;

[0064] FIG8 is a schematic diagram of a topological structure of a network problem judgment decision tree provided in an embodiment of the present application;

[0065] FIG9 is a schematic diagram of a topological structure of a decision tree for determining a network problem solution according to an embodiment of the present application;

[0066] FIG10 is a schematic diagram of a communication device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0067] In the description of this application, unless otherwise specified, " / " indicates that the objects associated before and after are in an "or" relationship, for example, A / B can represent A or B; "and / or" in this application is merely a description of the association relationship of associated objects, indicating that three relationships may exist, for example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural.

[0068] In the description of this application, unless otherwise specified, "plurality" means two or more than two. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b and / or c can be represented by: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or plural.

[0069] In addition, to facilitate the clear description of the technical solutions of the embodiments of the present application, in the embodiments of the present application, the words "first" and "second" are used to distinguish between identical or similar items with substantially the same functions and effects. Those skilled in the art will understand that the words "first" and "second" do not limit the quantity or execution order, and the words "first" and "second" do not necessarily mean different.

[0070] In the embodiments of this application, words such as "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be construed as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner to facilitate understanding.

[0071] It will be understood that the “embodiment” mentioned throughout the specification means that the specific features, structures or characteristics related to the embodiment are included in at least one embodiment of the present application. Therefore, the various embodiments throughout the specification do not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. It will be understood that in the various embodiments of the present application, the size of the sequence number of each process does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present application.

[0072] It can be understood that in this application, "when" and "if" both mean that corresponding processing will be taken under certain objective circumstances, and do not limit the time, nor do they require any judgment action when implementing, nor do they mean that there are other limitations.

[0073] It is understood that some optional features in the embodiments of the present application may, in certain scenarios, be implemented independently of other features, such as the solution on which they are currently based, to solve corresponding technical problems and achieve corresponding effects. They may also be combined with other features in certain scenarios as needed. Accordingly, the devices provided in the embodiments of the present application may also implement these features or functions accordingly, which will not be described in detail here.

[0074] In this application, unless otherwise specified, the same or similar parts between the various embodiments can refer to each other. In the various embodiments of this application, unless otherwise specified and there is no logical conflict, the terms and / or descriptions between different embodiments are consistent and can be referenced to each other, and the technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships. The following description of the embodiments of this application does not constitute a limitation on the scope of protection of this application.

[0075] In order to facilitate understanding of the technical solutions of the embodiments of the present application, a brief introduction to the relevant technologies and terms of the present application is first given.

[0076] 1.LLM.

[0077] LLM is an AI model that continuously collects and builds high-quality datasets (including manual annotations), enabling continuous optimization, rapid iteration, and data reflow. When the training data included in LLM exceeds a certain scale, it can achieve unexpected capabilities, such as language comprehension, intent understanding, and multi-turn conversation memory. LLM stores knowledge through network parameters, and its reasoning process is "black box," meaning it is unexplainable.

[0078] Table 1 shows a comparison of the advantages and disadvantages of LLM represented by artificial intelligence (AI) chatbots. As shown in Table 1, although AI chatbots have many advantages in question-answering, they have defects in logical reasoning.

[0079] Table 1

[0080] In the mobile communications field, there are many network issues related to operation, maintenance, and optimization. LLM can be used to provide solutions to these network issues. However, applying LLM to the mobile communications field presents challenges, as shown in Table 2.

[0081] Table 2

[0082] As can be seen from Table 2, the current LLM is essentially still learning the distribution law of data and lacks causal analysis and logical reasoning capabilities.

[0083] 2. Knowledge graph.

[0084] Knowledge graphs store knowledge in triples, for example, and represent entities and relationships using a topological structure of nodes and edges. This structured knowledge base is easy to understand. Knowledge graphs represent the logic of knowledge display and decouple it from the specific application of knowledge, making the reasoning process clear and explicit. This makes it interpretable and enables causal analysis and logical reasoning.

[0085] 3. Use the knowledge from the knowledge graph to train LLM.

[0086] In embodiments of the present application, knowledge from the knowledge graph can be used to train the LLM. For example, the pre-training corpus can be cleaned, filtered, and organized based on the knowledge from the knowledge graph to improve the quality of the corpus. Alternatively, for example, the LLM model structure can be modified to enable it to simultaneously process text sequences or improve pre-training tasks to improve the effectiveness of pre-training.

[0087] Figure 1 is a schematic diagram of a communication system provided in an embodiment of the present application. As shown in Figure 1, the communication system includes a device configured with a knowledge graph and a device configured with a digital twin network.

[0088] Among them, a device configured with a knowledge graph is used to send verification instructions of solutions corresponding to network problems to a device configured with a digital twin network and receive verification results, wherein the knowledge graph includes solutions to multiple network problems output by the LLM model.

[0089] The device configured with the digital twin network is used to receive verification instructions from the device configured with the knowledge graph, simulate the execution of the solution according to the verification instructions, and then send the verification results to the device configured with the knowledge graph.

[0090] Optionally, as shown in Figure 1, the communication system provided by the embodiment of the present application also includes: a management and control center, which is used to send an identifier of the business to which the network problem belongs to the device configured with the knowledge graph, and is also used to receive feedback information from the device configured with the knowledge graph.

[0091] Optionally, as shown in FIG1 , the communication system provided in an embodiment of the present application further includes: a device configured with an LLM, which is used to output solutions to multiple network problems and is also used to train an LLM model.

[0092] It should be noted that the communication system described in the embodiment of the present application is intended to more clearly illustrate the technical solution of the embodiment of the present application, and does not constitute a limitation on the technical solution provided in the embodiment of the present application. Ordinary technicians in this field can know that with the evolution of network architecture and the emergence of new business scenarios, the technical solution provided in the embodiment of the present application is also applicable to similar technical problems.

[0093] Optionally, the relevant functions of the device equipped with a knowledge graph, the digital network twin device, the device equipped with LLM, and the control center involved in this application can be implemented by one device, or by multiple devices together, or by one or more functional modules within a device, or it can be one or more chips, or a system on chip (SOC) or a chip system. The chip system can be composed of chips, or it can include chips and other discrete devices. The embodiments of this application do not specifically limit this.

[0094] It is understandable that the above functions can be network elements in hardware devices, software functions running on dedicated hardware, or a combination of hardware and software, or virtualized functions instantiated on a platform (e.g., a cloud platform).

[0095] In conjunction with the communication system shown in Figure 1, for example, Figure 2 is a schematic diagram of device interaction of the communication system provided in an embodiment of the present application. As shown in Figure 2, in order to improve the reliability of the solution corresponding to the network problem obtained by the user, on the one hand, the knowledge graph of the device configured with the knowledge graph comes from the multiple problem pairs output by the device configured with the LLM, and the device configured with the LLM can use the multiple problem pairs corresponding to the knowledge graph in the device configured with the knowledge graph to train the LLM model; on the other hand, the device configured with the digital twin network obtains the solution corresponding to each network problem corresponding to the multiple problem pairs from the device configured with the LLM, simulates and verifies the solution corresponding to each network problem, obtains the score of the solution corresponding to each network problem, and then feeds back the problem, the corresponding solution, and the score to the device configured with the LLM. The device configured with the LLM uses the problem, the corresponding solution, and the score to train the reward model of the LLM model; on the other hand, the device configured with the knowledge graph sends a verification instruction to the device configured with the digital twin network, and the device configured with the digital twin network simulates the solution corresponding to the network problem in the problem pair according to the verification instruction, and feeds back the verification result to the device configured with the knowledge graph to further update the knowledge graph in the device configured with the knowledge graph.

[0096] In combination with the communication system shown in Figure 1, for example, Figure 3 is a schematic diagram of the interaction between internal devices of the communication system provided in an embodiment of the present application and a schematic diagram of the interaction between the communication system and the physical network.

[0097] As shown in Figure 3, the device configured with the knowledge graph can update the knowledge graph based on the problem output by the LLM model of the device configured with LLM; the device configured with the knowledge graph can send a verification instruction for verifying the solution to the network problem in the knowledge graph to the device configured with the digital twin network, and receive the verification result of the device configured with the digital twin network. The device configured with the knowledge graph can update the knowledge graph based on the verification result; the device configured with the digital twin network can simulate and verify the solution to the network problem according to the verification instruction to obtain a verification result.

[0098] In one possible implementation, the device configured with the digital twin network can issue instructions to the physical network based on the verification results, instructing the physical network to apply the solution. When the device configured with the digital twin network verifies the solution to the network problem, it synchronizes the data of the physical network.

[0099] In another possible implementation, a device configured with a knowledge graph sends instruction information to the control center to instruct the application of the verified solution to the network problem to the physical network. The control center can send instruction information to the physical network based on the instruction information to instruct the application of the verified solution to the network problem to the physical network, so that the verified solution to the network problem is implemented in the physical network. After the physical network implements the verified solution to the network problem, the information collection can be reported to the control center.

