Network optimization method and device, and storage medium

Abstract

The invention discloses a network optimization method and device and a storage medium, and the method comprises the steps: obtaining a user intention, processing the user intention through an intention engine, and obtaining a first intention configuration which comprises an intention optimization strategy corresponding to a new intention type, the intention optimization strategy is obtained by predicting based on an initial network index; and optimizing a current network through the first intention configuration, so that a target network index of the current network reaches an intention target corresponding to the user intention. According to the invention, the optimization efficiency of the intention engine on the communication network is improved.

Description

Technical Field

[0001] This application relates to the technical field of wireless communication network assurance, and particularly to a network optimization method, device, and storage medium. Background Art

[0002] In recent years, with the application of intent networks in communication networks, when a communication network encounters an emergency, resulting in a sudden increase in specific service traffic and deterioration of terminal network metrics, specific intent goals can be issued to the intent network, enabling network optimization and greatly reducing the workload of operation and maintenance personnel.

[0003] Currently, the types of intents supported by intent networks are very limited. There are local differences in the problems and solutions existing in different networks. Often, the problem of insufficient achievement rate of original intent goals is encountered. It is necessary to support such new problems from intent collection to solution design and then to R & D delivery, resulting in the inability to expand the types of intents without relying on the productivity of the development team, slowing down the speed of intent type expansion, and leading to low efficiency in communication network optimization. Summary of the Invention

[0004] The main purpose of this application is to provide a network optimization method, device, and storage medium, aiming to solve the technical problem in related technologies that when using an intent network to optimize a communication network, it is necessary to support new intent types from intent collection to solution design and then to R & D delivery, slowing down the speed of intent type expansion and resulting in low efficiency in communication network optimization.

[0005] To achieve the above object, an embodiment of this application provides a network optimization method, which includes:

[0006] Obtain a user intent, and process the user intent using an intent engine to obtain a first intent configuration, where the first intent configuration includes an intent optimization strategy corresponding to a new intent type, and the intent optimization strategy is predicted based on initial network metrics;

[0007] Optimize the current network through the first intent configuration so that the target network metrics of the current network reach the intent goal corresponding to the user intent.

[0008] This application also provides a network optimization device. The network optimization device is an entity node device, and the network optimization device includes: a memory, a processor, and a program of the network optimization method stored on the memory and executable on the processor. When the program of the network optimization method is executed by the processor, the steps of the network optimization method as described above can be implemented.

[0009] To achieve the above object, a storage medium is further provided. A network optimization program is stored on the storage medium. When the network optimization program is executed by a processor, the steps of any one of the above network optimization methods are implemented.

[0010] The present application provides a network optimization method, device, and storage medium. Compared with the related art, where using an intent network to optimize a communication network requires collecting intents, designing solutions, and then developing and delivering to support new intent types, which slows down the speed of intent type expansion and results in low efficiency of communication network optimization. In the present application, by obtaining user intents and using an intent engine to process the user intents to obtain a first intent configuration. Since the first intent configuration includes an intent optimization strategy corresponding to the new intent type, and the intent optimization strategy is predicted based on the initial network metrics of the current network, therefore, the first intent configuration conforms to the optimization strategy corresponding to the new intent type. Furthermore, the current network can be optimized in a timely manner according to the generated first intent configuration, thus improving the network optimization efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 It is a schematic flowchart of the first embodiment of the network optimization method of the present application;

[0012] Figure 2 It is a schematic diagram of the system architecture of the network optimization method of the present application;

[0013] Figure 3 It is a schematic diagram of the device structure of the hardware operating environment involved in the solution of the embodiment of the present application;

[0014] Figure 4 It is a schematic diagram of the execution process involved in the network optimization method of the present application;

[0015] Figure 5 It is a schematic diagram of the device structure involved in the network optimization method of the present application;

[0016] Figure 6 It is a schematic flowchart of the refinement of S10 in the first embodiment of the network optimization method of the present application;

[0017] Figure 7 It is a schematic flowchart of the refinement of S11 in the first embodiment of the network optimization method of the present application;

[0018] Figure 8 It is a schematic flowchart of the refinement of S13 in the first embodiment of the network optimization method of the present application;

[0019] Figure 9 It is a schematic flowchart of the second embodiment of the network optimization method of the present application;

[0020] Figure 10Schematic flowchart of the third embodiment involved in the network optimization method of the present application;

[0021] Figure 11 Schematic flowchart of the refinement of S110 in the third embodiment involved in the network optimization method of the present application;

[0022] Figure 12 Schematic flowchart of the fourth embodiment involved in the network optimization method of the present application;

[0023] Figure 13 Schematic flowchart of the model training process involved in the fourth embodiment of the network optimization method of the present application. Detailed implementation manners

[0024] It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0025] The embodiments of the present application provide a network optimization method. Figure 1 Schematic flowchart of the first embodiment of the network optimization method of the present application. Refer to Figure 1 The method includes the following steps:

[0026] Step S10: Obtain the user intention and process the user intention by using an intention engine to obtain a first intention configuration. Among them, the first intention configuration includes an intention optimization strategy corresponding to a new intention type, and the intention optimization strategy is predicted based on initial network metrics.

