Inference method, system and device

Abstract

The embodiment of the invention provides a reasoning method, system and device, which can be used for a cognitive reasoning service producer to generate a reasoning scheme with confidence and / or a risk value, so that the reasoning scheme is fed back to a cognitive reasoning service consumer based on the confidence and / or the risk value of the reasoning scheme. And a cognitive reasoning service consumer can obtain a required reasoning scheme. The method comprises the following steps that: a cognitive inference service producer receives an inference request, wherein the inference request is used for requesting to provide an inference scheme for completing a user intention target; according to the reasoning request, obtaining a first reasoning scheme set comprising at least one reasoning scheme and at least one item of confidence or risk value corresponding to each reasoning scheme, the confidence being used for indicating the probability of reaching a reasoning task when the reasoning scheme is used, the risk value is used for indicating the probability of changing the user experience when the reasoning task is achieved by using the reasoning scheme; and feeding back at least one scheme in the first reasoning scheme set.

Description

Technical Field

[0001] This application relates to the field of communications, and in particular, to an inference method, system, and device. Background Art

[0002] In a communication system, for example, in scenarios such as the optimization or maintenance of a communication network, a cognitive inference service consumer may request an inference solution from a task inference service producer, such as a solution for achieving the intention of optimizing or maintaining a communication network. After generating the inference solution, the task inference service producer distributes it to the cognitive inference service consumer. However, due to the complexity of the network system topology and the dependency relationship of fault propagation, there may be multiple root cause inference solutions. Therefore, how to enable the cognitive inference service consumer to obtain the required inference solution has become an urgent problem to be solved. Summary of the Invention

[0003] Embodiments of this application provide an inference method, system, and device, which can be used for a cognitive inference service producer to generate an inference solution with a confidence level and / or a risk value, and thus feedback the inference solution to the cognitive inference service consumer based on the confidence level and / or the risk value of the inference solution, so that the cognitive inference service consumer can obtain the required inference solution.

[0004] In view of this, in a first aspect, this application provides an inference method, including:

[0005] Receiving an inference request, where the inference request is used to request a solution for achieving an inference task, and the inference task includes a task for completing a user intention goal; then obtaining a first inference solution set according to the inference request, where the first inference solution set includes at least one inference solution and at least one of the confidence level or risk value corresponding to each inference solution, each inference solution is used to achieve the inference task, the confidence level is used to indicate the probability of achieving the inference task when using the inference solution, and the risk value is used to indicate the probability of changing the user experience when using the inference solution to achieve the inference task; sending a second inference solution set, where the second inference solution set includes at least one solution in the first inference solution set.

[0006] In an implementation manner of this application, while generating the inference solution, the confidence level and / or risk value of each inference solution are also generated. Therefore, when feeding back the inference solution, the confidence level and / or risk value of each inference solution can be fed back at the same time, or the inference solution is fed back after being screened based on the confidence level of each inference solution, so that the consumer can more accurately select the required solution based on the received inference solution.

[0007] For ease of understanding, hereinafter, the sender of the inference request will be referred to as the consumer, and the receiver of the inference request will be referred to as the producer, which will not be elaborated hereinafter.

[0008] In a possible implementation, the foregoing sending of the second set of inference schemes may include: using the first set of inference schemes as the second set of inference schemes; and sending the second set of inference schemes. Therefore, in the embodiments of the present application, the confidence level and / or risk value corresponding to each inference scheme can be fed back simultaneously, so that the data received by the consumer can accurately select the required inference scheme based on the confidence level and / or risk value of each inference scheme.

[0009] In a possible implementation, the foregoing inference request further carries a confidence level indication and / or a risk indication. The confidence level indication is used to request the provision of the confidence level corresponding to the inference scheme, and the risk indication is used to request the provision of the risk value corresponding to the inference scheme. Therefore, the consumer can carry the confidence level indication and / or the risk indication in the inference request, so that the producer can feed back the confidence level and / or risk value of each inference scheme based on the inference request, so that the data received by the consumer can accurately select the required inference scheme based on the confidence level and / or risk value of each inference scheme.

[0010] In a possible implementation, the foregoing sending of the second set of inference schemes may include: screening out the schemes with a confidence level higher than the confidence level threshold and / or the risk value higher than the risk threshold from the first set of inference schemes to obtain the second set of inference schemes; and sending the second set of inference schemes.

[0011] Therefore, in the embodiments of the present application, the producer can screen the generated inference schemes before feeding them back to the consumer, so that the schemes that meet the consumer's needs can be directly fed back, enabling the consumer to more accurately obtain the required inference scheme.

[0012] In a possible implementation, the inference request further carries a confidence level threshold and / or a risk threshold. Therefore, in the embodiments of the present application, the consumer can determine the confidence level threshold and / or the risk threshold, so that the producer can directly feed back the inference scheme that meets the consumer's needs.

[0013] In a possible implementation, the foregoing sending of the second set of inference schemes may further include: screening out one of the inference schemes from the first set of inference schemes as the scheme of the second set of inference schemes according to at least one of the confidence level or risk value corresponding to each inference scheme; and sending the second set of inference schemes. In the embodiments of the present application, the producer can directly feed back one inference scheme to the consumer. Therefore, the consumer can accurately obtain the required inference scheme without making a selection.

[0014] In a possible implementation, screening out one inference scheme from the foregoing first inference scheme set according to at least one of the confidence level or risk value corresponding to each inference scheme may include: screening out one inference scheme that meets the screening conditions from the first inference scheme set, where the screening conditions include at least one of the following: one inference scheme with the highest confidence level, one inference scheme with the lowest risk value, or one inference scheme with the highest combined value after combining the confidence level and the risk value. In the implementation of this application, one inference scheme with the highest confidence level, the lowest risk value, or obtained by combining the confidence level and the risk value can be selected, so as to adapt to various scenarios.

[0015] In a possible implementation, a policy indication is further carried in the inference request, and the policy indication is used to indicate at least one of the screening conditions. Therefore, in the implementation of this application, the consumer can select the screening conditions, so that the producer can directly screen out the inference scheme that meets the consumer's needs.

[0016] In a possible implementation, the second inference scheme set further includes at least one of the confidence level or risk value corresponding to the inference scheme. Therefore, the producer can also feedback the confidence level and / or risk value to the consumer, so that the consumer can know the confidence level and / or risk value corresponding to each inference scheme. When multiple inference schemes are feedback, the consumer can further select the required scheme based on the confidence level and / or risk value of each inference scheme.

[0017] In a possible implementation, the foregoing obtaining the first inference scheme set according to the inference request may include: obtaining monitoring data corresponding to the inference task and a knowledge base, where the monitoring data includes data of devices related to the inference task, and the knowledge in the knowledge base includes rules for completing the inference task; obtaining the first inference scheme set according to the monitoring data and the knowledge. Therefore, in the implementation of this application, inference can be performed based on the knowledge stored locally and the real-time monitoring data collected, so as to obtain an accurate inference scheme based on the preset rules.

[0018] In a possible implementation, the foregoing method may further include: obtaining an execution result, where the execution result is obtained after executing the inference scheme included in the second inference scheme set; updating the knowledge according to the execution result to obtain the updated knowledge. Therefore, in the implementation of this application, the knowledge can be updated based on the execution result of the inference scheme, so that the knowledge stored in the knowledge base can be more accurate and the accuracy of the subsequently generated inference scheme can be improved.

[0019] In a possible implementation manner, the foregoing inference tasks may specifically include at least one of a fault location task, a fault repair task, a configuration verification task, a potential problem identification task, or a solution recommendation task. The fault location task is used to locate the root cause of a fault in a device. The fault repair task is used to repair the device with a fault. The configuration verification task is used to verify the configuration of a device. The potential problem identification task is used to obtain the probability of a device having a fault. The solution recommendation task is used to obtain at least one solution for performing a target task. Therefore, the method provided in this application can be applied to various network optimization or maintenance scenarios and has very strong generalization ability.

[0020] In a second aspect, this application provides an inference method, including: First, send an inference request, where the inference request is used to request a solution for achieving an inference task, and the inference task includes a task for achieving a user intent target. Subsequently, receive a second inference solution set, where the second inference solution set includes at least one inference solution. The second inference solution set is obtained from a first inference solution set, and the first inference solution set is a set generated based on the inference request. The first inference solution set includes at least one inference solution and at least one of the confidence level or risk value corresponding to each inference solution. Each inference solution is used to achieve the inference task. The confidence level is used to indicate the probability of achieving the inference task when using the inference solution, and the risk value is used to indicate the probability of changing the user experience when using the inference solution to achieve the inference task.

[0021] In the implementation manner of this application, when generating an inference solution, the producer also generates the confidence level and / or risk value of each inference solution. Therefore, when feedbacking the inference solution, the confidence level and / or risk value of each inference solution can be feedbacked simultaneously, or the inference solutions can be filtered based on the confidence level of each inference solution and then feedbacked, so that the consumer can more accurately select the required solution based on the received inference solutions.

[0022] In addition, the effects achieved by the second aspect and any optional implementation manner of the second aspect can refer to the effects achieved by the foregoing first aspect or any optional implementation manner of the first aspect, and the following will not repeat the similar parts.

[0023] In a possible implementation manner, the foregoing second inference solution set further includes at least one of the confidence level or risk value corresponding to each inference solution in the at least one inference solution. Therefore, in the implementation manner of this application, the producer can feedback the confidence level and / or risk value corresponding to each inference solution simultaneously, so that the consumer can accurately select the required inference solution based on the confidence level and / or risk value of each inference solution from the received data.

[0024] In a possible implementation, the foregoing inference request further carries a confidence indication and / or a risk indication. The confidence indication is used to request the provision of a confidence level corresponding to the inference scheme, and the risk indication is used to request the provision of a risk value corresponding to the inference scheme.

[0025] In a possible implementation, the confidence level of the inference schemes included in the foregoing second inference scheme set is higher than a confidence threshold and / or the risk value is higher than a risk threshold.

[0026] In a possible implementation, the foregoing inference request further carries a confidence threshold and / or a risk threshold.

[0027] In a possible implementation, the foregoing second inference scheme set includes one inference scheme, and one inference scheme is selected from the first inference scheme set according to the confidence level and / or the risk value.

[0028] In a possible implementation, the foregoing inference request further carries a policy indication, and the policy indication is used to indicate the provision of one inference scheme.

[0029] In a possible implementation, the foregoing policy indication further indicates at least one of the screening conditions. The screening conditions include at least one of the following: one inference scheme with the highest confidence level, one inference scheme with the lowest risk value, or one inference scheme with the highest fusion value after fusing the confidence level and the risk value.

[0030] In a possible implementation, the foregoing method may further include:

[0031] Sending the inference schemes in the second inference scheme set;

[0032] Receiving an execution result, where the execution result is an execution result obtained by using the inference schemes in the second inference scheme set to complete the inference task;

[0033] Sending the execution result, where the execution result is used to update knowledge, and the knowledge is used to determine the second inference scheme set. The knowledge includes the rules for completing the inference task.

[0034] In a third aspect, the present application provides an inference method, including:

[0035] A first device sends an inference request to a second device. The inference request is used to request the provision of a scheme for achieving an inference task, and the inference task includes a task of completing a user intent goal;

[0036] The second device obtains a first set of inference solutions according to an inference request. The first set of inference solutions includes at least one inference solution, and at least one of the confidence level or risk value corresponding to each inference solution. Each inference solution is used to achieve an inference task. The confidence level is used to indicate the probability of achieving the inference task when using the inference solution, and the risk value is used to indicate the probability of changing the user experience when using the inference solution to achieve the inference task;

[0037] The second device sends a second set of inference solutions to the first device. The second set of inference solutions includes at least one solution from the first set of inference solutions.

[0038] In addition, the effects achieved by any optional implementation manner of the third aspect can refer to the effects achieved by the foregoing first aspect or any optional implementation manner of the first aspect. The following will not repeat the similar points.

[0039] In a possible implementation manner, the foregoing second device sending the second set of inference solutions to the first device includes:

[0040] The second device uses the first set of inference solutions as the second set of inference solutions;

[0041] The second device sends the second set of inference solutions to the first device.