[0100] For example, as shown in Figure 3, the physical network in the embodiment of the present application may include core network equipment, access network equipment, and terminal equipment. Among them, the data plane of the physical network is used for information collection and then reported to the control center.

[0101] The devices configured with knowledge graphs, devices configured with digital twin networks, devices configured with LLMs, and related functions of the control center involved in this application can be implemented through the communication device 400 in Figure 4. Figure 4 is a structural diagram of the communication device 400 provided in an embodiment of the present application. The communication device 400 includes one or more processors 401, a communication line 402, and at least one communication interface (Figure 2 is only exemplary, taking the communication interface 404 and a processor 401 as an example for explanation), and optionally may also include a memory 403.

[0102] The processor 401 may be a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program of the present application.

[0103] The communication line 402 may include a path for connecting different components.

[0104] Communication interface 404 may be a transceiver module for communicating with other devices or communication networks, such as Ethernet, a radio access network (RAN), a terminal, or a wireless local area network (WLAN). For example, the transceiver module may be a device such as a transceiver or a transceiver. Alternatively, communication interface 404 may be a transceiver circuit or input / output interface within processor 401, for implementing signal input and output to the processor.

[0105] The memory 403 may be a device having a storage function. For example, it may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory may exist independently and be connected to the processor via a communication line 402. The memory may also be integrated with the processor.

[0106] Memory 403 is used to store computer-executable instructions for executing the solution of the present application, and is controlled by processor 401. Processor 401 is used to execute the computer-executable instructions stored in memory 403, thereby implementing the verification method or model training method for the network problem solution provided in the embodiment of the present application.

[0107] Alternatively, optionally, in an embodiment of the present application, the processor 401 may also execute the processing-related functions in the verification method or model training method of the network problem solution provided in the following embodiments of the present application, and the communication interface 404 is responsible for communicating with other devices or communication networks, which is not specifically limited in the embodiment of the present application.

[0108] Optionally, the computer-executable instructions in the embodiments of the present application may also be referred to as application code, which is not specifically limited in the embodiments of the present application.

[0109] In a specific implementation, as an embodiment, the processor 401 may include one or more CPUs, such as CPU0 and CPU1 in FIG. 2 .

[0110] In a specific implementation, as an embodiment, the communication device 400 may include multiple processors, such as the processor 401 and the processor 407 in FIG2 . Each of these processors may be a single-core processor or a multi-core processor. The processors here may include, but are not limited to, at least one of the following: a central processing unit (CPU), a microprocessor, a digital signal processor (DSP), a microcontroller unit (MCU), or an artificial intelligence processor, etc., and each computing device may include one or more cores for executing software instructions to perform calculations or processing.

[0111] In a specific implementation, as an embodiment, the communication device 400 may further include an output device 405 and an input device 406. The output device 405 communicates with the processor 401 and can display information in a variety of ways. For example, the output device 405 can be a liquid crystal display (LCD), a light emitting diode (LED) display device, a cathode ray tube (CRT) display device, or a projector. The input device 406 communicates with the processor 401 and can receive user input in a variety of ways. For example, the input device 406 can be a mouse, a keyboard, a touch screen device, or a sensor device.

[0112] The communication device 400 described above may also sometimes be referred to as a communication apparatus, and may be a general-purpose device or a dedicated device. For example, the communication device 400 may be a desktop computer, a portable computer, a network server, a personal digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, an embedded device, or a device having a similar structure to that shown in FIG1 . The embodiments of the present application do not limit the type of the communication device 400.

[0113] In addition, the composition structure shown in FIG4 does not constitute a limitation on the communication device. In addition to the components shown in FIG4, the communication device may include more or fewer components than shown, or combine certain components, or arrange the components differently.

[0114] The following describes the model training method and the verification method of the network problem solution provided in the embodiment of the present application in conjunction with the communication system shown in Figure 2.

[0115] Figure 5 is a schematic diagram of an example of a model training method provided in an embodiment of the present application. The method is illustrated by taking the interaction between a device configured with a digital twin network and a device configured with an LLM as an example. Of course, the subject that executes the action of the device configured with a digital twin network in the method can also be a device / module in a device configured with a digital twin network, such as a chip, processor, processing unit, etc. in a device configured with a digital twin network; the subject that executes the action of the device configured with an LLM in the method can also be a device / module in a device configured with an LLM, such as a chip, processor, processing unit, etc. in a device configured with an LLM, and the embodiment of the present application does not make specific limitations on this. The processing performed by a single execution subject (for example, a device configured with a digital twin network or a device configured with an LLM) in the embodiment of the present application can also be divided into executions by multiple execution subjects, and these execution subjects can be logically and / or physically separated. Exemplarily, as shown in Figure 5, the model training method 500 provided in the embodiment of the present application includes:

[0116] S510: The device configured with the LLM outputs a solution corresponding to each problem in a plurality of problem pairs to the device configured with the digital twin network. Correspondingly, the device configured with the digital twin network receives a solution corresponding to each problem in a plurality of problem pairs from the device configured with the LLM.

[0117] In the embodiment of the present application, the plurality of problem pairs include network problems and their corresponding solutions in the network management data set.

[0118] For example, the solution to the network problem may be provided by an operation and maintenance expert.

[0119] For example, the network problems in the network management data set may be network problems in a real network, or network problems raised by operation and maintenance experts, which is not limited in the embodiments of the present application.

[0120] Exemplarily, network problems can be triple verification problems. For example, a triple is a radio access network A, a base station B, and a cell C. A device equipped with an LLM provides several solutions for the inclusion relationships of A, B, and C. A device equipped with a digital twin network simulates these inclusion relationships. For example, after simulation, the device equipped with the digital twin network outputs: radio access network A includes base station B, and base station B includes cell C. Both inclusion relationships are correct, and the score is 100; radio access network A includes base station B, and base station B does not include cell C. Only one of the two inclusion relationships is correct, and the score is 50; radio access network A does not include base station B, and base station B does not include cell C. Both inclusion relationships are incorrect, and the score is 0.

[0121] Exemplarily, the network problem in the embodiment of the present application may be a causal relationship problem. For example, the network problem is that the user's average signal interference noise ratio (SINR) is less than or equal to 0db, and the corresponding solution is to change the number of co-frequency cells to determine whether the number of co-frequency cells is the cause of the user's average SINR being less than or equal to 0db; or, for example, the number of co-frequency cells is 3 or 6, and the device configured with the digital twin network simulates the situation where the number of co-frequency cells is 3 or 6 respectively. If the user's average SINR changes when the number of co-frequency cells is 3 and the number of co-frequency cells is 6, it can be considered that the number of co-frequency cells is the cause of the user's average SINR being less than or equal to 0db, and the score is 100; if the user's average SINR does not change when the number of co-frequency cells is 3 and the number of co-frequency cells is 6, it can be considered that the number of co-frequency cells is not the cause of the user's average SINR being less than or equal to 0db, and the score is 0.

[0122] Exemplarily, the network problem in the embodiment of the present application may be a network alarm type problem, for example, the power of the main cell is low, and the corresponding solution D is power control based on the deep Q network (DQN); the corresponding solution E is power control based on the data channel, and the device configured with the digital twin network simulates the power control based on DQN. The simulation result shows that the power of the main cell increases, and the score is 100; the device configured with the digital twin network simulates the power control based on the data channel. The simulation result shows that the power of the main cell is still low, and the score is 0.

[0123] In the embodiments of the present application, the network problem may also be other problems. The above are just a few examples and the embodiments of the present application do not limit this.

[0124] S520: The device configured with the digital twin network sends a score of the solution corresponding to each network problem to the device configured with the LLM. Correspondingly, the device configured with the LLM receives the score of the solution corresponding to each network problem from the device configured with the digital twin network.

[0125] In an embodiment of the present application, after a device configured with a digital twin network receives a solution corresponding to each problem in multiple problem pairs from a device configured with an LLM, it can perform verification on the solution corresponding to each problem and obtain a score for the solution corresponding to each problem after evaluation based on the verification results.

[0126] S530 , the device configured with the LLM performs model training based on the multiple problem pairs and the scores of the solutions corresponding to each network problem to obtain the LLM.

[0127] In one possible implementation, a device equipped with an LLM performs model training based on multiple question pairs with the highest score as the goal to obtain an LLM. In the embodiment of the present application, the LLM obtained by training with the highest score as the goal has higher reliability.

[0128] Optionally, the LLM includes a network policy model, which is trained based on multiple question pairs.

[0129] Optionally, the LLM further includes a reward model, which is trained based on multiple problem pairs and scores of solutions corresponding to each network problem in the multiple problem pairs.

[0130] In an embodiment of the present application, the training data of the reward model includes the score of the solution corresponding to each network problem. Therefore, the reward model can output the score of the solution corresponding to each network problem, that is, evaluate the reliability of the solution corresponding to each network problem, so that the LLM model obtained by training based on the reward model is more reliable.