[0027] It should be noted that the network optimization method can be applied to a network optimization device, and the network optimization device belongs to a network optimization system, and this network optimization system belongs to network optimization equipment.

[0028] It should be noted that the execution subject of this method is a network optimization system. In this system, by receiving the user intention, identifying the user intention, collecting samples, and inputting the network metrics related to the user intention into the model for processing, the required intention configuration is obtained, and then the current network is optimized with the obtained intention configuration. Among them, when it is determined that the user intention is an intention type stored in the system, the intention configuration stored in the system can be directly applied to optimize the network. When it is determined that the user intention is a new intention type, the model predicts the network metrics related to the user intention, and combines with operation and maintenance experts and a simulation environment to quickly obtain the optimal intention configuration corresponding to the new intention type.

[0029] It should be noted that the user intention can be the network metrics that the user hopes to achieve. For example, the user intention can be: UE average delay < 100ms, UE average delay < 80ms, etc., which can be changed according to the user's needs and are not specifically limited.

[0030] It should be noted that the intent engine is used to process user intents and output intent configurations for optimizing the current network. The intent engine serves as the carrier for the general capabilities of intents, responsible for docking the intent inputs of business systems, driving the intent models on the AI platform to identify intents, generate configurations, perceive the real-time status of the network management system, record and feedback to operation and maintenance experts, and regularly and automatically optimize the intent implementation model in combination with the configuration experience of operation and maintenance experts.

[0031] It should be understood that the first intent configuration is an optimization strategy for optimizing the current network obtained by the intent engine after processing user intents. For example, when the user intent is that the average UE delay < 100ms, the generated first intent configuration can be "increase the single-batch message processing volume by 10", or "reduce the data message processing interval by 10ms". The first intent configuration can change according to the different current status of metrics, and the first intent configuration includes multiple output optimization strategies. Among these optimization strategies, the optimal value of the selection scheme is used to optimize the current network.

[0032] It should be understood that the initial network metrics are the metrics to be adjusted associated with the new intent type. For example, when the user intent is that the average UE delay < 100ms, the network metrics to be adjusted are "single-batch message processing volume" and "data message processing interval", then these two metrics are the initial network metrics, and the initial network metrics are predicted through a preset configuration model.

[0033] It should be understood that in the existing intent engines, intent configurations can only be generated for fixed user intents, while the first intent configuration involved in this application is the intent configuration output for the new intent type. Among them, the new intent type is the new metric requirement that the current network metrics need to meet. After predicting the user intent / new intent type through the preset configuration model to obtain the intent optimization strategy, the optimization strategy required for the new intent can be obtained more accurately. If the simulation result obtained by simulating the obtained optimization strategy does not reach the required network target, after analyzing the optimization strategy, the optimization strategy is adjusted, and the remaining relevant metrics are input into the model for processing until the required network optimization strategy is obtained.

[0034] Specifically, Figure 2 is a schematic diagram of the system architecture of the network optimization method of this application, as Figure 2 shown, in which the system includes: the interface where the operation and maintenance expert is located, the interface where the intent user is located, the intent engine, the network management, and the AI platform. The functions of each module in the system mainly include:

[0035] Intention-guaranteed UI: It is for the main interface of intention users' operations, mainly responsible for intention input, intention recognition and confirmation, recommended configuration confirmation, intention list presentation, and view of intention achievement status. Among them, intention users are generally low-threshold operator personnel without advanced operation and maintenance experience, responsible for issuing system operation and maintenance intentions in natural language to guarantee specific business scenarios.

[0036] The intention engine includes modules such as the expert UI, work order management, intention management, sample collection, intention model management, intention recognition, environment perception, and configuration generation. The main functions of each module are as follows:

[0037] Expert UI: The interface for operation and maintenance experts, where all intention tasks, as well as the configurations for intention modification, the achievement status of intentions, and real-time environment information such as relevant KPI indicators and alarms can be seen. The expert UI provides an entry for configuration tuning of intention generation for expert users, can provide suggestions for modifying the implementation configurations of intentions, and for adding or deleting relevant metric monitoring items. Among them, operation and maintenance experts are operation and maintenance personnel with advanced operation and maintenance knowledge, familiar with the functions of each module on UME, and are responsible for system status monitoring, alarm handling, network tuning, etc.