[0042] In a possible implementation manner, the inference request further carries a confidence level indication and / or a risk indication. The confidence level indication is used to request the provision of the confidence level corresponding to the inference solution, and the risk indication is used to request the provision of the risk value corresponding to the inference solution.

[0043] In a possible implementation manner, the foregoing second device sending the second set of inference solutions to the first device may include:

[0044] The second device filters out the solutions with a confidence level higher than the confidence level threshold and / or the solutions with a risk value higher than the risk threshold from the first set of inference solutions to obtain the second set of inference solutions;

[0045] The second device sends the second set of inference solutions to the first device.

[0046] In a possible implementation manner, the foregoing inference request further carries a confidence level threshold and / or a risk threshold.

[0047] In a possible implementation manner, the foregoing second device sending the second set of inference solutions to the first device may further include:

[0048] The second device filters out one inference solution from the first set of inference solutions as the solution of the second set of inference solutions according to at least one of the confidence level or risk value corresponding to each inference solution;

[0049] The second device sends a second set of inference solutions to the first device.

[0050] In a possible implementation, the above-mentioned second device selects one inference solution from the first set of inference solutions as the solution of the second set of inference solutions according to at least one of the confidence levels or risk values corresponding to each inference solution, which may include:

[0051] The second device selects one inference solution that meets the screening conditions from the first set of inference solutions. The screening conditions include at least one of the following: one inference solution with the highest confidence level, one inference solution with the lowest risk value, or one inference solution with the highest fusion value after fusing the confidence level and the risk value.

[0052] In a possible implementation, a policy indication is further carried in the above-mentioned inference request, and the policy indication is used to indicate at least one of the screening conditions.

[0053] In a possible implementation, at least one of the confidence level or risk value corresponding to the inference solution is further included in the above-mentioned second set of inference solutions.

[0054] In a possible implementation, the above-mentioned second device obtains the first set of inference solutions according to the inference request, including:

[0055] The second device obtains the monitoring data and the knowledge base corresponding to the inference task. The monitoring data includes the data of the devices related to the inference task, and the knowledge in the knowledge base includes the rules for completing the inference task;

[0056] The second device obtains the first set of inference solutions according to the monitoring data and the knowledge.

[0057] In a possible implementation, the above-mentioned method may further include:

[0058] The first device sends an execution result to the second device, and the execution result is obtained after executing the inference solution included in the second set of inference solutions;

[0059] The second device updates the knowledge according to the execution result to obtain the updated knowledge.

[0060] In a possible implementation, the above-mentioned inference task includes at least one of a fault location task, a fault repair task, a configuration verification task, a hidden danger identification task, or a solution recommendation task. The fault location task is used to locate the root cause of the device fault, the fault repair task is used to repair the device with the fault, the configuration verification task is used to verify the configuration of the device, the hidden danger identification task is used to obtain the probability of the device having a fault, and the solution recommendation task is used to obtain at least one solution for executing the target task.

[0061] In a fourth aspect, the present application provides an inference device. The inference device can be used to execute the steps performed by the inference device in the above-mentioned first aspect and any possible implementation manner provided by the first aspect.

[0062] In a possible design, the present application can divide the inference device into functional modules according to the above-mentioned first aspect and any possible implementation manner provided by the first aspect. For example, each functional module can be divided corresponding to each function, or two or more functions can be integrated into one processing module.

[0063] Exemplarily, the present application can divide the inference device into a transceiver module, a processing module, etc. according to functions. The descriptions of the possible technical solutions and beneficial effects executed by each of the above-mentioned divided functional modules can refer to the technical solutions provided in the above-mentioned first aspect or its corresponding possible implementation manner, and will not be elaborated here.

[0064] In another possible design, the inference device includes: a memory and a processor, and the memory and the processor are coupled. The memory is used to store computer instructions, and the processor is used to call the computer instructions to execute the method provided in the first aspect or its corresponding possible implementation manner.

[0065] In a fifth aspect, the present application provides an inference device. The inference device can be used to execute the steps performed by the inference device in the above-mentioned second aspect and any possible implementation manner provided by the second aspect.

[0066] In a possible design, the present application can divide the inference device into functional modules according to the above-mentioned second aspect and any possible implementation manner provided by the second aspect. For example, each functional module can be divided corresponding to each function, or two or more functions can be integrated into one processing module.

[0067] Exemplarily, the present application can divide the inference device into a transceiver module, a processing module, etc. according to functions. The descriptions of the possible technical solutions and beneficial effects executed by each of the above-mentioned divided functional modules can refer to the technical solutions provided in the above-mentioned second aspect or its corresponding possible implementation manner, and will not be elaborated here.

[0068] In another possible design, the inference device includes: a memory and a processor, and the memory and the processor are coupled. The memory is used to store computer instructions, and the processor is used to call the computer instructions to execute the method provided in the second aspect or its corresponding possible implementation manner.

[0069] In a sixth aspect, the present application provides a communication system, which may also be referred to as a communication network, etc. The communication system may include multiple inference devices, where at least one inference device may be used to execute the method steps in the foregoing second aspect or any optional implementation manner of the second aspect, and at least one of the remaining inference devices may also be used to execute the method steps in the foregoing first aspect or any optional implementation manner of the first aspect.

[0070] In a seventh aspect, the present application provides a computer-readable storage medium, such as a non-transitory computer-readable storage medium. A computer program (or instruction) is stored thereon. When the computer program (or instruction) runs on a computer device, the computer device is caused to execute the method provided in the first aspect or its corresponding possible implementation manner, or execute the method provided in the second aspect or its corresponding possible implementation manner.

[0071] In an eighth aspect, the present application provides a computer program product. When it runs on a computer device, the method provided in the first aspect or its corresponding possible implementation manner is caused to be executed, or the method provided in the second aspect or its corresponding possible implementation manner is caused to be executed.

[0072] In a ninth aspect, the present application provides a chip system, including: a processor, where the processor is used to call and run a computer program stored in a memory and execute the method provided in the first aspect or its corresponding possible implementation manner, or execute the method provided in the second aspect or its corresponding possible implementation manner.

[0073] It can be understood that any of the above-provided systems, devices, computer storage media, computer program products, or chip systems, etc. can be applied to the corresponding methods provided in the first aspect or the second aspect.

[0074] For the specific implementation steps of the fourth aspect to the ninth aspect of the present application and various possible implementation manners, as well as the beneficial effects brought by each possible implementation manner, reference can be made to the descriptions in various possible implementation manners in the first aspect or the second aspect, and details are not repeated here. Description of the Drawings

[0075] Figure 1 It is a schematic diagram of a system architecture provided by the present application;

[0076] Figure 2 It is another schematic diagram of a system architecture provided by the present application;

[0077] Figure 3 It is another schematic diagram of a system architecture provided by the present application;

[0078] Figure 4 It is another schematic diagram of a system architecture provided by the present application;

[0079] Figure 5 Flow schematic diagram of an inference method provided for this application;

[0080] Figure 6 Flow schematic diagram of another inference method provided for this application;

[0081] Figure 7 Structural schematic diagram of a data and knowledge base provided for this application;

[0082] Figure 8 Flow schematic diagram of another inference method provided for this application;

[0083] Figure 9 Flow schematic diagram of another inference method provided for this application;

[0084] Figure 10 Flow schematic diagram of another inference method provided for this application;

[0085] Figure 11 Flow schematic diagram of another inference method provided for this application;

[0086] Figure 12 Flow schematic diagram of another inference method provided for this application;

[0087] Figure 13 Flow schematic diagram of another inference method provided for this application;

[0088] Figure 14 Structural schematic diagram of an inference device provided for this application;

[0089] Figure 15 Structural schematic diagram of another inference device provided for this application;

[0090] Figure 16 Structural schematic diagram of another inference device provided for this application. Specific embodiments

[0091] Next, the technical solutions in the embodiments of this application will be described in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in this application without creative efforts shall fall within the protection scope of this application.

[0092] Refer to Figure 1, in the network operation and maintenance architecture based on 5G communication technology, it usually includes three layers, namely: network element (NE), element management system (EMS), and network management system (NMS).

[0093] Among them, NE are various network devices that make up the network, which can include but are not limited to base stations, access points, various core network devices, etc. Optionally, NE can be physical entity devices or virtual network functions, and this application does not limit this. EMS can not only manage NE, but also be a provider of network operation and maintenance management services, and can execute specific network operation and maintenance management service operations according to the service call requests sent by NMS. As an invoker of network operation and maintenance management services, NMS can send service call requests to EMS according to the network operation and maintenance management requirements of customers, so as to call the network operation and maintenance management services in EMS that meet the operation and maintenance management requirements of customers. Both EMS and NMS are composed of at least one device. The devices in EMS can be called network element management devices, and the devices in NMS can be called network management devices.

[0094] Network operation and maintenance includes network optimization. Network optimization usually can include improving user experience, reducing network energy consumption or increasing network capacity. The embodiments of this application do not specifically limit this, and the requirements and goals of network optimization can be set according to actual needs. Traditional optimization solutions based on expert experience are difficult to achieve satisfactory results in increasingly complex communication scenarios. In order to bring better network experience to customers and improve operation and maintenance efficiency, introducing artificial intelligence for network operation and maintenance has become a mainstream trend. Machine learning is a branch of artificial intelligence. Machine learning is to use algorithms to analyze data, learn from it, and then make decisions and predictions on events in the real world. In the network optimization scenario, the decisions and predictions made by the machine learning model are optimization suggestions, that is, changes in the values of the parameters of the devices in the network (sometimes also referred to as parameter values or simply parameters in the embodiments of this application). Taking a wireless base station as an example, these parameters include scheduling parameters, resource management parameters, radio frequency (RF) parameters, power parameters, handover parameters, and so on.

[0095] For example, in some solutions, cognitive reasoning service consumers create fault location cognitive reasoning service tasks, including the location of the fault, fault-related alarms, fault events and other information; after receiving the task, the producer generates a fault chain by reasoning and infers the root cause of the fault; then, the producer returns the reasoning results to the consumer. Due to the topological complexity of the network system and the dependency of fault propagation, there are likely to be multiple root cause reasoning results. However, faced with multiple results, users cannot know which root cause is the real root cause, which affects the overall root cause location and fault repair time, affecting the user experience.

[0096] For example, in some scenarios, cognitive reasoning service consumers create energy-saving solution recommendation reasoning service requests, including energy-saving action areas, energy-saving goals and experience constraints, and action time periods. After receiving the request, producers perform reasoning and return reasoning solutions. However, cognitive reasoning service consumers usually receive multiple sets of reasoning solutions from producers, and users may not be able to choose the most appropriate solution according to their needs. Random selection leads to low efficiency in closed-loop problems.

[0097] Therefore, the present application provides a reasoning method, which generates a confidence level and / or a risk value for each reasoning scheme when generating a reasoning scheme, so that the user can select an appropriate scheme based on the needs.

[0098] The system architecture and method provided by this application are introduced in more detail below.

[0099] The reasoning method provided in this application can be applied to various communication networks, such as various optical networks or wireless networks. Figure 2 , is a schematic diagram of another architecture of a communication system provided in an embodiment of the present application. The communication network includes an NMS, an EMS, and network elements. Figure 2 NMS is used to indicate the equipment included in NMS, and EMS is used to indicate the equipment included in EMS. The following will not be repeated. For the relevant description of NMS, EMS and network elements, please refer to Figure 1Understanding from the relevant descriptions therein, NMS is responsible for the management of the network operation, administration, and maintenance functions. The EMS, i.e., the network element management system, is responsible for managing one or more categories of network elements. A network element, i.e., a network unit. In a radio access network, it refers to a base station or a base station controller, including but not limited to: one or more base stations in a global system for mobile communications (GSM), a universal mobile telecommunications system (UMTS), long term evolution (LTE), and new radio (NR), as well as base station controllers of GSM and UMTS. In a bearer network, it refers to one or more of an access router (Access, ACC), a metro edge gateway (mEG), a metro aggregation edge gateway (mAEG), a metro backbone (mBB), and a 5G backbone (5G BB). In a core network, it refers to one or more of a mobile switching center (MSC), a serving general packet radio system support node (SGSN), an equipment identity register (EIR), a packet data network gateway for user plane (PGW-U), an access and mobility management function (AMF), and a unified data management (UDM).