[0131] It should be understood that in the embodiment of the present application, by adjusting the parameters of the network policy model so that the result output by the network policy model has the highest score in the reward model, after repeated training, a network policy model based on the reward model, namely LLM, can be obtained.

[0132] In one possible implementation, when a new network problem exists in a real network, the operation and maintenance experts obtain a solution to the new network problem. The new network problem and / or the solution to the new network problem can be used as training data to retrain the network policy model, and / or the reward model, and / or the LLM.

[0133] The model training method provided in the embodiment of the present application is that a device configured with LLM inputs the solution corresponding to each network problem in multiple problem pairs to a device configured with a digital twin network, so that the device configured with the digital twin network simulates and executes the solution corresponding to each network problem, and obtains a score for the solution corresponding to each network problem, thereby realizing an automatic scoring mechanism and avoiding errors caused by scoring by operation and maintenance experts. Furthermore, the device configured with LLM uses multiple problem pairs to train the LLM model with the highest score as the goal. Since the scoring of the device configured with the digital twin network is introduced in the LLM model training process, the trained LLM model can be made more reliable.

[0134] Exemplarily, Figure 6 is a flow chart of the training of the LLM model in an embodiment of the present application. As shown in Figure 6, in step one, the device configured with LLM trains the network policy model based on multiple problem pairs; in step two, the device configured with LLM inputs the solution corresponding to each problem to the device configured with the digital twin network for simulation, and the scores obtained after simulation are used as data training reward models for multiple problem pairs and the scores of the solutions corresponding to each network problem in the multiple problem pairs; in step three, the device configured with LLM adjusts the parameters based on multiple problem pairs and the solutions corresponding to each problem through reinforcement learning (RLHF) and other methods to maximize the output solution score, thereby training LLM. Exemplarily, the process includes:

[0135] Step 1: Training the network policy model may include inputting network problems and their corresponding solutions into the network policy model for training. The network problem may be a random network problem in a network management dataset, or a network alarm problem. In one possible implementation, the solution to the network problem may be provided by an operations and maintenance expert. In another possible implementation, the solution to the network problem may be determined based on historical operational data or logs. For example, the solution may be the cause of the alarm problem.

[0136] Step 2: Training the reward model may include: 1. Randomly extracting a network problem Y from multiple problem pairs; 2. The device configured with LLM adjusts the parameters of the network policy model to obtain N solutions corresponding to the network problem Y; 3. The device configured with LLM inputs the N solutions into the device configured with the digital twin network, and the device configured with the digital twin network simulates the N solutions to obtain the scores of the N solutions; 4. The device configured with LLM uses the network problem Y, N solutions, and the scores of the N solutions to train the reward model; repeating the above steps can train the reward model.

[0137] Step 3: Training the LLM model may include: 1. The LLM-equipped device randomly selects a network problem X; 2. The LLM-equipped device adjusts the parameters of the network policy model to obtain M solutions corresponding to network problem X; 3. The LLM-equipped device inputs network problem X and its M corresponding solutions into the reward model, causing the reward model to output scores for the M reward solutions; 4. Through reinforcement learning from human feedback (RLHF), the goal is to maximize the score of the solution output by the network policy model. Repeated training is performed to obtain the LLM.

[0138] In an embodiment of the present application, a device equipped with an LLM can be trained based on multiple question pairs, aiming for the highest score, to ultimately obtain an LLM. In other words, the device equipped with an LLM adjusts the parameters of the network policy model so that the output results score high in the reward model. Through reinforcement learning, the output results of the network policy model receive the maximum reward, and the training is repeated continuously, ultimately obtaining an LLM. Based on the above approach, the reliability of the LLM can be improved.

[0139] Figure 7 is a schematic diagram of an example of a verification method for a network problem solution provided in an embodiment of the present application. The method is illustrated by taking the interaction between a device configured with a digital twin network and a device configured with a knowledge graph as an example. Of course, the subject that executes the action of the device configured with a digital twin network in the method can also be a device / module in a device configured with a digital twin network, such as a chip, processor, processing unit, etc. in a device configured with a digital twin network; the subject that executes the action of the device configured with a knowledge graph in the method can also be a device / module in a device configured with a knowledge graph, such as a chip, processor, processing unit, etc. in a device configured with a knowledge graph, and the embodiment of the present application does not make specific limitations on this. The processing performed by a single execution subject (for example, a device configured with a digital twin network or a device configured with a knowledge graph) in the embodiment of the present application can also be divided into executions by multiple execution subjects, and these execution subjects can be logically and / or physically separated. Exemplarily, as shown in Figure 7, method 700 includes:

[0140] S710: A device configured with a knowledge graph obtains an identifier of a service to which the first network problem belongs.

[0141] Illustratively, the service to which the first network problem belongs may be a verification service of the network problem.

[0142] In one possible implementation, a device configured with a knowledge graph receives an identifier of a business to which the first network problem belongs from a control center.

[0143] Optionally, the knowledge graph is pre-stored on a device configured with the knowledge graph, which is not limited in the embodiments of the present application.

[0144] Optionally, the control center receives a first network problem from a real network and sends an identifier of the service to which the first network problem belongs to a device configured with the knowledge graph. For example, the first network problem may be information such as an alarm, a user complaint, or an optimization request. It should be understood that other network problems may also fall within the scope of the first network problem, and this embodiment of the application is not limited to this.

[0145] S720: The device configured with the knowledge graph determines a solution corresponding to the first network problem from the knowledge graph according to an identifier of the business to which the first network problem belongs.

[0146] In the embodiment of the present application, the knowledge graph includes solutions to multiple network problems output by the LLM. The first network problem is one of the multiple network problems, and the first solution is one of the one or more solutions corresponding to the first network problem. It should be understood that the knowledge graph also includes other network problems and one or more solutions corresponding to other network problems, and the embodiment of the present application is not limited to this.

[0147] In a possible implementation of an embodiment of the present application, the LLM is obtained by training based on multiple problem pairs, where the multiple problem pairs include network problems in a network management data set and their corresponding solutions. This embodiment of the present application does not limit this.

[0148] In another possible implementation of the embodiment of the present application, the LLM is obtained by scoring and training based on multiple problem pairs and the solutions corresponding to each network problem. Exemplarily, the score is obtained after the solution corresponding to each problem in the model training method described in method 400 is input into a device simulation configured with a digital twin network; or, exemplarily, the score is obtained by the operation and maintenance experts scoring the solutions corresponding to each problem. The embodiment of the present application does not limit the source of the score. In the embodiment of the present application, the multiple problem pairs include network problems and their corresponding solutions in the network management data set, and the embodiment of the present application does not limit this.

[0149] To facilitate the reader's understanding, the knowledge graph is introduced as an example through Figures 8 and 9. Figure 8 is a schematic diagram of a topological structure of a network problem judgment decision tree in the knowledge graph. As shown in Figure 8, for example, the first-level nodes of the network problem judgment decision tree can indicate that the network problem is that the average SINR of the user is less than or equal to 0dB; the second-level nodes of the network problem judgment decision tree can indicate the reasons for the average SINR of the user being less than or equal to 0dB. For example, one of the reasons is a cell interference problem. Optionally, the weight of the cell interference problem in the network problem judgment decision tree can be 0.6 (the weight can represent the proportion of the cell interference problem among all the reasons causing the network problem, which will not be repeated in the following examples); for another example, one of the reasons is an interference problem caused by a sudden increase in the number of users. Optionally, the weight of the interference problem caused by the sudden increase in the number of users in the network problem judgment decision tree can be 0.2; for another example, one of the reasons is that there is no network problem. Optionally, the weight of the no network problem in the network problem judgment decision tree can be 0.01. The third-level nodes of the network problem judgment decision tree can represent more detailed causes of the causes in the second-level nodes. For example, for the cell interference type problems in the second-level nodes, one cause of the cell interference type problems is the inter-cell interference problem. Optionally, the weight of the inter-cell interference problem in the cell interference type problems is 0.7; for another example, one cause of the cell interference type problems is the internal cell interference problem. Optionally, the weight of the internal cell interference problem in the interference type problems is 0.2; for another example, one cause of the cell interference type problems is the new interference source problem. Optionally, the weight of the new interference source problem in the cell interference type problems is 0.1.

[0150] It should be understood that the first network problem in the embodiment of the present application can be a network problem corresponding to any layer of nodes in the network problem judgment decision tree, or, in other words, for example, the cause of the network problem of the first layer node in the second layer node can also be regarded as a network problem, and the embodiment of the present application does not limit this.