[0100] Based on the communication network described above, the solution provided by the embodiments of the present application can be jointly executed by the NMS and the EMS, or can be jointly executed by the EMS and the network element, or can also be deployed in other devices.

[0101] For example, as Figure 3 shown, it is another schematic architecture diagram provided by the present application. Among them, the EMS, the NMS, and the network element can refer to the relevant descriptions above Figure 1 and will not be elaborated here.

[0102] The NMS can send an inference request to the EMS, and this inference request can be used to request from the EMS an inference solution for performing maintenance or optimization tasks of the communication network.

[0103] After receiving the inference request, the EMS sends alternative inference solutions to the NMS. Among them, the EMS can pre-collect data from network elements, and the network elements can upload their own data or the data of the devices they manage to the EMS, and then the EMS can generate an inference solution based on the collected data.

[0104] After receiving the alternative inference solutions, the NMS determines the final inference solution and sends it to the network elements through the EMS.

[0105] After receiving the final inference solution, the network elements can execute the final inference solution and feedback the execution results to the NMS.

[0106] Refer to Figure 4 , which is another architecture schematic diagram provided by this application.

[0107] Furthermore, the NMS can be divided into an intent service producer (such as a user's inference device or user interface, etc.) and a task inference service consumer (such as an NMS entity server or controller, etc.), and the EMS can include a cognitive inference service producer (such as an EMS server or server cluster) and a knowledge base (such as a database or file system, etc.).

[0108] Cognitive inference service consumer: It can be used to call the cognitive inference service provided by the cognitive inference service producer and send a service request to the cognitive inference service producer.

[0109] Cognitive inference service producer: It is used to provide cognitive inference services and return an inference solution to the cognitive inference service consumer. The cognitive inference service producer can be deployed within the EMS, and the producer also maintains the knowledge base. The producer has the ability of knowledge construction and knowledge update in combination with this knowledge base.

[0110] In the implementation manner of this application, the intent service producer can be used to generate intent service requirements, such as maintenance intent or optimization intent for the communication network, etc., and send the intent to the cognitive inference service consumer, and the cognitive inference service consumer sends the intent service requirement to the cognitive inference service producer.

[0111] The cognitive inference service producer maintains the knowledge base. Specifically, it can use the data collected from network elements to update the knowledge in the knowledge base and use the knowledge in the knowledge base to complete the inference task, that is, generate an inference solution for the intent service requirement, and feedback the generated inference solution as an alternative inference solution to the cognitive inference service consumer.

[0112] After the cognitive service inference consumer sends the final inference solution to the network element, it receives the execution result of the network element and feeds back the intent execution result to the intent service producer.

[0113] When performing network operation and maintenance, the network operation and maintenance solutions used can be inferred and generated based on data-driven, or generated based on knowledge-driven, or can also be a solution that combines data-driven and knowledge-driven. As shown in Table 1, for data-driven, its advantage is that it does not require precise modeling and can continuously learn and evolve, but its interpretability is poor, it needs to rely on high-quality large-scale data, and it also needs to rely on the powerful computing power of the device. For knowledge-driven, its advantages are strong interpretability and high execution efficiency, but usually the cost of knowledge acquisition is high, it is difficult to continuously learn and evolve, and the knowledge mechanism is not clear for large-scale complex problems.

[0114] Therefore, based on the above analysis, this application can combine knowledge-driven and data-driven, utilize their respective advantages, and form a new technical route of knowledge and data fusion (i.e., cognitive inference). Combining with the business scenario of the autonomous driving network, knowledge mainly serves as the carrier of expert experience and is used to express some expert experience in operation and maintenance (such as fault propagation knowledge, configuration constraint knowledge), and data mainly reflects the status of the existing network. Typical applications of knowledge + data fusion cognitive inference in ADN include fault location (such as hardware faults in communication systems), fault repair (such as switching, restarting, rolling back, parameter adjustment), configuration verification, hidden danger identification (such as which devices may have faults), solution recommendation (such as energy-saving solution recommendation), and so on.

[0115]

[0116] Table 1

[0117] The machine learning model mentioned in this application may include artificial intelligence models such as neural networks. The performance of most neural networks depends on a large amount of training data. Taking the network optimization scenario as an example to illustrate this, the real network scenario is complex and dynamically changing, and there are many factors affecting network performance. In addition to the parameters of network devices, the surrounding geographical environment (such as the terrain, surrounding buildings, etc.), the distribution and behavior of users, and even the climate will affect network performance. This leads to a huge state space and decision space faced by the machine learning model, requiring a large amount of data as training data for learning. Or it can be understood that due to the numerous factors affecting network performance, during the iterative training process of the machine learning model, there is also a large amount of knowledge to be learned, that is, the machine learning model needs to obtain the relationship between a large number of parameter changes and performance changes (or network state changes) as training data to learn the relationship between network configuration parameters and network performance changes from it. It should be noted that the model in this application includes a neural network model, and this application does not limit the specific category of the model. For example, the model in this application can be a convolutional neural network model, a recurrent neural network model, a deep residual network model, etc., and will not be repeated hereinafter.

[0118] Next, in combination with the foregoing system architecture, the method flow provided by this application will be introduced.

[0119] Refer to Figure 5 , the schematic flowchart of an inference method provided by this application is as follows.

[0120] 501. The NMS sends an inference request to the EMS.

[0121] Among them, the inference request is used to request an inference solution for achieving an inference task (or it can also be called an inference intention or a cognitive inference service request, etc.). The inference task may include a task for completing a user intention goal, and the inference task may specifically include a task or intention for managing a communication network, such as a network optimization intention or a network maintenance intention, etc.

[0122] The inference request may specifically come from the NMS or other management devices for managing the communication network. For example, when the management device obtains the intention of the user to manage the communication network, it can generate an inference request and send it to the EMS.

[0123] Optionally, the inference task may include, but is not limited to, one or more of the following: at least one of a fault location task, a fault repair task, a configuration verification task, a potential problem identification task, or a solution recommendation task. The fault location task is used to locate the root cause of a fault in a device. The fault repair task is used to repair the device that has a fault. The configuration verification task is used to verify the configuration of a device. The potential problem identification task is used to obtain the probability of a device having a fault. The solution recommendation task is used to obtain at least one solution for performing a target task.

[0124] 502. The EMS obtains a first set of inference solutions based on an inference request.

[0125] After receiving the inference request, the EMS can generate one or more inference solutions, which are referred to as the first set of inference solutions for easy distinction.

[0126] Among them, when generating the inference solutions, the EMS can synchronously generate at least one of a confidence level or a risk value. The confidence level can be used to indicate the probability of achieving the inference task when using the inference solution. The risk value is used to indicate the probability of changing the user experience when using the inference solution to achieve the inference task. The user experience can be specifically reflected by indicators such as a user service level agreement (SLA) and a user quality of service (QoS).

[0127] Optionally, when generating one or more inference solutions, data-driven and / or knowledge-driven methods can be combined to generate the inference solutions. For example, the monitoring data corresponding to the inference task and the knowledge stored in the knowledge base can be obtained. The monitoring data may include data of devices related to the inference task. The knowledge may include rules for completing the inference task. Subsequently, the first set of inference solutions is obtained based on the monitoring data and the knowledge. Therefore, in the implementation manner of this application, the inference solutions can be generated by combining the actual data generated by the devices in the communication network and the pre-known knowledge, so that more accurate inference solutions can be obtained.

[0128] 503. The EMS sends a second set of inference solutions to the NMS.

[0129] After generating the first set of inference solutions, the EMS can feedback one or more inference solutions to the NMS as alternative inference solutions for the inference task. That is, the second set of inference solutions may include one or more inference solutions in the first set of inference solutions.

[0130] In the embodiments of the present application, when the EMS generates an inference solution, it can simultaneously generate the confidence level and / or risk value corresponding to each inference solution. Therefore, when feeding back alternative inference solutions to the NMS, the confidence level and / or risk value of each inference solution can be fed back simultaneously, or a solution that meets the user's needs, such as a solution with a higher confidence level and / or a lower risk value, can be selected, enabling the user to more accurately select the required inference solution.

[0131] Therefore, when the EMS feeds back the second set of inference solutions to the NMS, it can be divided into multiple cases. For example, it can be divided into but not limited to the following multiple ways:

[0132] 1. Feed back all inference solutions and the confidence level and / or risk value of each solution

[0133] After the EMS obtains the first set of inference solutions, it can directly use this first set of inference solutions as the second set of inference solutions and feed back this second set of inference solutions to the NMS. That is, it can directly feed back the inference solutions in the first set of inference solutions and the confidence level and / or risk value of each solution to the NMS, and the user can select the required inference solution based on the confidence level and / or risk value of each solution.

[0134] Correspondingly, the inference request can also include a confidence level indication and / or a risk indication. The confidence level indication can be used to request the provision of the confidence level corresponding to the inference solution, and the risk indication is used to request the provision of the risk value corresponding to the inference solution. Of course, when the EMS feeds back the recorded inference solutions, it can also default to feeding back the confidence level and / or risk value corresponding to each inference solution to the NMS.

[0135] Therefore, when the user selects the final inference solution, they can more accurately select the required inference solution based on the confidence level and / or risk value of each inference solution, thereby improving the compliance of the selected solution with the user's needs. Taking energy conservation as an example, in a conservative scenario, the user tends to select a solution with a low risk (even if the confidence level is not the highest), and in an aggressive scenario, the user tends to select a solution with a high confidence level (even if the risk is also high). For example, the remote radio unit (RRU) deep sleep has good energy-saving effects and a high confidence level, but also a high risk.

[0136] 2. Feed back one or more inference solutions that meet the condition that the confidence level is higher than the confidence level threshold and / or the risk value is lower than the risk value threshold

[0137] After the EMS obtains the first set of inference solutions, it can screen out one or more inference solutions with a confidence level higher than the confidence level threshold and / or a risk value lower than the risk value threshold from it and feed them back to the NMS. Therefore, it can directly feed back the inference solutions that meet its needs to the NMS, enabling it to directly select the inference solutions that meet the needs.

[0138] Optionally, the confidence level and / or risk value of each inference scheme in the one or more inference schemes can also be fed back synchronously. So that NMS can know the specific confidence level and / or risk value of each inference scheme, and thus can know in advance the expected result when executing the inference scheme. And when EMS feeds back multiple inference schemes, the required inference scheme can be selected more accurately based on the confidence level and / or risk value of each inference scheme.

[0139] Optionally, NMS can also carry a confidence level threshold and / or a risk value threshold in the inference request to instruct EMS to provide inference schemes with a confidence level higher than the confidence level threshold and / or a risk value lower than the risk value threshold.

[0140] Of course, in some scenarios, the default confidence level threshold and / or risk value threshold can also be set by the cognitive inference method service producer, and there is no need to carry the confidence level threshold and / or risk value threshold in the inference request, so as to reduce the data transmission volume and thus reduce the communication overhead.

[0141] In the embodiment of the present application, EMS can directly feed back to NMS the inference schemes with a confidence level higher than the confidence level threshold and / or a risk value lower than the risk value threshold. This is equivalent to a preliminary screening, providing inference schemes that meet the user's needs. Therefore, the inference schemes fed back to the user are all schemes that meet the user's needs, enabling the user to obtain inference schemes with higher usability.

[0142] 3. Feed back an inference scheme that meets the user's specified requirements

[0143] In the embodiment of the present application, in order to further reduce the selection decisions that the user needs to make, an inference scheme that meets the user's specified requirements can be directly fed back based on the user's needs, so that the user does not need to make a re-selection and the communication overhead can be reduced.

[0144] After EMS obtains the first inference scheme set, one inference scheme that meets the screening conditions can be selected from it and fed back to NMS. Specifically, one inference scheme can be selected based on the confidence level and / or risk value of each inference scheme as the scheme in the aforementioned second inference scheme set and fed back to NMS. Therefore, the user does not need to select the received inference scheme, and the received one inference scheme can be used as the final inference scheme.

[0145] The screening conditions can specifically include but are not limited to at least one of the following: one inference scheme with the highest confidence level, one inference scheme with the lowest risk value, or one inference scheme with the highest fusion value after fusing the confidence level and the risk value.

[0146] For example, EMS can feedback a reasoning solution with higher confidence, or feedback a reasoning solution with higher confidence, or feedback a solution that meets the requirements by combining confidence and risk value, such as the highest ranked solution among the top N solutions in terms of confidence and risk value, or the highest value obtained after fusion calculation of confidence and risk value, etc.

[0147] Optionally, the NMS may carry a policy indication in the inference request, and the policy indication may be used to indicate a screening condition, so that the EMS may screen out one of the generated one or more inference schemes based on the policy indication and feed it back to the NMS.

[0148] Of course, the EMS can select one of the generated one or more inference solutions based on pre-set screening conditions and feed it back to the NMS. The specific conditions can be determined according to the actual application scenario, and this application does not limit this.

[0149] Therefore, in the implementation manner of the present application, an inference scheme can be directly screened out and fed back to the NMS, so that the user can directly obtain the inference scheme that meets the needs, and the user can directly determine the final inference scheme without re-screening.

[0150] The above is a general introduction to the method flow provided by the present application. The following is a more detailed introduction to the method flow provided by the present application in combination with the above method flow and application scenarios.

[0151] In addition, after receiving one or more candidate reasoning solutions, the NMS can continue to send the final reasoning solution to the network element. After obtaining the execution result of the final reasoning solution, the data and knowledge base are updated so that the actual feedback of the execution result of the reasoning solution can be combined to improve the accuracy of the next reasoning.

[0152] See also Figure 6 , a flowchart of another reasoning method provided by this application is described as follows.

[0153] 601. EMS collects data from network elements and updates data and knowledge base.

[0154] Among them, EMS can collect information related to various scenarios from network elements, such as data generated or collected by various devices. After the data is collected, the data and knowledge base can be updated.

[0155] In the implementation mode of the present application, when the EMS performs reasoning to generate a reasoning scheme, the reasoning scheme can be generated by combining data-driven and knowledge-driven. Accordingly, in the data collection stage, the collected data can be used to update the data and knowledge base, so as to use the data and knowledge base to perform subsequent reasoning steps.

[0156] Specifically, when collecting data, data generated by the operation of each device itself can be collected, information accessed by each device can be collected, and environmental information where each device is located can also be collected, etc. For example, fault alarm information of the device, information on the energy consumption of the device, configuration information of the device, or access information of the device can be collected.

[0157] For example, as Figure 7 shown, a knowledge base can be established, which includes a variety of pre-set knowledge. The knowledge usually can include expert experience or knowledge learned from the collected data. The knowledge can be used to indicate the rules of various types of functions or tasks, or the logical chain between the types of data on which each function depends. In the embodiment of the present application, the knowledge base can include knowledge of the rules for completing the inference task.

[0158] When generating an inference plan, data-driven can also be combined. Therefore, machine learning can be pre-performed based on the collected data to obtain a machine learning model, such as Figure 7 the neural network shown in. When generating an inference plan, the data of the device collected can be used as the input of the neural network, and the inference plan can be output. And for different inference tasks, different neural networks can be set, or a cross-modal large model for implementing multiple functions can also be set, etc.

[0159] It should be noted that step 601 can be executed when each device is initialized, or can be actively reported when the data of the network element changes, or can be periodically collected by the EMS. Specifically, it can be determined according to the actual application scenario. The present application does not limit the execution timing of step 601.

[0160] 602. The NMS sends an inference request to the EMS.

[0161] This inference request can be used to request the EMS to provide an inference plan for achieving the inference task (or it can also be called an inference intention or a cognitive inference service request, etc.). The inference task can include tasks or intentions for managing the communication network, such as tasks or intentions for maintaining or optimizing the communication network, etc.

[0162] Among them, the information of the inference task can be carried in the inference request so that the EMS can identify the inference task service to be executed. For example, the task name, task field, or task identifier of the inference task can be carried, and task description information can also be carried. For example, for a fault location task, the task description information that can be carried can include the time period when the fault occurred, the location where the fault occurred, the content of the fault event, etc.

[0163] In addition, NMS can actively send an inference request to the EMS, or send an inference request to the EMS when triggered by other devices or users, which can be specifically limited according to the actual application scenario and is not limited in this application.

[0164] For the three different scenarios described above, different contents may be carried in the inference request. For specific details, please refer to the following Figures 8 to 10 corresponding introduction and will not be elaborated here.

[0165] 603. The EMS obtains one or more inference solutions and the confidence level and / or risk value of each solution according to the inference request.

[0166] After receiving the inference request, the EMS can obtain one or more alternative inference solutions and the confidence level and / or risk value of each solution based on the inference request in combination with data-driven and / or knowledge-driven, that is, the first inference solution set described above.

[0167] Specifically, when generating an inference solution, one or more inference solutions can be generated based on data-driven. For example, the data of the devices in the communication network that need to be maintained or optimized can be used as the input of the neural network, so as to output the inference solution and the confidence level and / or risk value corresponding to each inference solution. Or, one or more inference solutions can also be generated based on knowledge-driven. For example, the data of the devices in the communication network that need to be maintained or optimized can be read, and the maintenance or optimization solutions corresponding to each device can be determined based on knowledge, so as to form the maintenance or optimization solution of the communication network and the confidence level and / or risk value corresponding to each inference solution. One or more inference solutions can also be generated by combining data-driven and knowledge-driven; for example, the collected device data can be used as the input of the neural network, and knowledge can be used as a constraint to output one or more inference solutions and the confidence level and / or risk value corresponding to each inference solution, etc. The specific calculation method can be adjusted according to the actual application scenario and is not limited in this application.

[0168] Generally, the value of the confidence level is related to the probability of achieving the inference task. In this application, by way of example, it is introduced by taking the positive correlation between the value of the confidence level and the probability of achieving the inference task as an example. That is, the higher the confidence level, the higher the probability of achieving the inference task when using the inference solution; the lower the confidence level, the lower the probability of achieving the inference task when using the inference solution. Of course, in some scenarios, it can also be replaced by a negative correlation between the value of the confidence level and the probability of achieving the inference task, which is not limited in this application.

[0169] Accordingly, the risk value is related to the probability of changing the user service level agreement (SLA) when performing the reasoning task using the reasoning scheme. Exemplarily in this application, an example is given where the risk value is positively correlated with the probability of changing the protocol SLA when performing the reasoning task using the reasoning scheme. That is, the lower the risk value, the lower the probability of changing the SLA when performing the reasoning task using the reasoning scheme, and the higher the risk value, the higher the probability of changing the SLA when performing the reasoning task using the reasoning scheme. Of course, it can also be replaced with a negative correlation between the risk value and the probability of changing the protocol SLA when performing the reasoning task using the reasoning scheme, which can be specifically replaced according to the actual application scenario, and this application does not make any limitations in this regard.

[0170] 604. The EMS sends one or more alternative reasoning schemes to the NMS.

[0171] After the EMS generates one or more reasoning schemes and the confidence level and / or risk value of each scheme, it can, based on the one or more reasoning schemes and the confidence level and / or risk value of each scheme, send one or more alternative reasoning schemes to the NMS, that is, the aforementioned second set of reasoning schemes.

[0172] Optionally, when sending the alternative reasoning schemes, the confidence level and / or risk value of each alternative reasoning scheme can be sent. For example, when sending multiple alternative reasoning schemes, the confidence level and / or risk value of each reasoning scheme can be sent synchronously; or the generated reasoning schemes can be screened according to the confidence level and / or risk value of each reasoning scheme, and one or more schemes that meet the user's needs are selected for feedback. At this time, the confidence level and / or risk value of each reasoning scheme can be sent synchronously or not synchronously; or, if only one reasoning scheme is feedback, at this time, the confidence level and / or risk value of each reasoning scheme can also be selected to be sent synchronously or not sent.

[0173] Accordingly, for the aforementioned different feedback methods, different contents may be carried in the reasoning request, and specific details can be referred to the following Figures 8 to 10 corresponding introduction, which will not be elaborated here.

[0174] 605. The NMS sends the final reasoning scheme to the network element.

[0175] If the NMS receives one alternative reasoning scheme sent by the EMS, it can directly use this one alternative reasoning scheme as the final reasoning scheme and send it to the network element; if the NMS receives multiple alternative reasoning schemes sent by the EMS, it can screen out one of the multiple alternative reasoning schemes as the final reasoning scheme.

[0176] Optionally, if the NMS receives multiple alternative inference solutions sent by the EMS, the NMS can randomly select one inference solution; if the confidence level and / or risk value corresponding to each inference solution are received simultaneously, one inference solution can be selected from multiple inference solutions based on the confidence level and / or risk value of each inference solution. For example, the inference solution with the highest confidence level or the lowest risk value can be selected, etc.

[0177] In the embodiments of the present application, the EMS can feedback alternative inference solutions to the NMS based on the confidence level and / or risk value of the inference solutions. Therefore, the NMS can more accurately determine a better inference solution.

[0178] In addition, when the NMS sends the final inference solution to the network element, usually the EMS can forward the inference solution, and issue the maintenance or optimization solutions for each device in the communication network included in the inference solution to each network element device. For example, each network element is instructed to execute the specified configuration, thereby improving the performance of the communication network. The following embodiments of the present application will not elaborate on the sending method of this final inference solution.

[0179] 606. The network element feeds back the execution result to the NMS.

[0180] After each network element executes the inference solution, it can feed back the execution result to the NMS, such as the operation status of the network element after the fault is repaired, or the energy consumption of the network element after the energy-saving solution is adopted, etc.

[0181] 607. The NMS sends the execution result to the EMS.

[0182] After receiving the execution result, the NMS can send the execution result to the EMS.

[0183] 608. The EMS updates the data or knowledge base based on the execution result.

[0184] After receiving the execution result, the EMS can update the data or knowledge base based on the execution result, so that the EMS can obtain a more accurate inference solution when generating the inference solution for the next time.

[0185] In the embodiments of the present application, when generating the inference solution, the EMS can synchronously generate the confidence level and / or risk value of each inference solution, that is, it can feedback alternative inference solutions based on the confidence level and / or risk value of each inference solution. For example, after screening the generated inference solutions based on the confidence level and / or risk value of each inference solution and then issuing them, or synchronously issuing the confidence level and / or risk value of each inference solution, etc., so that the NMS can more accurately obtain the required inference solution.

[0186] For ease of understanding, the following will introduce different inference solution feedback strategies respectively.

[0187] Embodiment 1: Feedback all inference schemes and the confidence level and / or risk value of each scheme

[0188] Refer to Figure 8 , a flowchart of another inference method provided by this application is described as follows.

[0189] 801. The EMS collects data from network elements and updates the data and the knowledge base.

[0190] Among them, step 801 can refer to the relevant description of the foregoing step 601, and will not be elaborated here.

[0191] 802. The NMS sends an inference request to the EMS, and the inference request may carry a confidence indication and / or a risk indication.

[0192] When the NMS needs the EMS to provide an inference scheme for performing an inference task, the NMS can send an inference request to the EMS. This inference request can refer to the relevant description of the foregoing step 602, and the similarities will not be elaborated.

[0193] Among them, optionally, the inference request may carry a confidence indication (confidence reporting indication) and / or a risk indication (risk reporting indication). The confidence indication can be used to indicate that the EMS synchronously feedbacks the confidence level of each inference scheme when providing the inference scheme, and the risk indication can be used to indicate that the EMS synchronously feedbacks the risk value of each inference scheme when providing the inference scheme.

[0194] 803. The EMS obtains one or more inference schemes and the confidence level and / or risk value of each scheme.

[0195] Among them, step 803 can refer to the relevant description of the foregoing step 603, and will not be elaborated here.

[0196] 804. The EMS sends one or more alternative inference schemes and the confidence level and / or risk value of each scheme to the NMS.

[0197] After the EMS generates one or more inference schemes and the confidence level and / or risk value of each scheme, the one or more inference schemes and the confidence level and / or risk value of each scheme can be sent to the NMS.