[0151] FIG9 is a schematic diagram of a topological structure of a decision tree for network problem solutions in a knowledge graph. As shown in FIG9 , for example, the network problem may be an inter-cell interference problem. The first-layer nodes of the decision tree for network problem solutions may indicate that the cause of the inter-cell interference problem is low power in the primary cell. The second-layer nodes of the decision tree for network problem solutions may indicate that one of the solutions to the problem of low power in the primary cell is to adopt a power control algorithm. Optionally, the weight of the power control algorithm in the decision tree for network problem solutions is 0.7. The second-layer nodes of the decision tree for network problem solutions may indicate that one of the solutions to the problem of low power in the primary cell is to adopt a load control algorithm. Optionally, the weight of the load control algorithm in the decision tree for network problem solutions is 0.2. The third-layer nodes of the decision tree for network problem solutions may indicate that the second-layer nodes A more detailed solution of the solution, for example, one of the more detailed solutions for the solution of using the power control algorithm in the second-layer node is a power control algorithm based on DQN, and optionally, the weight of the power control algorithm based on DQN in the power control algorithm is 0.7; for another example, one of the more detailed solutions for the solution of using the power control algorithm in the second-layer node is a power control algorithm based on a control channel, and optionally, the weight of the power control algorithm based on the control channel in the power control algorithm is 0.1; for another example, one of the more detailed solutions for the solution of using the power control algorithm in the second-layer node is a power control algorithm based on a data channel, and optionally, the weight of the power control algorithm based on the data channel in the power control algorithm is 0.1.

[0152] Exemplarily, the first solution corresponding to the first network problem in the embodiment of the present application may be the DQN-based power control algorithm shown in Figure 9; or, exemplary, the first solution corresponding to the first network problem may be the control channel-based power control algorithm shown in Figure 9; or, exemplary, the first solution corresponding to the first network problem may be the data channel-based power control algorithm shown in Figure 9, and the embodiment of the present application is not limited to this.

[0153] It should be noted that the blank nodes in the network problem judgment decision tree in the knowledge graph shown in FIG8 and the network problem solution decision tree in the knowledge graph shown in FIG9 can be completed by the solution output by the LLM, that is, the branches in the topology structure are added by the solution output by the LLM. For example, a network problem Z that is not in the knowledge graph appears in the real network. The network management center can input the network problem Z into the LLM of a device equipped with the LLM, and the LLM outputs the solution to the network problem Z. Furthermore, the device equipped with the LLM can input the network problem Z and the solution corresponding to the network problem Z into the device equipped with the knowledge graph. The device equipped with the knowledge graph can correspondingly complete the blanks in the knowledge graph. Optionally, the device equipped with the knowledge graph can adjust the weights corresponding to each node according to the network problem Z and the solution corresponding to the network problem Z output by the device equipped with the LLM.

[0154] S730: The device configured with the knowledge graph sends a first verification instruction to the device configured with the digital twin network. Correspondingly, the device configured with the digital twin network receives the first verification instruction from the device configured with the knowledge graph.

[0155] In an embodiment of the present application, the first verification instruction is used to instruct a device configured with a digital twin network to simulate and execute a first solution corresponding to the first network problem.

[0156] In one possible implementation, the first solution corresponding to the first network problem may be the solution that is most likely to solve the first network problem (for example, the solution with the highest weight in the knowledge graph). In this way, a device configured with a digital twin network can simulate fewer solutions, saving resource overhead for simulating solutions; the first solution corresponding to the first network problem may also be any one of one or more solutions corresponding to the first network problem, and this embodiment of the present application is not limited to this.

[0157] S740: The device configured with the digital twin network sends a first verification result to the device configured with the knowledge graph. Correspondingly, the device configured with the knowledge graph receives the first verification result from the device configured with the digital twin network.

[0158] In an embodiment of the present application, the first verification result is the result obtained after the device configured with the digital twin network executes the first solution corresponding to the first network problem.

[0159] Optionally, the first verification result indicates that the first network problem is resolved. For example, if the first network problem is caused by an internal cell interference problem and the average SINR of the user is less than or equal to 0dB, and after the device configured with the digital twin network uses a DQN-based power control algorithm for simulation, the average SINR of the user is greater than 0dB, then the first network problem can be considered to be resolved; alternatively, if the average SINR of the user is greater than a first threshold, then the first network problem can be considered to be resolved, and this embodiment of the present application is not limited to this.

[0160] Optionally, the first verification result indicates that the first network problem has not been resolved. For example, if the first network problem is an intra-cell interference problem causing an average user SINR of less than or equal to 0 dB, and after a device configured with a digital twin network performs simulation using a DQN-based power control algorithm, the average user SINR is still less than or equal to 0 dB, then the first network problem can be considered unresolved.

[0161] Exemplarily, the first verification result may be a specific value, for example, the average SINR value of the user is 5, and the device configured with the knowledge graph determines that the first network problem is solved based on the value; for another example, the average SINR value of the user is -5, and the device configured with the knowledge graph determines that the first network problem is not solved based on the value; or, exemplarily, the first verification result may be an indication information, and if the indication information is a positive acknowledgment (ACK), it indicates that the first network problem is solved; and if the indication information is a negative acknowledgment (NACK), it indicates that the first network problem is not solved.

[0162] Optionally, in an embodiment of the present application, the first verification instruction may include at least one of the following: an identifier of the service to which the first network problem belongs, an identifier of the first network problem, an identifier of the first solution, an area location identifier, a control instruction, an execution time, and configuration information corresponding to the first solution. Exemplarily, the area location identifier may be a location identifier of a cell, which may indicate a cell where a network problem, such as a network alarm problem, occurs; exemplarily, the control instruction may be an instruction to adjust the transmit power of the cell; exemplarily, the execution time may be the time for a device configured with a digital twin network to simulate the first solution; exemplarily, the configuration information corresponding to the first solution may be the information required for a device configured with a digital twin network to simulate the first solution. According to the configuration information corresponding to the first solution, the first solution may be digitally verified in the digital twin network. If the verification result indicates that the problem has not been solved, the solution may not be applied to the physical network, which may save resources for implementing the solution in the physical network.

[0163] Optionally, in an embodiment of the present application, the first verification result may include at least one of the following: an identifier of the service to which the first network problem belongs, an identifier of the first network problem, an identifier of the first solution, an identifier of the regional location, and network status information. Exemplarily, the network status information may include information such as the average SINR of the user.

[0164] In the verification method of the network problem solution provided in the embodiment of the present application, a device configured with a knowledge graph sends a first verification instruction to a device configured with a digital twin network, instructing the device configured with the digital twin network to simulate and execute the first solution corresponding to the first network device. After the device configured with the digital twin network simulates the first solution according to the first verification instruction, it sends a first verification result to the device configured with the knowledge graph, so that the device configured with the digital twin network simulates and verifies the first solution corresponding to the first network problem and sends the first verification result. In particular, since the first network problem and the first solution are both output by LLM, LLM has a powerful ability to output network problems and corresponding solutions, so that the device configured with the knowledge graph can use the network problems and corresponding solutions output by LLM to build a more complete knowledge graph. Furthermore, the solution in the knowledge graph is verified by the device configured with the digital twin network, and the verification result is fed back to the device configured with the knowledge graph. The knowledge graph in the device configured with the knowledge graph can be further updated, thereby improving the reliability of the solution corresponding to the network problem obtained by the user.

[0165] Furthermore, as shown in FIG7 , optionally, when the first verification result indicates that the first network problem is solved, the method for verifying the network problem solution provided in the embodiment of the present application further includes:

[0166] S751: The device configured with the knowledge graph sends first indication information to the control center. Correspondingly, the control center receives the first indication information from the device configured with the knowledge graph.

[0167] In the embodiment of the present application, the first indication information is used to instruct the control center to apply the first solution to the real network.

[0168] S752: The control center sends second feedback information to the device configured with the knowledge graph. Correspondingly, the device configured with the knowledge graph receives the second feedback information from the control center.

[0169] In an embodiment of the present application, the second feedback information is used to indicate the result after the first solution is applied to the real network. Exemplarily, there is a user average SINR less than or equal to 0db in the real network (i.e., the first network problem). After power control is performed using a DQN-based power control algorithm (i.e., the first solution corresponding to the first network problem) in the real network, the user average SINR is greater than 0db, which means that the result after the first solution is applied to the real network can be obtained (for example, the result is that the network problem in the real network is solved). Or, exemplarily, there is a user average SINR less than or equal to 0db in the real network (i.e., the first network problem). After power control is performed using a DQN-based power control algorithm (i.e., the first solution corresponding to the first network problem) in the real network, the user average SINR is still less than or equal to 0db, which means that the result after the first solution is applied to the real network can be obtained (the result is not good, or the result is that the network problem in the real network is not solved).

[0170] S753, the device equipped with the knowledge graph updates the knowledge graph according to the second feedback information.