[0198] 805. The NMS sends the final inference scheme to the network element.

[0199] If the NMS only receives one alternative reasoning scheme, it can determine whether to use the alternative reasoning scheme as the final reasoning scheme based on the confidence and / or risk value of the alternative reasoning scheme. For example, if the confidence of the alternative reasoning scheme exceeds the first threshold or the risk value is lower than the second threshold, the alternative reasoning scheme can be used as the final reasoning scheme. Of course, the alternative reasoning scheme can also be directly used as the final reasoning scheme, which is adjusted according to the actual application scenario. In addition, the final reasoning scheme can also be called the actual reasoning scheme, the target reasoning scheme or other replaceable names, etc., and this application does not limit this.

[0200] If the NMS receives multiple alternative reasoning schemes, one of the reasoning schemes can be selected as the final reasoning scheme based on the confidence and / or risk value of the multiple alternative reasoning schemes. For example, a reasoning scheme with the highest confidence is selected as the final reasoning scheme, or a reasoning scheme with the lowest risk value is selected as the final reasoning scheme, or a reasoning scheme with the lowest risk value is selected as the final reasoning scheme from the reasoning schemes with confidence higher than the third threshold, or a reasoning scheme with the highest confidence is selected from the reasoning schemes with risk values ​​lower than the fourth threshold, etc. The specific selection can be made according to actual user needs, and this application does not limit this.

[0201] For example, when EMS sends an inference plan to NMS, there are multiple situations. For different situations, the way to select the final inference plan may also be different:

[0202] (1) For solutions that only return confidence: NMS can select the solution with the highest confidence.

[0203] (2) For only returning risk: NMS can choose the solution with the lowest risk;

[0204] (3) Regarding the returned confidence and risk: NMS can customize the selection strategy (such as confidence priority, risk value priority, or weighting confidence and risk and then selecting based on the weighted value, etc.).

[0205] 806. The network element feeds back the execution result to the NMS.

[0206] 807. The NMS sends the execution result to the EMS.

[0207] 808. The EMS updates the data or knowledge base based on the execution results.

[0208] Among them, steps 806 to 808 can refer to the introduction of the aforementioned steps 606 to 608, which will not be repeated here.

[0209] Therefore, in the implementation manner of the present application, when EMS provides an inference scheme to NMS, it can also simultaneously provide the confidence and / or risk value of each inference scheme, so that when NMS selects the final inference scheme, it can more accurately select an inference scheme that meets the needs based on the confidence and / or risk value of each inference scheme.

[0210] Implementation Plan 2: Feedback one or more reasoning solutions with a confidence level higher than a confidence threshold and / or a risk value lower than a risk value threshold. Figure 9 , a flowchart of another reasoning method provided by this application is described as follows.

[0211] 901. EMS collects data from network elements and updates data and knowledge base.

[0212] Among them, step 901 can refer to the relevant description of the aforementioned step 601, which will not be repeated here.

[0213] 902. The NMS sends a reasoning request to the EMS. The reasoning request may carry a confidence threshold and / or a risk threshold.

[0214] The inference request may refer to the related description of the aforementioned step 602, and similarities will not be repeated here.

[0215] The difference is that, in the reasoning request, optionally, a confidence threshold and / or a risk threshold may be carried. The confidence threshold may be used to instruct the EMS to provide a reasoning solution with a confidence value higher than the confidence threshold, and the risk threshold may be used to instruct the EMS to provide a reasoning solution with a risk value lower than the risk threshold.

[0216] 903. The EMS obtains one or more reasoning solutions and the confidence and / or risk value of each solution.

[0217] Among them, step 903 can refer to the introduction of the aforementioned step 603, which will not be repeated here.

[0218] 904. The EMS sends one or more candidate reasoning solutions to the NMS, and the confidence of each solution exceeds the confidence threshold and / or the risk value is lower than the risk threshold.

[0219] After the EMS generates one or more reasoning schemes, it can screen out reasoning schemes whose confidence exceeds a confidence threshold and / or whose risk value is lower than a risk threshold from the one or more reasoning schemes as candidate reasoning schemes, and send the one or more candidate reasoning schemes to the NMS.

[0220] The confidence threshold and / or risk threshold may come from an inference request, or may be a pre-set default value, or a value received when the inference plan was last generated, etc. It may be adjusted according to the actual application scenario, and this application does not limit this.

[0221] Optionally, when sending one or more alternative reasoning schemes to the NMS, the EMS may also simultaneously send the confidence and / or risk value of each scheme, so that the NMS can be aware of the possible effects of using each alternative reasoning scheme to achieve the reasoning task.

[0222] 905. The NMS sends the final inference solution to the network element.

[0223] After receiving one or more candidate reasoning solutions sent by the EMS, the NMS may select one of the one or more candidate reasoning solutions as the final reasoning solution.

[0224] Specifically, when the NMS receives an alternative reasoning scheme, it can directly use the alternative reasoning scheme as the final reasoning scheme; when the NMS receives multiple alternative reasoning schemes, it can randomly select an alternative reasoning scheme as the final reasoning scheme.

[0225] Optionally, if the EMS simultaneously sends the confidence and / or risk value of each alternative reasoning scheme when sending the alternative reasoning scheme to the NMS, then when the NMS selects the final reasoning scheme from multiple alternative reasoning schemes, it can select the final reasoning scheme based on the confidence and / or risk value of each scheme. For example, select a reasoning scheme with the highest confidence as the final reasoning scheme, or select a reasoning scheme with the lowest risk value as the final reasoning scheme, or select the reasoning scheme with the lowest risk value as the final reasoning scheme among the reasoning schemes with confidence higher than the third threshold, or select the reasoning scheme with the highest confidence among the reasoning schemes with risk values ​​lower than the fourth threshold, etc. The specific selection can be based on actual user needs, and this application does not limit this.

[0226] 906. The network element feeds back the execution result to the NMS.

[0227] 907. The NMS sends the execution result to the EMS.

[0228] 908. The EMS updates the data or knowledge base based on the execution results.

[0229] Among them, steps 906 to 908 can refer to the introduction of the aforementioned steps 606 to 608, which will not be repeated here.

[0230] Therefore, in the implementation mode of the present application, when providing an inference solution to the NMS, the EMS can provide an inference solution with a confidence exceeding the confidence threshold and / or a risk value below the risk threshold, thereby providing an inference solution that meets the needs of the user, so that the user can more accurately screen out the inference solution that meets the needs. This is equivalent to expressing the user's preference for confidence and risk through confidenceThld (confidence threshold) and riskThld (risk threshold), and can filter out inference results that do not meet the user's preferences in advance, reduce the number of returned solutions, reduce the burden on users to check the results, and improve efficiency.

[0231] Implementation 3: Feedback a reasoning solution that meets the user's specified requirements

[0232] See also Figure 10 , a flowchart of another reasoning method provided by this application is described as follows.

[0233] 1001. EMS collects data from network elements and updates data and knowledge base.

[0234] Among them, step 1001 can refer to the relevant description of the aforementioned step 601, which will not be repeated here.

[0235] 1002. The NMS sends an inference request to the EMS, and the inference request may carry a policy indication.

[0236] The inference request may refer to the related description of the aforementioned step 602, and similarities will not be repeated here.

[0237] The difference is that, in the reasoning request, a policy indication can be optionally carried to instruct the EMS to provide a reasoning solution selected in accordance with the policy indicated by the policy indication. For example, the policy indication can be confidence priority, risk value priority, or a fusion of confidence and risk value, etc., which can be determined according to the actual application scenario, and this application does not limit this.

[0238] 1003. The EMS obtains one or more reasoning solutions and the confidence and / or risk value of each solution.

[0239] Among them, step 1003 can refer to the introduction of the aforementioned step 603, which will not be repeated here.

[0240] 1004. The EMS sends an alternative reasoning solution to the NMS, where the solution is a solution selected based on the policy indicated by the policy indication.

[0241] After EMS obtains one or more inference schemes, it can select one of the inference schemes as an alternative inference scheme based on the policy indication. The policy indication can come from the inference request, or it can be a pre-set policy indication, or it can be an indication received during the last inference scheme screening, etc. It can be determined according to the actual application scenario, and this application does not limit this.

[0242] Specifically, the strategy indicated by the strategy indication may include but is not limited to: confidence priority, risk value priority, or fusion of confidence and risk value indication, etc. Confidence priority can indicate that EMS can select alternative reasoning schemes based on the confidence of each reasoning scheme, such as selecting the reasoning scheme with the highest confidence as the alternative reasoning scheme; risk value priority can indicate that EMS can select alternative reasoning schemes based on the risk value of each reasoning scheme, such as selecting the scheme with the lowest risk value as the alternative reasoning scheme; fusion of confidence and risk value indication can indicate that EMS can select alternative reasoning schemes based on the confidence and risk value of each reasoning scheme, such as weighted fusion of the confidence and risk value of each reasoning scheme, and selecting an alternative reasoning scheme based on the value obtained after fusion, etc. Optionally, the risk value is usually negatively correlated with its probability of being selected, such as the lower the risk value, the higher the probability of being selected, and the higher the risk value, the lower the probability of being selected, thereby selecting an alternative reasoning scheme with a lower risk value.

[0243] 1005. The NMS sends the final inference solution to the network element.

[0244] In the implementation of the present application, the NMS can receive an alternative reasoning scheme sent by the EMS, so the NMS does not need to make a selection and can directly send the alternative reasoning scheme to the EMS as the final reasoning scheme. This is equivalent to the EMS having screened the alternative reasoning schemes based on the policy indication when sending the alternative reasoning schemes, thereby reducing the screening steps of the NMS and improving the efficiency of the NMS in selecting the final reasoning scheme.

[0245] 1006. The network element feeds back the execution result to the NMS.

[0246] 1007. The NMS sends the execution result to the EMS.

[0247] 1008. The EMS updates the data or knowledge base based on the execution results.

[0248] Among them, steps 1006 to 1008 can refer to the introduction of the aforementioned steps 606 to 608, which will not be repeated here.

[0249] In the implementation manner of the present application, the EMS can directly send the reasoning scheme that meets the requirements to the NMS based on the policy indication. The NMS can directly obtain the reasoning scheme that meets the requirements without further screening, thereby obtaining the required reasoning scheme more accurately.

[0250] The following is an illustrative introduction to possible application scenarios of the method provided in this application in combination with some possible application scenarios.

[0251] Application scenario 1: Fault location and repair scenario

[0252] The fault location cognitive reasoning service task is to locate hardware faults or software faults in the communication network, which may include but is not limited to the location of the fault, fault-related alarms, fault events and other information. The fault repair cognitive reasoning service task may include but is not limited to the fault handling methods, such as restarting, adjusting parameters, master-slave switching or resetting the board.

[0253] See also Figure 11 , a flowchart of another reasoning method provided by this application is described as follows.

[0254] 1101. Cognitive reasoning service producers collect data.

[0255] Among them, the cognitive reasoning service producer can collect data in advance, which may specifically include information about the network elements it manages, such as the logs, configurations, status, or detected events of each network element.

[0256] For example, the alarms, events, traffic statistics, statistical information, packet capture data, configuration information or operation logs generated by the network elements managed by the EMS can be collected. Among them, the alarms may include the alarms generated by the equipment when a fault is detected; the traffic statistics may include the statistical traffic information; the packet capture data includes the data intercepted by the equipment during operation, which is usually used to detect network security; the configuration information may include the configuration information used by each device during operation; the operation log includes the operation records of each device when it is actively or passively operated.

[0257] After receiving multiple requests, the cognitive reasoning service producer collects data generated by each device, or actively reports the changed data to the cognitive reasoning service producer after the data of the device changes, or after the execution result is generated after the previous execution of the reasoning task, the knowledge is stored in the form of triples, such as adding confidence attributes or risk attributes to the knowledge. If the use of certain knowledge ultimately leads to problems when the solution is issued, the risk of the corresponding knowledge will be increased.

[0258] 1102. The cognitive reasoning service consumer creates a fault location and repair service request to the cognitive reasoning service producer under the triggering of the intention service producer.