[0171] Exemplarily, the device configured with the knowledge graph updates the knowledge graph according to the second feedback information, and can increase or decrease the corresponding weight of the first solution in the knowledge graph according to the second feedback information for the device configured with the knowledge graph. For example, as shown in the examples in Figure 8 or Figure 9, the second feedback information indicates that the first network problem is solved after the first solution is applied to the real network, then the weight corresponding to the first solution in the knowledge graph is increased; optionally, when the device configured with the knowledge graph receives the second feedback information indicating that the first network problem is solved after the first solution is applied to the real network for a certain number of times, the device configured with the knowledge graph increases the weight corresponding to the first solution in the knowledge graph. For example, when the second feedback information indicating that the first network problem is solved after the first solution is applied to the real network reaches 1000 times, the device configured with the knowledge graph increases the weight corresponding to the first solution in the knowledge graph by 0.1; or, optionally, each time the device configured with the knowledge graph receives the second feedback information indicating that the first network problem is solved after the first solution is applied to the real network, the device configured with the knowledge graph increases the weight corresponding to the first solution in the knowledge graph. For example, each time the device configured with the knowledge graph receives the second feedback information indicating that the first network problem is solved after the first solution is applied to the real network, the device configured with the knowledge graph increases the weight corresponding to the first solution in the knowledge graph by 0.0001. The description of reducing the weight corresponding to the first solution in the knowledge graph by a device configured with a knowledge graph is similar to this and will not be repeated here.

[0172] Exemplarily, a device configured with a knowledge graph updates the knowledge graph according to the second feedback information, and the topological structure of the knowledge graph can be updated for the device configured with the knowledge graph according to the second feedback information. For example, as shown in Figure 8 or Figure 9, if the second feedback information indicates that the first network problem is not solved after the first solution is applied to the real network, the branch corresponding to the first solution in the knowledge graph is removed; optionally, when the device configured with the knowledge graph receives a certain number of second feedback information indicating that the first network problem is not solved after the first solution is applied to the real network, the device configured with the knowledge graph removes the branch corresponding to the first solution in the knowledge graph. For example, when the second feedback information indicating that the first network problem is not solved after the first solution is applied to the real network reaches 1000 times, the device configured with the knowledge graph removes the branch corresponding to the first solution in the knowledge graph.

[0173] Alternatively, as shown in FIG7 , optionally, when the first verification result indicates that the first network problem is not resolved, the method for verifying the network problem solution provided in the embodiment of the present application further includes:

[0174] S761, the device configured with the knowledge graph determines a second solution corresponding to the first network problem from the knowledge graph according to the identifier of the business to which the first network problem belongs.

[0175] In the embodiment of the present application, the second solution is one of the one or more solutions corresponding to the first network problem.

[0176] S762: The device configured with the knowledge graph sends a second verification instruction to the device configured with the digital twin network. Correspondingly, the device configured with the digital twin network receives the second verification instruction from the device configured with the knowledge graph.

[0177] In an embodiment of the present application, the second verification instruction is used to instruct a device configured with a digital twin network to simulate and execute a second solution corresponding to the first network problem.

[0178] S763: The device configured with the digital twin network sends a second verification result to the device configured with the knowledge graph. Correspondingly, the device configured with the knowledge graph receives the second verification result from the device configured with the digital twin network.

[0179] In an embodiment of the present application, the second verification result is the result obtained after the device configured with the digital twin network executes the second solution corresponding to the first network problem.

[0180] It should be understood that steps S761 to S763 can be repeated until the verification result sent by the device configured with the digital twin network to the device configured with the knowledge graph indicates that the first network problem is solved, or the verification result indicates that the first network problem is not solved but the number of simulations of the device configured with the digital twin network reaches a second threshold.

[0181] Exemplarily, a device configured with a knowledge graph can update the knowledge graph according to the second verification result. Specifically, the topological structure of the knowledge graph is updated according to the second verification result. For how a device configured with a knowledge graph updates the knowledge graph, please refer to the above step S753, which will not be repeated here.

[0182] Optionally, as shown in FIG7 , the verification method for the network problem solution provided in the embodiment of the present application may further include:

[0183] S770: The device configured with the knowledge graph sends first feedback information to the control center. Correspondingly, the control center receives the first feedback information from the device configured with the knowledge graph.

[0184] In the embodiment of the present application, the first feedback information is used to indicate that the verification service corresponding to the first network problem is completed.

[0185] In one possible implementation, the completion of the verification service corresponding to the first network problem can mean that the first network problem is solved, and the control center can be informed that the verification service is completed and instruct the operation of the physical network based on the first feedback information; in another possible implementation, the completion of the verification service corresponding to the first network problem can be the verification result of the device configured with the digital twin network, indicating that the first network problem is not solved but the number of simulations of the device configured with the digital twin network reaches a second threshold, which can make the control center aware that the solution corresponding to the verified first network problem cannot be applied to the physical network, thereby avoiding the waste of resources caused by applying unfeasible solutions to the physical network.

[0186] Optionally, as shown in FIG7 , the verification method for the network problem solution provided in the embodiment of the present application may further include:

[0187] S780, the device configured with the knowledge graph updates the knowledge graph according to the first verification result.

[0188] Exemplarily, a device configured with a knowledge graph updates the knowledge graph according to the first verification result, and may increase or decrease the weight corresponding to the first solution in the knowledge graph according to the first verification result; or, exemplary, a device configured with a knowledge graph updates the knowledge graph according to the first verification result, and may update the topological structure of the knowledge graph according to the first verification result. For how a device configured with a knowledge graph updates the knowledge graph, please refer to the above-mentioned step S753, which will not be repeated here.

[0189] Optionally, the device configured with the LLM may retrain the LLM according to the first verification result, so that the solution output by the retrained LLM may be more reliable.

[0190] It should be noted that step S780 is performed after step S740 , and the order of step S780 and other optional steps except steps S710 to S740 is not limited.

[0191] To facilitate the reader's understanding, the following will introduce the verification method of the network problem solution provided by the embodiment of the present application with a specific example. The example includes:

[0192] 1. In a real network, the average SINR of the user is less than or equal to 0dB. For example, a network element in the core network sends an alarm message to the control center, indicating that the average SINR of the user is less than or equal to 0dB.

[0193] 2. The control center receives the alarm information and generates an identifier of the service to which the first network problem corresponding to the alarm information belongs, such as IDa. The control center sends IDa to the device configured with the knowledge graph. Accordingly, the device configured with the knowledge graph analyzes the network problem corresponding to IDa, that is, the reason why the user's average SINR is less than or equal to 0db, based on IDa. Combining various types of data, it is analyzed that the cause is most likely an inter-cell interference problem, and the cause is set as the first network problem ID1; the device configured with the knowledge graph determines the first solution corresponding to the first network problem ID1 from the knowledge graph corresponding to the first network problem ID1 (for example, using the DQN power control algorithm to control power) and sets the first solution to ID1.

[0194] 3. The device configured with the knowledge graph sends a first verification instruction to the device configured with the digital twin network, wherein the first verification instruction includes IDa, ID1, and id1. Optionally, the first verification instruction also includes at least one of the area location identifier (i.e., cell location identifier), control instruction (i.e., using a power control algorithm to control power), execution time (i.e., the time to execute the control instruction), and information collection setting information (i.e., relevant data of the cell involved in the first solution).

[0195] 4. The device configured with the digital twin network simulates the first solution corresponding to id1.

[0196] 5. The device configured with the digital twin network sends a first verification result to the device configured with the knowledge graph. Optionally, the first verification result includes a result indicating whether the first solution corresponding to id1 solves the problem of the user's average SINR being less than or equal to 0db after being simulated by the device configured with the digital twin network; wherein, in one possible implementation, the first verification result indicates that the user's average SINR is greater than 0db after being simulated by the device configured with the digital twin network, and the device configured with the knowledge graph can increase the weight of the first solution in the knowledge graph according to the first verification result; or, the first verification result indicates that the user's average SINR is less than or equal to 0db after being simulated by the device configured with the digital twin network, and the device configured with the knowledge graph can reduce the weight of the first solution in the knowledge graph according to the first verification result.

[0197] 6. In step 5, if the first verification result indicates that the average SINR of the user after simulation by the device configured with the digital twin network is less than or equal to 0db, the device configured with the knowledge graph determines the second solution corresponding to the second network problem ID2 from the knowledge graph corresponding to the first network problem ID1 (i.e., power is controlled by a control channel power control algorithm), and sets the second solution to id2; the device configured with the knowledge graph sends a second verification instruction to the device configured with the digital twin network to obtain a second verification result. The process is similar to steps 3 to 5 and will not be repeated here.

[0198] 7. After the first verification results of steps 3 to 5 indicate that the average SINR of the user is greater than 0db after simulation of the device configured with the digital twin network, or, after the first verification results of steps 3 to 5 indicate that the average SINR of the user is less than or equal to 0db after simulation of the device configured with the digital twin network, and then after one or more steps 6, the verification results indicate that the average SINR of the user is greater than 0db after simulation of the device configured with the digital twin network, or, after simulation of the device configured with the digital twin network, the average SINR of the user is still less than or equal to 0db but the number of steps 6 exceeds the first threshold, the device configured with the knowledge graph sends a first feedback message to the control center, and the first feedback message indicates that the verification service corresponding to IDa is completed.