[0259] Typically, cognitive reasoning services may be generated by an intent service producer and trigger cognitive reasoning service consumers to create fault location and repair service requests to the cognitive reasoning service producer.

[0260] The request may include but is not limited to:

[0261] Task name: such as fault location, fault repair

[0262] Task description information: fault object (FailurePredictionObjec), potential fault type (PotentialFailureType), fault occurrence time (EventTime), fault severity, fault location, fault event content, etc.

[0263] The specific information carried in the request may also be referred to the records of relevant standards. For example, the standard document 3GPP TS28.104V1.1.0 records the information carried in the request in the fault location scenario.

[0264] In addition, in combination with the above-mentioned embodiments 1 to 3, the indication that can be carried in the inference request may include but is not limited to one or more of the following:

[0265] 1. Reporting a confidence indication (i.e. the aforementioned confidence indication) and / or reporting a risk indication (i.e. the aforementioned risk indication);

[0266] 2. confidenceThld (i.e. the aforementioned confidence threshold) and / or riskThld (i.e. the aforementioned risk threshold);

[0267] 3. sortPolicy (i.e. the aforementioned policy indication).

[0268] Among them, reporting the confidence indication can be used to instruct the cognitive reasoning service producer to provide feedback on the confidence (confidence) of each solution when providing fault location and repair solutions, and reporting the risk indication can be used to instruct the cognitive reasoning service producer to provide feedback on the risk value (risk) of each reasoning solution when providing fault location and repair solutions.

[0269] confidenceThld can be used to instruct cognitive reasoning service producers to provide solutions with a confidence level not lower than or higher than confidenceThld when providing fault location and repair solutions; riskThld can be used to instruct cognitive reasoning service producers to provide solutions with a risk value lower than or no higher than riskThld when providing fault location and repair solutions.

[0270] sortPolicy can be used to instruct cognitive reasoning service producers to select a solution for feedback according to the strategy indicated by sortPolicy when providing fault location and repair solutions. For example, sortPolicy can be set to confidencePriority (confidence priority), riskPriority (low risk priority), or a custom confidence and risk weighted calculation formula, such as sortPolicy = 0.5*confidence+0.5*(1-risk).

[0271] In the implementation mode of the present application, by enhancing the interface between the cognitive reasoning service producer and the cognitive reasoning service consumer, the confidence and risk information of the cognitive reasoning request return field is supplemented, which assists the users on the cognitive reasoning service consumer side to better select and judge solutions and improve the efficiency of the problem closure loop.

[0272] 1103. The cognitive reasoning service producer performs fault location and repair service reasoning, generates and selects reasoning solutions that meet the requirements.

[0273] After receiving a request to create a fault location and repair service, the cognitive reasoning service producer can read relevant knowledge and data according to the request, including the confidence and risk information in the knowledge triple. Execute the cognitive reasoning algorithm or run the neural network, calculate the confidence and / or risk of the root cause of the fault and the repair operation during the reasoning process, and generate a fault location and repair service reasoning solution. The fault location and repair reasoning solution may include

[0274] Specifically, when executing fault location and repair service reasoning, it can be completed through the algorithm within the producer. For example, when reasoning about the root cause, the probability of the root cause can be calculated through algorithms such as random walk. This method requires enhancing the algorithm capabilities within the cognitive reasoning producer; alternatively, the knowledge in the knowledge base can be combined to generate a fault location and repair service reasoning solution.

[0275] For example, in a knowledge base, knowledge can be stored in the form of triples. The knowledge base itself carries confidence and risk information. By enriching the definition of triple attributes in the knowledge graph, confidence and risk attributes are added to the traditional triple of head entity-relationship-tail entity. Figure 12 As shown in the figure, for example, for a certain fault phenomenon, the probability of causing alarm 1 is 0.6, the probability of causing alarm 2 is 0.2, the probability of causing KPI x abnormality is 0.8, and in the case of KPI x abnormality, the probability of causing alarm 3 is 0.7, the probability of causing alarm 4 is 0.6, etc. In the case of alarm 1, the risk of repairing by restarting is 0.5, and so on.

[0276] 1104. The cognitive reasoning service producer feeds back one or more alternative fault location and repair solutions to the cognitive reasoning service consumer.

[0277] When sending back a fault location and repair service reasoning solution, the specific returned solution can be determined based on the instructions carried in the reasoning request.

[0278] Such situations can be divided into but not limited to the following:

[0279] 1. For the confidence indication and / or risk indication carried in the reasoning request: when feeding back the reasoning solution, confidence and / or risk are synchronously returned, which can be expressed as: [{"solution":x, "confidence":0.5, "risk":0.7},...];

[0280] 2. For confidenceThld and / or riskThld carried in the reasoning request: when feeding back the reasoning solution, confidence>(≥)confidenceThld and / or risk<(≤)riskThld can be filtered out from the generated solutions;

[0281] 3. For the reasoning request with sortPolicy: For different strategies, you can filter out a reasoning solution from the generated reasoning solutions based on sortPolicy. For example, if sortPolicy = confidencePriority, you can filter out the reasoning solution with the highest confidence for feedback; if sortPolicy = riskPriority, you can directly return the solution with the lowest risk value; if sortPolicy is a custom weighted fusion formula, you can calculate according to the formula and return a reasoning solution based on the calculated score, such as returning the solution with the highest score.

[0282] In the implementation manner of the present application, corresponding reasoning solutions can be adaptively returned for different instructions carried in the reasoning request. Therefore, for various situations, the requesting party can accurately select the required solution, thereby achieving more accurate fault location and repair.

[0283] For the above situation 1, when the cognitive reasoning service consumer sends a cognitive reasoning service request, in addition to the existing basic task information description, the confidence reporting indication and risk reporting indication are added. The cognitive reasoning service producer returns the reasoning result and the corresponding confidence and risk according to the request content, indicating the probability that the result is correct and the probability that the result may cause risk. This allows users to have a clearer understanding of the reliability of each solution, choose the solution they need, and improve the efficiency of the problem closed loop. That is, when the cognitive reasoning service producer returns multiple reasoning results, it carries the confidence and / or risk information of the results, indicating the confidence and risk of the results, helping users to better choose the appropriate solution according to their own needs. That is, when the cognitive reasoning service producer returns multiple reasoning results, it carries the confidence and / or risk information of the results, indicating the confidence and risk of the results, helping users to better choose the appropriate solution according to their own needs.

[0284] For the above situation 2, when the cognitive reasoning service consumer sends a request, the request content contains the confidenceThld (confidence threshold) and / or riskThld (risk threshold) fields. The cognitive reasoning service producer can perform pruning in advance based on the threshold requirements during the reasoning process to speed up the reasoning efficiency and only return the results that meet the conditions of confidence>confidenceThld AND riskThld. <riskThld的结果,减少返回的无效结果数量,降低消费者选择方案的负担。即通过confidenceThld与riskThld表达用户对置信度和风险的偏好,可以提前过滤掉不满足用户偏好的推理结果,减少返回方案的数量,降低用户排查结果的负担,提高效率。

[0285] For situation 3, a solution selection strategy sortPolicy is added. When sending a cognitive reasoning service request, the selection strategy of the user's preference is carried. The cognitive reasoning service producer directly completes the selection of multiple solutions internally, and the cognitive reasoning service consumer directly receives the optimal solution, which saves the time of individual selection and analysis on the consumer side and improves the efficiency of the problem closed loop. That is, the cognitive reasoning service request carries the user-defined solution selection strategy field sortPolicy, and the cognitive reasoning service producer directly returns only one optimal result based on the user's preference, eliminating the need to return multiple results for the consumer to choose.

[0286] 1105. The cognitive reasoning service consumer sends the final fault location and repair plan to the network element.

[0287] After receiving one or more fault location and repair solutions sent by the cognitive reasoning service producer, the cognitive reasoning service consumer selects a fault location and repair solution as the final fault location and repair solution, and sends the final fault location and repair solution to the relevant network elements.

[0288] Generally, if a cognitive reasoning service consumer receives a fault location and repair solution, the fault location and repair solution can be used as the final fault location and repair solution and sent to the corresponding network element.

[0289] If the cognitive reasoning service consumer receives multiple fault location and repair solutions, then one fault location and repair solution can be selected from the multiple fault location and repair solutions as the final fault location and repair solution. It can be selected randomly or according to a pre-set rule. For example, if the confidence and / or risk of each solution is received at the same time, the solution with the highest confidence or the lowest confidence can be selected, or the confidence and risk can be weighted and fused, and the solution with the highest or lowest value can be selected based on the fused score, etc. The specific adjustment can be made according to the actual application scenario.

[0290] In some possible scenarios, when multiple reasoning solutions are received, the intent service producer may also select one of the reasoning solutions as the final reasoning solution. For example, in the fault location and repair solution of the embodiment of the present application, the intent service producer may select the final fault location and repair solution.

[0291] In addition, cognitive reasoning service consumers can directly send received fault location and repair solutions to network elements, or they can generate different control data based on different network elements and then send them. For example, for fault location and repair solutions, data or configurations for fault repair can be sent to relevant network elements based on the fault location results to achieve repair after fault location.

[0292] 1106. The network element feeds back the repair result to the cognitive reasoning service consumer.

[0293] After receiving the fault location and repair plan sent by the cognitive reasoning service consumer, the network element can execute the corresponding plan to achieve fault repair. For example, for fault repair, switching, restarting, rollback or parameter adjustment can be used. Each network element can perform the corresponding operation and feedback the repair results after execution to the cognitive reasoning service consumer, such as whether the fault has been repaired after executing the received repair plan.

[0294] 1107. The cognitive reasoning service consumer feeds back the repair result to the intent service producer.

[0295] After the cognitive reasoning service consumer receives the repair result fed back by the network element, it can feed back the repair result to the intent service producer so that the intent service producer can be informed of the repair status of the fault, or determine whether its intention has been achieved. If not, the aforementioned step 1102 can be re-executed.

[0296] 1108. The cognitive reasoning service consumer sends the repair result to the cognitive reasoning service producer.

[0297] The cognitive reasoning service consumer may also send a repair result to the cognitive reasoning service producer to instruct the cognitive reasoning service producer to update the knowledge base.

[0298] 1109. The cognitive reasoning service producer updates the knowledge base.

[0299] After the cognitive reasoning service producer receives the execution result of the reasoning solution, it can update the knowledge base. Based on the final execution result, the cognitive reasoning service consumer sends a request to update the knowledge confidence and risk to the cognitive reasoning service producer, requesting to update the knowledge base in the producer. Generally, if the intended goal is finally achieved based on some knowledge in the knowledge base, it means that the confidence of the knowledge is high, and the confidence attribute of the knowledge triple can be increased accordingly; if the use of some knowledge eventually leads to problems when the solution is issued, the risk of the corresponding knowledge should be increased.

[0300] Generally, the knowledge-driven approach needs to enhance knowledge mining capabilities, supplement the confidence and risk information in the knowledge base, and can update the knowledge base in reverse based on the application results of the solution (consumer feedback). Figure 12 ,After the fault repair plan is generated and executed in the network element, the knowledge in the knowledge base can be modified in reverse ,based on the execution result, for example, the confidence and / or risk corresponding to each fault repair ,operation under each alarm can be adjusted.

[0301] In the implementation manner of the present application, for the situation of fault repair and location, the cognitive reasoning service producer can generate a variety of feasible fault location and repair solutions, as well as the confidence and / or risk corresponding to each solution. When feeding back the fault and repair solutions to the cognitive reasoning service consumer, the confidence and / or risk corresponding to each solution can be fed back synchronously, so that the cognitive reasoning service consumer can select the required fault location and repair solution based on the confidence and / or risk corresponding to each solution. Furthermore, the cognitive reasoning service producer can filter the generated fault location and repair solutions based on the confidence and / or risk corresponding to each solution, such as filtering according to confidenceThld and / or riskThld, or according to sortPolicy, so that the solution that meets its needs can be fed back to the cognitive reasoning service consumer, reducing the steps that the cognitive reasoning service consumer needs to choose, or eliminating the need for the cognitive reasoning service consumer to make a choice, etc., so that the cognitive reasoning service can obtain the required solution more directly and accurately.

[0302] Application scenario 2: Recommended energy-saving scenarios

[0303] See also Figure 13 , a flowchart of another reasoning method provided by this application is described as follows.

[0304] 1301. Cognitive reasoning service producers collect data.

[0305] Among them, step 1301 is similar to the aforementioned step 1101. Please refer to the relevant description of the aforementioned step 1101 for details, which will not be repeated here.

[0306] For energy-saving recommendation scenarios, the data usually collected by cognitive reasoning service producers may include but are not limited to: network element operating performance, quality of experience (QoE) data, network element configuration data or network analysis data, etc. The required data can be obtained according to the actual application scenario.

[0307] 1302. The cognitive reasoning service consumer creates an energy-saving decision cognitive reasoning service request to the cognitive reasoning service producer under the triggering of the intention service producer.

[0308] Similar to the aforementioned step 1102, the similarities are not repeated here, and the differences are introduced below.

[0309] The energy-saving decision-making cognitive reasoning service request may include an energy-saving action area, an energy-saving target, and an experience constraint or an action time period.

[0310] 1303. The cognitive reasoning service producer performs energy-saving decision-making service reasoning to generate energy-saving solutions.

[0311] After the producer of the careful reasoning service receives the cognitive reasoning service request for creating an energy-saving decision, it can perform energy-saving decision service reasoning based on the information carried in the cognitive reasoning service request for creating an energy-saving decision, and generate an energy-saving plan based on knowledge-driven and / or data-driven.

[0312] The energy-saving plan may specifically include a plan for energy-saving management of the equipment to be managed for the energy-saving area or energy-saving time period mentioned in the request. Specifically, the energy-saving plan may include, but is not limited to: objects / network elements to be energy-saving, energy-saving types, network load trends, energy-saving suggestions (energy-saving strategies, such as shutting down certain network elements during energy-saving periods, shutting down certain network elements in energy-saving areas, etc.) or grid status after energy-saving and other information.

[0313] 1304. The cognitive reasoning service producer feeds back one or more energy savings to the cognitive reasoning service.

[0314] 1305. The cognitive reasoning service consumer sends the final energy-saving plan to the network element.

[0315] 1306. The network element feeds back the energy-saving results to the cognitive reasoning service consumer.

[0316] 1307. The cognitive reasoning service consumer feeds back the energy-saving results to the intention service producer.

[0317] 1308. The cognitive reasoning service consumer sends the energy saving result to the cognitive reasoning service producer.

[0318] 1309. Update knowledge base.

[0319] Among them, steps 1304 to 1309 are similar to the aforementioned steps 1104 to 1109, except that the aforementioned fault location and repair are replaced by energy saving. For details, please refer to the relevant descriptions of the aforementioned steps 1104 to 1109, which will not be repeated here.

[0320] In the implementation mode of the present application, for the energy-saving recommendation scenario, the cognitive reasoning service producer can still generate one or more energy-saving schemes (or energy-saving recommendation schemes or energy-saving decision schemes, etc.), as well as the confidence and / or risk value corresponding to each energy-saving scheme, and feedback the generated energy-saving schemes and the confidence and / or risk value corresponding to each energy-saving scheme according to the instructions of the cognitive reasoning service consumer, so that the cognitive reasoning service consumer can screen the schemes based on the confidence and / or risk value of each scheme to accurately obtain the required energy-saving scheme. Or instruct the cognitive reasoning service producer to screen the generated energy-saving schemes and then feedback, so that the cognitive reasoning service consumer can directly and accurately obtain the required energy-saving scheme without screening.

[0321] To facilitate understanding of the effects achieved by the solution provided in this application, a comparative explanation is provided below in combination with existing solutions and the solution provided in this application.

[0322] The following is a comparison between the existing solutions and the solution provided by this application from the two dimensions of NMS and EMS.

[0323] See Table 2 for a comparison between the steps performed by the EMS in the existing scheme and the scheme provided in this application.

[0324]

[0325] Table 2

[0326] See Table 3 for a comparison between the steps performed by the NMS in the existing solution and the solution provided in this application.

[0327]

[0328]

[0329] Table 3

[0330] Obviously, through the solution provided by the embodiment of the present application, cognitive reasoning service consumers can obtain more accurate reasoning solutions. For example, they can filter based on the confidence and / or risk of each received reasoning solution to obtain a reasoning solution that meets their needs; or they can carry instructions in the reasoning request, such as confidenceThld / riskThld / sortPolicy, so that the cognitive reasoning service producer can filter the generated solutions, so that the cognitive reasoning service consumer can directly obtain the reasoning solution that meets their needs.

[0331] The above is an introduction to the method flow provided by the present application. The following is an introduction to the device for executing the above method.

[0332] See also Figure 14, a structural schematic diagram of an inference device provided in this application is described as follows.

[0333] The reasoning device may be deployed in the aforementioned EMS or cognitive reasoning service producer, etc. The reasoning device may specifically include:

[0334] The transceiver module 1401 is used to receive a reasoning request, wherein the reasoning request is used to request a solution for achieving a reasoning task, wherein the reasoning task includes a task for completing a user's intended goal;

[0335] A processing module 1402 is configured to obtain a first reasoning scheme set according to the reasoning request, wherein the first reasoning scheme set includes at least one reasoning scheme and at least one of a confidence level or a risk value corresponding to each reasoning scheme, wherein each reasoning scheme is used to achieve the reasoning task, the confidence level is used to indicate a probability of achieving the reasoning task when the reasoning scheme is used, and the risk value is used to indicate a probability of changing a user experience when the reasoning scheme is used to achieve the reasoning task;

[0336] The transceiver module 1401 is further configured to send a second reasoning solution set, where the second reasoning solution set includes at least one solution in the first reasoning solution set.

[0337] In a possible implementation, the processing module 1402 is further configured to use the first reasoning solution set as the second reasoning solution set;

[0338] The transceiver module 1401 is specifically configured to send the second reasoning solution set.

[0339] In a possible implementation, the reasoning request also carries a confidence indication and / or a risk indication, wherein the confidence indication is used to request provision of a confidence corresponding to the reasoning scheme, and the risk indication is used to request provision of a risk value corresponding to the reasoning scheme.

[0340] In a possible implementation, the processing module 1402 is specifically configured to filter out, from the first reasoning solution set, solutions whose confidences are higher than a confidence threshold and / or solutions whose risk values ​​are higher than a risk threshold, to obtain the second reasoning solution set;

[0341] The transceiver module 1401 is specifically configured to send the second reasoning solution set.

[0342] In a possible implementation manner, the reasoning request also carries the confidence threshold and / or the risk threshold.

[0343] In a possible implementation, the processing module 1402 is specifically configured to select one of the reasoning schemes from the first reasoning scheme set as a scheme of the second reasoning scheme set according to at least one of the confidence or risk value corresponding to each reasoning scheme;

[0344] The transceiver module 1401 is specifically configured to send the second reasoning solution set.

[0345] In a possible implementation, the processing module 1402 is specifically used to filter out one of the inference schemes that meets the filtering conditions from the first inference scheme set, and the filtering conditions include at least one of the following: an inference scheme with the highest confidence, an inference scheme with the lowest risk value, or an inference scheme with the highest fusion value after integrating the confidence and risk values.

[0346] In a possible implementation, the inference request further carries a policy indication, where the policy indication is used to indicate at least one of the screening conditions.

[0347] In a possible implementation, the second reasoning scheme set further includes at least one of a confidence level or a risk value corresponding to the reasoning scheme.

[0348] In one possible implementation, the processing module 1402 is specifically used to: obtain monitoring data and a knowledge base corresponding to the reasoning task, the monitoring data including data of equipment related to the reasoning task, and the knowledge in the knowledge base including rules for completing the reasoning task; and obtain the first reasoning scheme set based on the monitoring data and the knowledge.

[0349] In a possible implementation, the transceiver module 1401 is further configured to obtain an execution result, where the execution result is obtained after executing the reasoning scheme included in the second reasoning scheme set;

[0350] The processing module 1402 is further configured to update the knowledge according to the execution result to obtain updated knowledge.

[0351] In a possible implementation, the reasoning task includes: at least one of a fault location task, a fault repair task, a configuration verification task, a hidden danger identification task or a solution recommendation task, wherein the fault location task indicates locating the root cause of the device fault, the fault repair task indicates repairing the device that has caused the fault, the configuration verification task indicates verifying the configuration of the device, the hidden danger identification task indicates obtaining the probability of the device causing a fault, and the solution recommendation task indicates obtaining at least one solution for executing the target task.

[0352] See also Figure 15, a structural schematic diagram of an inference device provided in this application is described as follows.

[0353] The reasoning device may be deployed in the aforementioned NMS or cognitive reasoning service consumer, and may include:

[0354] The transceiver module 1501 is used to send a reasoning request, wherein the reasoning request is used to request a solution for achieving a reasoning task, wherein the reasoning task includes a task for completing a user's intended goal;

[0355] The transceiver module 1501 is also used to receive a second reasoning scheme set, the second reasoning scheme set includes at least one reasoning scheme, the second reasoning scheme set is obtained from the first reasoning scheme set, the first reasoning scheme set is a set generated based on the reasoning request, the first reasoning scheme set includes at least one reasoning scheme, and at least one of the confidence or risk value corresponding to each reasoning scheme, each reasoning scheme is used to achieve the reasoning task, the confidence is used to indicate the probability of achieving the reasoning task when using the reasoning scheme, and the risk value is used to indicate the probability of changing the user experience when using the reasoning scheme to achieve the reasoning task.

[0356] In a possible implementation, the second reasoning scheme set further includes: at least one of a confidence level or a risk value corresponding to each reasoning scheme in the at least one reasoning scheme.

[0357] In a possible implementation, the reasoning request also carries a confidence indication and / or a risk indication, wherein the confidence indication is used to request provision of a confidence corresponding to the reasoning scheme, and the risk indication is used to request provision of a risk value corresponding to the reasoning scheme.

[0358] In a possible implementation manner, the confidence of the reasoning scheme included in the second reasoning scheme set is higher than a confidence threshold and / or the risk value is higher than a risk threshold.

[0359] In a possible implementation, the reasoning request also carries the confidence threshold and / or the risk threshold.

[0360] In a possible implementation, the second reasoning solution set includes one reasoning solution, and the one reasoning solution is screened from the first reasoning solution set according to the confidence level and / or the risk value.

[0361] In a possible implementation manner, the inference request further carries a policy indication, where the policy indication is used to indicate providing an inference solution.

[0362] In a possible implementation, the policy indication also indicates at least one of the screening conditions, and the screening condition includes at least one of the following: an inference scheme with the highest confidence, an inference scheme with the lowest risk value, or an inference scheme with the highest fusion value after integrating the confidence and the risk value.

[0363] In a possible implementation manner, the transceiver module 1501 is further configured to:

[0364] Sending a reasoning solution in the second reasoning solution set;

[0365] receiving an execution result, where the execution result is an execution result obtained by using a reasoning scheme in the second reasoning scheme set to accomplish the reasoning task;

[0366] The execution result is sent, where the execution result is used to update knowledge, where the knowledge is used to determine the second reasoning solution set, and the knowledge includes rules for completing the reasoning task.

[0367] like Figure 16 FIG. 1 is a schematic diagram of the hardware structure of a computer device 160 provided in an embodiment of the present application. The computer device 160 can be used to implement the functions of the NMS or EMS in the above method.

[0368] Figure 16 The computer device 160 shown may include: a processor 1601 , a memory 1602 , a communication interface 1603 , and a bus 1604 . The processor 1601 , the memory 1602 , and the communication interface 1603 may be connected via the bus 1604 .

[0369] The processor 1601 is the control center for generating the computer device 160, and may be a general-purpose central processing unit (CPU) or other general-purpose processors, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0370] As an example, the processor 1601 may include one or more CPUs, such as Figure 16 CPU 0 and CPU 1 are shown in .

[0371] The memory 1602 may be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM) or other type of dynamic storage device that can store information and instructions, an electrically erasable programmable read-only memory (EEPROM), a 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 to these.