[0199] 8. If, after simulation by a device configured with a digital twin network, the DQN power control algorithm is used to control power so that the average SINR of the user is greater than 0 dB, the device configured with the knowledge graph may send a first indication message to the control center. The first indication message is used to indicate the use of the DQN power control algorithm to control power. Further, the control center may instruct the real network to apply the DQN power control algorithm to control power in solving the network problem where the average SINR of the user is less than or equal to 0 dB.

[0200] 9. If the DQN power control algorithm is used to control power in a real network and the average SINR of users is greater than 0db, the real network can provide feedback to the control center. Furthermore, the control center sends a second feedback information to the device configured with the knowledge graph. The second feedback information is used to indicate that the DQN power control algorithm is used to control power in the real network, and the average SINR of users is greater than 0db. Furthermore, the device configured with the knowledge graph can increase the weight of the first solution corresponding to id1, that is, the DQN power control algorithm, in the knowledge graph according to the second feedback information.

[0201] The above method 500 mainly introduces the solution provided by the embodiment of the present application from the perspective of the interaction between the device configured with LLM and the device configured with the digital twin network, and the control center. Accordingly, the embodiment of the present application also provides a communication device, which is used to implement the above various methods. The communication device can be a device configured with LLM in the above method embodiment, or a device including the device configured with LLM, or a component that can be used to configure a device with LLM; or, the communication device can be a device configured with a digital twin network in the above method embodiment, or a device including the device configured with digital twin network, or a component that can be used to configure a device with digital twin network. It can be understood that in order to implement the above functions, the communication device includes hardware structures and / or software modules corresponding to the execution of each function. It should be easy for those skilled in the art to realize that, in combination with the units and algorithm steps of the various examples described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in a hardware or computer software-driven hardware manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0202] In the embodiment of the present application, the communication device can be divided into functional modules according to the embodiment of the above-mentioned method 500. For example, each functional module can be divided according to each function, or two or more functions can be integrated into one processing module. The above-mentioned integrated modules can be implemented in the form of hardware or in the form of software functional modules. It should be understood that the division of modules in the embodiment of the present application is schematic and is only a logical functional division. In actual implementation, other division methods may be used.

[0203] For example, FIG10 is a schematic diagram of a communication device provided in an embodiment of the present application. Taking the communication device as a device configured with an LLM in the above method embodiment (which may be a chip of a device configured with an LLM, or a module of a device configured with an LLM, or an internal device of a device configured with an LLM) as an example, the device configured with an LLM includes a transceiver module 1010 and a processing module 1020. The transceiver module 1010, which may also be referred to as a transceiver unit, is used to implement transceiver functions, and may be, for example, a transceiver circuit, a transceiver, a transceiver, or a communication interface.

[0204] In one solution, the transceiver module 1010 is used to input a solution corresponding to each network problem in a plurality of problem pairs to a device configured with a digital twin network, where the plurality of problem pairs include network problems and their corresponding solutions in a network management data set.

[0205] The transceiver module 1010 is also used to receive the score of the solution corresponding to each network problem from the device configured with the digital twin network, wherein the score is obtained by the device configured with the digital twin network verifying the solution corresponding to each problem and evaluating it based on the verification results.

[0206] The processing module 1020 is used to perform model training based on multiple question pairs and the scores of the solutions corresponding to each network problem to obtain a large language model (LLM).

[0207] In an embodiment of the present application, the LLM includes a network strategy model, which is obtained by training based on multiple question pairs.

[0208] In an embodiment of the present application, the LLM includes a reward model, which is trained based on multiple problem pairs and the scores of the solutions corresponding to each network problem in the multiple problem pairs.

[0209] All relevant contents of each step involved in the above method embodiment can be referred to the functional description of the corresponding functional module and will not be repeated here. Optionally, the communication device may further include a storage module 1030, which may be used to store instructions and / or data, and the processing module 1020 may read the instructions and / or data in the storage module 1030.

[0210] In the embodiments of the present application, the device configured with the LLM is presented in the form of various functional modules divided in an integrated manner. "Module" here can refer to a specific ASIC, circuit, processor and memory that executes one or more software or firmware programs, integrated logic circuit, and / or other devices that can provide the above-mentioned functions. In one embodiment, those skilled in the art will appreciate that the device configured with the LLM can take the form of the communication device 400 shown in Figure 4.

[0211] Since the device configured with LLM provided in the embodiment of the present application (which may be a chip of a device configured with LLM, or a module of a device configured with LLM, or an internal device of a device configured with LLM) can execute the above-mentioned model training method, the technical effects that can be obtained can be referred to the above-mentioned method embodiment and will not be repeated here.

[0212] Alternatively, for example, FIG10 is a schematic diagram of a communication device provided in an embodiment of the present application. Taking the communication device as a device configured with a digital twin network in the above method embodiment (which may be a chip of a device configured with a digital twin network, or a module of a device configured with a digital twin network, or an internal device of a device configured with a digital twin network) as an example, the device configured with a digital twin network includes a transceiver module 1010 and a processing module 1020. The transceiver module 1010, which may also be referred to as a transceiver unit for implementing a transceiver function, may be, for example, a transceiver circuit, a transceiver, a transceiver, or a communication interface.

[0213] In one solution, the transceiver module 1010 is configured to receive a solution corresponding to each network problem in a plurality of problem pairs from a device configured with LLM, where the plurality of problem pairs include network problems and their corresponding solutions in a network management dataset.

[0214] The transceiver module 1010 is also used to send the score of the solution corresponding to each network problem to the device configured with LLM, wherein the score is obtained after the device configured with the digital twin network verifies the solution corresponding to each problem and evaluates it based on the verification results.

[0215] All relevant contents of each step involved in the above method embodiment can be referred to the functional description of the corresponding functional module and will not be repeated here. Optionally, the communication device may further include a storage module 1030, which may be used to store instructions and / or data, and the processing module 1020 may read the instructions and / or data in the storage module 1030.

[0216] In an embodiment of the present application, the device configured with a digital twin network is presented in the form of various functional modules divided in an integrated manner. The "module" here can refer to a specific ASIC, circuit, processor and memory that executes one or more software or firmware programs, integrated logic circuit, and / or other devices that can provide the above functions. In one embodiment, those skilled in the art can imagine that the device configured with a digital twin network can take the form of a communication device 400 shown in Figure 4.

[0217] Since the device configured with a digital twin network provided in the embodiment of the present application (which may be a chip of a device configured with a digital twin network, or a module of a device configured with a digital twin network, or an internal device of a device configured with a digital twin network) can execute the above-mentioned model training method, the technical effects that can be obtained can be referred to the above-mentioned method embodiments and will not be repeated here.

[0218] The above method 700 mainly introduces the solution provided by the embodiment of the present application from the perspective of the interaction between a device configured with a knowledge graph and a device configured with a digital twin network. Accordingly, the embodiment of the present application also provides a communication device, which is used to implement the above various methods. The communication device can be a device configured with a knowledge graph in the above method embodiment, or a device including the above device configured with a knowledge graph, or a component that can be used to configure a device with a knowledge graph; or, the communication device can be a device configured with a digital twin network in the above method embodiment, or a device including the above device configured with a digital twin network, or a component that can be used to configure a device with a digital twin network. It can be understood that in order to implement the above functions, the communication device includes hardware structures and / or software modules corresponding to the execution of each function. It should be easy for those skilled in the art to realize that, in combination with the units and algorithm steps of the various examples described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in a hardware or computer software-driven hardware manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0219] In the embodiment of the present application, the communication device can be divided into functional modules according to the embodiment of the above-mentioned method 700. For example, each functional module can be divided according to each function, or two or more functions can be integrated into one processing module. The above-mentioned integrated module can be implemented in the form of hardware or in the form of software functional modules. It should be understood that the division of modules in the embodiment of the present application is schematic and is only a logical functional division. In actual implementation, there may be other division methods.

[0220] For example, FIG10 is a schematic diagram of a communication device provided in an embodiment of the present application. Taking the communication device as a device configured with a knowledge graph in the above method embodiment (which may be a chip of a device configured with a knowledge graph, or a module of a device configured with a knowledge graph, or an internal device of a device configured with a knowledge graph) as an example, the device configured with a knowledge graph includes a transceiver module 1010 and a processing module 1020. The transceiver module 1010, which may also be referred to as a transceiver unit for implementing a transceiver function, may be, for example, a transceiver circuit, a transceiver, a transceiver, or a communication interface.

[0221] In one solution, the transceiver module 1010 is configured to receive an identifier of a service to which the first network problem belongs from a control center.

[0222] Processing module 1020 is used to determine a first solution corresponding to the first network problem from a knowledge graph based on an identifier of the first network problem; wherein the knowledge graph includes solutions to multiple network problems output by a large language model (LLM), the first network problem is one of the multiple network problems, and the first solution is one of the one or more solutions corresponding to the first network problem.

[0223] The transceiver module 1010 is further configured to send a first verification instruction to a device configured with a digital twin network, where the first verification instruction is used to instruct the device configured with the digital twin network to simulate and execute a first solution corresponding to the first network problem;

[0224] The transceiver module 1010 is also used to receive a first verification result from a device configured with a digital twin network. The first verification result is a result obtained after the device configured with the digital twin network executes a first solution corresponding to a first network problem.

[0225] All relevant contents of each step involved in the above method embodiment can be referred to the functional description of the corresponding functional module and will not be repeated here. Optionally, the communication device may further include a storage module 1030, which may be used to store instructions and / or data, and the processing module 1020 may read the instructions and / or data in the storage module 1030.

[0226] In an embodiment of the present application, the device configured with the knowledge graph is presented in the form of various functional modules divided in an integrated manner. The "module" here can refer to a specific ASIC, circuit, processor and memory that executes one or more software or firmware programs, integrated logic circuit, and / or other devices that can provide the above functions. In one embodiment, those skilled in the art can imagine that the device configured with the knowledge graph can take the form of the communication device 400 shown in Figure 4.

[0227] Since the device configured with a knowledge graph provided in the embodiment of the present application (which may be a chip of a device configured with a knowledge graph, or a module of a device configured with a knowledge graph, or an internal device of a device configured with a knowledge graph) can execute the above-mentioned method for verifying the solution to the network problem, the technical effects that can be obtained can be referred to the above-mentioned method embodiment and will not be repeated here.

[0228] In one solution, the transceiver module 1010 is used to receive a first verification instruction from a device configured with a knowledge graph, where the first verification instruction is used to instruct the device configured with a digital twin network to simulate and execute a first solution corresponding to a first network problem, wherein the first solution corresponding to the first network problem is determined from the knowledge graph by the device configured with the knowledge graph based on an identifier of a business to which the first network problem belongs, and the knowledge graph includes solutions to multiple network problems output by a large language model LLM, the first network problem is one of the multiple network problems, and the first solution is one of the one or more solutions corresponding to the first network problem.

[0229] The transceiver module 1010 is also used to send a first verification result to a device configured with a knowledge graph. The first verification result is the result obtained after the device configured with a digital twin network executes a first solution corresponding to the first network problem.

[0230] All relevant contents of each step involved in the above method embodiment can be referred to the functional description of the corresponding functional module and will not be repeated here. Optionally, the communication device may further include a storage module 1030, which may be used to store instructions and / or data, and the processing module 1020 may read the instructions and / or data in the storage module 1030.

[0231] In an embodiment of the present application, the device configured with the LLM is presented in the form of various functional modules divided in an integrated manner. The "module" here can refer to a specific ASIC, circuit, processor and memory that executes one or more software or firmware programs, integrated logic circuit, and / or other devices that can provide the above functions. In one embodiment, those skilled in the art can imagine that the device configured with the digital twin network can take the form of the communication device 400 shown in Figure 4.

[0232] Since the device configured with a digital twin network provided in the embodiment of the present application (which may be a chip of a device configured with a digital twin network, or a module of a device configured with a digital twin network, or an internal device of a device configured with a digital twin network) can execute the above-mentioned model training method, the technical effects that can be obtained can be referred to the above-mentioned method embodiments and will not be repeated here.

[0233] The processor 401 in the communication device 400 shown in FIG4 can call the computer-executable instructions stored in the memory 403 to enable the communication device 400 to execute the verification method of the network problem solution in the above method embodiment.

[0234] Specifically, the functions / implementation processes of the transceiver module 1010 and the processing module 1020 in FIG10 can be implemented by the processor 401 in the communication device 400 shown in FIG4 calling computer-executable instructions stored in the memory 403. Alternatively, the functions / implementation processes of the processing module 1020 in FIG10 can be implemented by the processor 401 in the communication device 400 shown in FIG4 calling computer-executable instructions stored in the memory 403, and the functions / implementation processes of the transceiver module 1010 in FIG10 can be implemented by the communication interface 404 in the communication device 400 shown in FIG4.

[0235] It should be understood that one or more of the above modules or units can be implemented by software, hardware, or a combination of the two. When any of the above modules or units is implemented in software, the software exists in the form of computer program instructions and is stored in a memory, and a processor can be used to execute the program instructions and implement the above method flow. The processor can be built into an SoC or ASIC, or it can be an independent semiconductor chip. In addition to the core used to execute software instructions to perform calculations or processing within the processor, it can further include necessary hardware accelerators, such as field programmable gate arrays (FPGAs), programmable logic devices (PLDs), or logic circuits that implement dedicated logic operations.

[0236] When the above modules or units are implemented in hardware, the hardware can be any one or any combination of a CPU, a microprocessor, a digital signal processing (DSP) chip, a microcontroller unit (MCU), an artificial intelligence processor, an ASIC, a SoC, an FPGA, a PLD, a dedicated digital circuit, a hardware accelerator or a non-integrated discrete device, which can run the necessary software or not rely on the software to execute the above method flow.

[0237] Optionally, an embodiment of the present application further provides a communication device (for example, the communication device may be a chip or a chip system), which includes a processor for implementing the method in any of the above method embodiments. In one possible design, the communication device also includes a memory. The memory is used to store necessary program instructions and data, and the processor can call the program code stored in the memory to instruct the communication device to execute the method in any of the above method embodiments. Of course, the memory may not be in the communication device. When the communication device is a chip system, it may be composed of a chip, or it may include a chip and other discrete devices, which is not specifically limited in the embodiment of the present application.

[0238] Optionally, an embodiment of the present application also provides a computer-readable storage medium, which stores a computer program or instruction. When the computer program or instruction is run on a communication device, the communication device can execute the method described in any of the above method embodiments or any of its implementation methods.

[0239] Optionally, an embodiment of the present application further provides a communication system, which includes the first network device described in the above method embodiment and the terminal device described in the above method embodiment.

[0240] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware, or any combination thereof. When implemented using a software program, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions according to the embodiments of the present application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more media integrated therein. The available media may be magnetic media (eg, floppy disks, hard disks, magnetic tapes), optical media (eg, DVDs), or semiconductor media (eg, solid state drives (SSDs)).

[0241] Although the present application is described herein in conjunction with various embodiments, in the process of implementing the claimed application, those skilled in the art may understand and implement other variations of the disclosed embodiments by reviewing the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude multiple situations. A single processor or other unit may implement several functions listed in the claims. Certain measures are recorded in different dependent claims, but this does not mean that these measures cannot be combined to produce good results.

[0242] Although the present application has been described with reference to specific features and embodiments thereof, it is apparent that various modifications and combinations may be made thereto without departing from the scope of the present application. Accordingly, this specification and the drawings are merely illustrative of the present application as defined by the appended claims and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, the present application is intended to encompass such modifications and variations as fall within the scope of the claims of the present application and their equivalents.

Claims

1. A method for verifying a solution to a network problem, characterized in that: The method comprises: Obtaining an identifier of the service to which the first network problem belongs; Determine, from a knowledge graph, a first solution corresponding to the first network problem according to an identifier of a service to which the first network problem belongs; wherein the knowledge graph includes solutions to multiple network problems output by a large language model (LLM), the first network problem is one of the multiple network problems, and the first solution is one of the one or more solutions corresponding to the first network problem; Sending a first verification instruction to a device configured with a digital twin network, wherein the first verification instruction is used to instruct the device configured with the digital twin network to simulate and execute the first solution corresponding to the first network problem; Receive a first verification result from the device configured with the digital twin network, where the first verification result is a result obtained after the device configured with the digital twin network executes the first solution corresponding to the first network problem.

2. The method according to claim 1, characterized in that The obtaining the identifier of the service to which the first network problem belongs includes: An identifier of the business to which the first network problem belongs is received from a control center.

3. The method according to claim 1 or 2, characterized in that: The first verification result indicates that the first network problem is solved, and the method further includes: Sending first feedback information to the control center, where the first feedback information is used to indicate that the verification service corresponding to the first network problem is completed.

4. The method according to any one of claims 1 to 3, characterized in that The first verification result indicates that the first network problem is not resolved, and the method further includes: Determine, from the knowledge graph, a second solution corresponding to the first network problem according to an identifier of the service to which the first network problem belongs; wherein the second solution is one of the one or more solutions corresponding to the first network problem; Sending a second verification instruction to the device configured with the digital twin network, wherein the second verification instruction is used to instruct the device configured with the digital twin network to simulate and execute the second solution corresponding to the first network problem; Receive a second verification result from the device configured with the digital twin network, where the second verification result is a result obtained after the device configured with the digital twin network executes the second solution corresponding to the first network problem.