[0372] In one possible implementation, the memory 1602 may exist independently of the processor 1601. The memory 1602 may be connected to the processor 1601 via a bus 1604 and used to store data, instructions, or program codes. When the processor 1601 calls and executes the instructions or program codes stored in the memory 1602, the method provided in the embodiment of the present application can be implemented.

[0373] In another possible implementation, the memory 1602 may also be integrated with the processor 1601 .

[0374] The communication interface 1603 is used for connecting the computer device 160 with other devices through a communication network, and the communication network may be Ethernet, a radio access network (RAN), a wireless local area network (WLAN), etc. The communication interface 1603 may include a receiving unit for receiving data and a sending unit for sending data.

[0375] The bus 1604 may be an industry standard architecture (ISA) bus, a peripheral component interconnect (PCI) bus, or an extended industry standard architecture (EISA) bus. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 16 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0376] It should be pointed out that Figure 16 The structure shown in the figure does not constitute a limitation on the computer device 160, exceptFigure 16 In addition to the components shown, computer device 160 may include more or fewer components than shown, or combine certain components, or arrange components differently.

[0377] Through the description of the above implementation mode, the technicians in the field can clearly understand that the present application can be implemented by means of software plus necessary general hardware, and of course, it can also be implemented by special hardware including special integrated circuits, special CPUs, special memories, special components, etc. In general, all functions completed by computer programs can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be various, such as analog circuits, digital circuits or special circuits. However, for the present application, software program implementation is a better implementation mode in more cases. Based on such an understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a readable storage medium, such as a computer floppy disk, a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, etc., including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0378] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product.

[0379] In an embodiment of the present application, a computer-readable storage medium is also provided, in which a program for training a model or performing an inference task is stored. When the program is run on a computer, the computer performs the above-mentioned Figures 4 to 12 All or part of the steps in the method described in the embodiment shown in FIG.

[0380] The embodiment of the present application also provides a digital processing chip. The digital processing chip integrates a circuit and one or more interfaces for implementing the above-mentioned processor or the function of the processor. When the digital processing chip integrates a memory, the digital processing chip can complete the method steps of any one or more embodiments in the above-mentioned embodiments. When the digital processing chip does not integrate a memory, it can be connected to an external memory through a communication interface. The digital processing chip implements the method steps of any one or more embodiments in the above-mentioned embodiments according to the program code stored in the external memory.

[0381] A computer program product is also provided in the embodiment of the present application, and the computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may 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 may be transmitted from a website site, a computer, a server, or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server, or data center. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a server or a data center that includes one or more available media integrations. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state hard disk Solid State Disk (SSD)), etc.

[0382] A person skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the relevant hardware through a program, and the program can be stored in a computer-readable storage medium, which can include: ROM, RAM, disk or CD, etc.

[0383] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described here can be implemented in a sequence other than the content illustrated or described here. The term "and / or" in the present application is only a kind of association relationship describing the associated objects, indicating that there can be three kinds of relationships, for example, A and / or B can be represented: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are an "or" relationship. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or modules is not necessarily limited to those steps or modules clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices. The naming or numbering of the steps in this application does not mean that the steps in the method flow must be executed in the time / logical sequence indicated by the naming or numbering. The named or numbered process steps can change the execution order according to the technical purpose to be achieved, as long as the same or similar technical effects can be achieved. The division of modules in this application is a logical division. There may be other division methods when it is implemented in actual applications. For example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between each other shown or discussed can be through some ports, and the indirect coupling or communication connection between modules can be electrical or other similar forms, which are not limited in this application. In addition, the modules or submodules described as separate components may or may not be physically separated, may or may not be physical modules, or may be distributed in multiple circuit modules, and some or all of the modules can be selected according to actual needs to achieve the purpose of the present application solution.

Claims

1. A reasoning method, characterized in that, it includes: Receiving a reasoning request for requesting a solution for achieving a reasoning task, where the reasoning task includes a task of completing a user intention goal; Obtaining a first set of reasoning solutions according to the reasoning request, where the first set of reasoning solutions includes at least one reasoning solution and at least one of the confidence level or risk value corresponding to each reasoning solution, and each reasoning solution is used to achieve the reasoning task, the confidence level is used to indicate the probability of achieving the reasoning task when using the reasoning solution, and the risk value is used to indicate the probability of changing the user experience when using the reasoning solution to achieve the reasoning task; Sending a second set of reasoning solutions, where the second set of reasoning solutions includes at least one solution in the first set of reasoning solutions.

2. The method according to claim 1, characterized in that, the sending of the second set of reasoning solutions includes: Regarding the first set of reasoning solutions as the second set of reasoning solutions; Sending the second set of reasoning solutions.

3. The method according to claim 2, characterized in that, the reasoning request also carries a confidence level indication and / or a risk indication, where the confidence level indication is used to request the provision of the confidence level corresponding to the reasoning solution, and the risk indication is used to request the provision of the risk value corresponding to the reasoning solution.

4. The method according to claim 1, characterized in that, the sending of the second set of reasoning solutions includes: Screening out the solutions with a confidence level higher than the confidence level threshold and / or the solutions with a risk value higher than the risk threshold from the first set of reasoning solutions to obtain the second set of reasoning solutions; Sending the second set of reasoning solutions.

5. The method according to claim 4, characterized in that, the reasoning request also carries the confidence level threshold and / or the risk threshold.

6. The method according to claim 1, characterized in that, the sending of the second set of reasoning solutions further includes: Screening out one reasoning solution from the first set of reasoning solutions as the solution of the second set of reasoning solutions according to at least one of the confidence level or risk value corresponding to each reasoning solution; Sending the second set of reasoning solutions.

7. The method according to claim 6, characterized in that, the screening out one reasoning solution from the first set of reasoning solutions according to at least one of the confidence level or risk value corresponding to each reasoning solution includes: Screening out one reasoning solution that meets the screening conditions from the first set of reasoning solutions, where the screening conditions include at least one of the following: one reasoning solution with the highest confidence level, one reasoning solution with the lowest risk value, or one reasoning solution with the highest combined value after fusing the confidence level and the risk value.

8. The method according to claim 7, characterized in that, the reasoning request also carries a strategy indication, where the strategy indication is used to indicate at least one of the screening conditions.

9. The method according to any one of claims 4-8, characterized in that, the second set of reasoning solutions also includes at least one of the confidence level or risk value corresponding to the reasoning solution.

10. The method according to any one of claims 1-9, characterized in that, obtaining the first set of inference solutions according to the inference request includes: obtaining monitoring data corresponding to the inference task and a knowledge base, where the monitoring data includes data of devices related to the inference task, and the knowledge in the knowledge base includes rules for completing the inference task; obtaining the first set of inference solutions according to the monitoring data and the knowledge.

11. The method according to claim 10, characterized in that, the method further includes: obtaining an execution result, where the execution result is obtained after executing the inference solutions included in the second set of inference solutions; updating the knowledge according to the execution result to obtain updated knowledge.

12. The method according to any one of claims 1-11, characterized in that, the inference task includes at least one of a fault location task, a fault repair task, a configuration verification task, a potential hazard identification task, or a solution recommendation task. The fault location task is used to locate the root cause of a fault in a device, the fault repair task is used to repair the device with a fault, the configuration verification task is used to verify the configuration of a device, the potential hazard identification task is used to obtain the probability of a device having a fault, and the solution recommendation task is used to obtain at least one solution for executing a target task.

13. An inference method, characterized in that, including: sending an inference request, where the inference request is used to request a solution for achieving an inference task, and the inference task includes a task for completing a user intention goal; receiving a second set of inference solutions, where the second set of inference solutions includes at least one inference solution, the second set of inference solutions is obtained from a first set of inference solutions, the first set of inference solutions is a set generated based on the inference request, the first set of inference solutions includes at least one inference solution, and at least one of the confidence level or risk value corresponding to each inference solution. Each inference solution is used to achieve the inference task, the confidence level is used to indicate the probability of achieving the inference task when using the inference solution, and the risk value is used to indicate the probability of changing the user experience when using the inference solution to achieve the inference task.

14. The method according to claim 13, characterized in that, the second set of inference solutions further includes at least one of the confidence level or risk value corresponding to each inference solution in the at least one inference solution.

15. The method according to claim 14, characterized in that, the inference request further carries a confidence level indication and / or a risk indication, where the confidence level indication is used to request a confidence level corresponding to the inference solution, and the risk indication is used to request a risk value corresponding to the inference solution.

16. The method according to claim 13, characterized in that, the confidence level of the inference solutions included in the second set of inference solutions is higher than a confidence level threshold and / or the risk value is higher than a risk threshold.

17. The method according to claim 16, characterized in that, the inference request further carries the confidence level threshold and / or the risk threshold.

18. The method according to claim 13, wherein, the second set of inference schemes includes one inference scheme, and the one inference scheme is screened from the first set of inference schemes according to the confidence level and / or the risk value.

19. The method according to claim 18, wherein, a policy indication is further carried in the inference request, and the policy indication is used to indicate to provide one inference scheme.

20. The method according to claim 19, wherein the policy indication further indicates at least one of the screening conditions, and the screening conditions include at least one of the following: one inference scheme with the highest confidence level, one inference scheme with the lowest risk value, or one inference scheme with the highest fusion value after fusing the confidence level and the risk value.

21. The method according to any one of claims 13-20, wherein, the method further includes: sending the inference scheme in the second set of inference schemes; receiving an execution result, where the execution result is an execution result obtained by using the inference scheme in the second set of inference schemes to achieve the inference task; sending the execution result, and the execution result is used to update knowledge, and the knowledge is used to determine the second set of inference schemes, and the knowledge includes rules for completing the inference task.

22. An inference method, wherein, it includes: a first device sends an inference request to a second device, and the inference request is used to request to provide a scheme for achieving an inference task, and the inference task includes a task for completing a user intention target; the second device obtains a first set of inference schemes according to the inference request, the first set of inference schemes includes at least one inference scheme, and at least one of the confidence level or the risk value corresponding to each inference scheme, each inference scheme is used to achieve the inference task, the confidence level is used to indicate the probability of achieving the inference task when using the inference scheme, and the risk value is used to indicate the probability of changing the user experience when using the inference scheme to achieve the inference task; the second device sends a second set of inference schemes to the first device, and the second set of inference schemes includes at least one scheme in the first set of inference schemes.

23. The method according to claim 22, wherein, the second device sending the second set of inference schemes to the first device includes: the second device takes the first set of inference schemes as the second set of inference schemes; the second device sends the second set of inference schemes to the first device.

24. The method according to claim 22, wherein, the second device sending the second set of inference schemes to the first device includes: the second device screens out the schemes with the confidence level higher than the confidence level threshold and / or the schemes with the risk value higher than the risk threshold from the first set of inference schemes to obtain the second set of inference schemes; the second device sends the second set of inference schemes to the first device.

25. The method according to claim 22, wherein, the second device sending the second set of inference schemes to the first device further includes: The second device selects, from the first set of inference schemes, one inference scheme as the scheme of the second set of inference schemes according to at least one of the confidence levels or risk values corresponding to each inference scheme; The second device sends the second set of inference schemes to the first device.

26. An inference device, characterized in that, it includes: a transceiver module configured to perform the step of receiving data or sending data as described in any one of claims 1-12; a processing module configured to perform the step of processing data as described in any one of claims 1-12.

27. An inference device, characterized in that, it includes: a transceiver module configured to perform the step of receiving data or sending data as described in any one of claims 13-21; a processing module configured to perform the step of processing data as described in any one of claims 13-21.

28. A communication system, characterized in that, the system includes a first device and a second device; the second device is configured to perform the method as described in any one of claims 13 to 21; the first device is configured to perform the method as described in any one of claims 1 to 12.

29. An inference device, characterized in that, it includes: a memory storing executable program instructions; and a processor, the processor is configured to be coupled with the memory, read and execute the instructions in the memory, and trigger the inference device to implement the method as described in any one of claims 1 to 12 or 13 to 21.

30. A computer-readable storage medium, including instructions, when running on a computer, cause the computer to execute the method as described in any one of claims 1 to 12, or execute the method as described in any one of claims 13 to 21.

Images