5. The method according to any one of claims 1 to 4, characterized in that The method further comprises: Update the knowledge graph according to the first verification result.

6. The method according to claim 5, characterized in that The updating of the knowledge graph according to the first verification result includes: Increase or decrease the weight corresponding to the first solution in the knowledge graph according to the first verification result, or update the topological structure of the knowledge graph according to the first verification result.

7. The method according to any one of claims 1 to 6, characterized in that The LLM is trained based on multiple problem pairs, or the LLM is trained based on multiple problem pairs and the scores of the solutions corresponding to each network problem, and the scores are obtained after inputting the solutions corresponding to each problem into the simulation of the device configured with the digital twin network, and the multiple problem pairs include network problems in the network management data set and their corresponding solutions.

8. The method according to any one of claims 1 to 7, characterized in that The first verification result indicates that the first network problem is solved, and the method further includes: Sending first instruction information to the control center, where the first instruction information is used to instruct the control center to apply the first solution to a real network; receiving second feedback information from the control center, where the second feedback information is used to indicate a result of applying the first solution to a real network; Update the knowledge graph according to the second feedback information.

9. The method according to any one of claims 1 to 8, characterized in that The first network problem includes at least one of the following: a network alarm problem, a user complaint problem, and an optimization demand problem.

10. The method according to any one of claims 1 to 9, characterized in that The first verification instruction includes at least one of the following: an identifier of the service to which the first network problem belongs, an identifier of the first network problem, an identifier of the first solution, Area location identification, control instructions, execution time, and information collection settings; The first verification result includes at least one of the following: The identifier of the service to which the first network problem belongs, the identifier of the first network problem, the identifier of the first solution, the regional location identifier, and network status information.

11. A verification system for a network problem solution, characterized in that: include: Devices configured with knowledge graphs and devices configured with digital twin networks; The device configured with the knowledge graph is used to obtain an identifier of a business to which the first network problem belongs, and determine a first solution corresponding to the first network problem from the knowledge graph according to the identifier of the business to which the first network problem belongs; wherein the knowledge graph includes solutions to multiple network problems output by a large language model LLM, the first network problem is one of the multiple network problems, and the first solution is one of the one or more solutions corresponding to the first network problem; The device configured with the knowledge graph is further used to send a first verification instruction to the device configured with the digital twin network, wherein the first verification instruction is used to instruct the device configured with the digital twin network to simulate and execute the first solution corresponding to the first network problem; The device configured with the digital twin network is used to receive a first verification instruction from the device configured with the knowledge graph, and after simulating the execution of the first solution according to the first verification instruction, send a first verification result to the device configured with the knowledge graph, wherein the first verification result is a result obtained after the device configured with the digital twin network executes the first solution corresponding to the first network problem; The device configured with the knowledge graph is also used to receive a first verification result from the device configured with the digital twin network.

12. The system according to claim 11, characterized in that The first verification result indicates that the first network problem is resolved; The device configured with the knowledge graph is also used to send first feedback information to the control center, and the first feedback information is used to indicate that the verification service corresponding to the first network problem is completed.

13. The system according to claim 11 or 12, characterized in that: The first verification result indicates that the first network problem is not resolved; The device configured with the knowledge graph is further used to determine a second solution corresponding to the first network problem from the knowledge graph according to the identifier of the first network problem; wherein the second solution is one of the one or more solutions corresponding to the first network problem; The device configured with the knowledge graph is further used to send a second verification instruction to the device configured with the digital twin network, wherein the second verification instruction is used to instruct the device configured with the digital twin network to simulate and execute the second solution corresponding to the first network problem; The device configured with the digital twin network is further configured to receive the second verification instruction from the device configured with the knowledge graph, and after simulating the execution of the second solution according to the second verification instruction, send a second verification result to the device configured with the knowledge graph; The device configured with the knowledge graph is also used to receive a second verification result from the device configured with the digital twin network.

14. The system according to any one of claims 11 to 13, characterized in that The device configured with the knowledge graph is also used to update the knowledge graph according to the first verification result, and / or update the LLM according to the first verification result.

15. The system according to claim 14, characterized in that The device configured with the knowledge graph is further used to update the knowledge graph according to the first verification result, including: The device configured with the knowledge graph is also used to increase or decrease the weight corresponding to the first solution in the knowledge graph according to the first verification result, or to update the topological structure of the knowledge graph according to the first verification result.

16. The system according to any one of claims 11 to 15, characterized in that The LLM is trained based on multiple problem pairs, or the LLM is trained based on multiple problem pairs and the scores of the solutions corresponding to each problem, and the scores are obtained after inputting the solutions corresponding to each problem into the simulation of the device configured with the digital twin network, and the multiple problem pairs include network problems in the network management data set and their corresponding solutions.

17. The system according to any one of claims 11 to 16, characterized in that The first verification result indicates that the first network problem is resolved; The device configured with the knowledge graph is further used to send first instruction information to the control center, where the first instruction information is used to instruct the control center to apply the first solution to a real network; The device configured with the knowledge graph is also used to receive second feedback information from the control center and update the knowledge graph according to the second feedback information, wherein the second feedback information is used to indicate the result after the first solution is applied to a real network.

18. A system according to any one of claims 11 to 17, characterized in that The first network problem includes at least one of the following: a network alarm problem, a user complaint problem, and an optimization demand problem.

19. The system according to any one of claims 11 to 18, characterized in that The first verification instruction includes at least one of the following: an identifier of the service to which the first network problem belongs, an identifier of the first network problem, an identifier of the first solution, an area location identifier, a control instruction, an execution time, and information collection setting information; The first verification result includes at least one of the following: The identifier of the service to which the first network problem belongs, the identifier of the first network problem, the identifier of the first solution, the regional location identifier, and network status information.

20. A model training method, characterized in that: include: Inputting a solution corresponding to each network problem in a plurality of problem pairs to a device configured with a digital twin network, wherein the plurality of problem pairs include network problems in a network management data set and their corresponding solutions; Receive a score of a solution corresponding to each network problem of the device configured with the digital twin network, wherein the score is obtained by the device configured with the digital twin network verifying the solution corresponding to each problem and evaluating the verification result; Based on the multiple question pairs and the scores of the solutions corresponding to each network problem, model training is performed to obtain a large language model LLM.

21. The method according to claim 20, characterized in that The LLM includes a network strategy model, and the network strategy model is trained based on the multiple question pairs.

22. The method according to claim 20 or 21, characterized in that The LLM includes a reward model, which is trained based on the multiple problem pairs and the scores of the solutions corresponding to each network problem in the multiple problem pairs.

23. A communication device, characterized in that: include: A transceiver module, used to obtain an identifier of the service to which the first network problem belongs; a processing module, configured to determine, from a knowledge graph, a first solution corresponding to the first network problem according to an identifier of the first network problem; wherein the knowledge graph includes solutions to multiple network problems output by a large language model (LLM), the first network problem is one of the multiple network problems, and the first solution is one of the one or more solutions corresponding to the first network problem; The transceiver module is further used to send a first verification instruction to a device configured with a digital twin network, wherein the first verification instruction is used to instruct the device configured with the digital twin network to simulate and execute the first solution corresponding to the first network problem; The transceiver module is also used to receive a first verification result from the device configured with the digital twin network, and the first verification result is a result obtained after the device configured with the digital twin network executes the first solution corresponding to the first network problem.

24. A communication device, characterized in that: include: A transceiver module, used to input a solution corresponding to each network problem in a plurality of problem pairs to a device configured with a digital twin network, wherein the plurality of problem pairs include network problems in a network management data set and their corresponding solutions; A transceiver module, used to receive a score of a solution corresponding to each network problem of the device configured with the digital twin network, wherein the score is obtained after the device configured with the digital twin network verifies the solution corresponding to each problem and evaluates the verification result; The processing module is used to perform model training based on the multiple problem pairs and the scores of the solutions corresponding to each network problem to obtain a large language model LLM.

25. A communication device, characterized in that: The communication device comprises means for executing the method according to any one of claims 1 to 10.

26. A communication device, characterized in that: The communication device comprises means for performing the method according to any one of claims 20 to 22.

27. A computer-readable storage medium, characterized in that: The computer-readable storage medium comprises instructions, which, when executed, enable the method according to any one of claims 1 to 10 to be implemented.

28. A computer-readable storage medium, characterized in that: The computer-readable storage medium comprises instructions which, when executed, cause the method according to any one of claims 20 to 22 to be implemented.

29. A computer program product, characterized in that The computer program product comprises instructions which, when executed, cause the method according to any one of claims 1 to 10 to be implemented.

30. A computer program product, characterized in that The computer program product comprises instructions which, when executed, cause the method according to any one of claims 20 to 22 to be implemented.

Images