[0038] Work order management: Responsible for the tracking and closed-loop of work orders for intention addition and intention implementation optimization. When the intention achievement rate is insufficient, relevant alarms occur, or the intention user enters an unsupported intention type, the intention engine automatically dispatches a work order to the operation and maintenance expert. The operation and maintenance expert views the information of each module according to the work order information, troubleshoots problems, and provides a configuration modification strategy. After the expert feedbacks that the work order is completed, the work order management system will push the configuration to the intention user. After the intention user confirms it, it will be sent to the configuration management of the network management by the intention configuration generation module. Then the work order system continuously pays attention to the intention achievement rate. Until the intention is achieved, the modified configuration is passed to the sample collection module and marked as a valid modification. If the intention is still not achieved after the modification is issued, a work order will be dispatched to the expert for further optimization and passed to the sample collection module, marked as an invalid modification; until the intention is achieved, or the user actively deletes or modifies the intention, the work order is closed.

[0039] Specifically, the work order module is embedded in the network management system for model optimization, enabling users to unconsciously feedback data to the model for optimizing the model. Different from the conventional work order module, the work order process of the present invention will not only promote the formulation of solutions and effect simulation, but also track the closed-loop of the execution effect of the solution and promote sample recording.

[0040] Sample Collection: Responsible for collecting automatically generated configurations, experts' modification strategies, KPI metrics before and after configuration deployment, and the achievement of intentions. Different from the existing sample collection method (manual sample collection), it is converted to automatically collect sample information in standard format. This module collects configuration samples, metric samples, and intention samples required for training through automatic acquisition, enabling users to provide sample data without awareness and using it to optimize the model.

[0041] Intention Model Management: Responsible for managing the training and inference of intention-related models on the AI platform, mainly including text intention recognition models, configuration pre-trained LMs, and configuration generation models. Among them, the functions of each model are respectively used for:

[0042] 1. Configuration Pre-trained LM: Can generate multiple sets of configurations based on KPIs and partial configuration information.

[0043] 2. Preset Configuration Model: A model generated after the configuration pre-trained LM undergoes supervised fine-tuning. It can automatically generate multiple sets of configurations and recommended rankings based on KPIs and partial configuration information.

[0044] 3. Text Intention Recognition Model / Preset Intention Recognition Model: Recognizes the user's natural language into several key elements of the intention, including scope, KPI target, time, and operation type.

[0045] 4. Intention Scenario Clustering Model / Preset Clustering Model: Clusters intention optimization scenarios based on the differences before and after the KPI snapshot.

[0046] This module is different from the existing intention model management (only including intention recognition models). This module includes text recognition, configuration generation, and intention type recognition. In addition to driving model inference, it also drives model training, increasing the online training speed by more than 10 times through GPU, enabling the system to achieve model optimization at a daily granularity, and turning the intention engine that originally only had specific intention recognition capabilities into a core of general intention recognition, implementation, and optimization.

[0047] Intention Recognition: Responsible for receiving the user's input natural language and sending it to the intention model management module. The management module drives the AI platform model inference to identify the type and key elements of the intention and feedback it to the intention user for confirmation. If the identified type is not an existing supported type, an intention addition work order will be automatically dispatched after user confirmation.

[0048] Environmental Sensing: Responsible for collecting network status, providing configuration snapshots, alarm monitoring, KPI metric monitoring, and network element status monitoring. This module is different from the existing intent system (where network element reports are directly fed back to the intent assurance module and processed by the service module). Instead, it triggers work orders and delegates experts to automatically handle them through the intent engine without the intervention of the service module. It provides information for experts to make decisions and environmental data for the sample collection module.

[0049] Configuration Generation: Responsible for identifying network elements and cells affected by the intent, collecting environmental information, configuration inference, pushing configuration modification content, and configuration distribution. Based on the results of intent recognition, it queries the relevant network element and cell environmental information from the environmental sensing module. Then, it sends the KPI targets in the intent recognition results, the current relevant configuration situation provided by the environmental sensing module, and the current KPI metric situation to the AI platform through the intent model management for configuration generation. The results of configuration generation are recommended to the intent user for confirmation. After the user confirms, it is automatically distributed to the configuration management of the network management system.

[0050] This module is different from the existing built-in rule-based configuration generation. Operation and maintenance experts do not need to directly formulate configuration data. They only need to input associated KPIs as input prompts for the configuration LM. This module will automatically use the KPI targets, associated KPI metric situations, and the existing configuration snapshot as inputs to the AI model to generate multiple sets of complete configurations, identify configuration modification points, and preferentially select 3 sets (it can also be 5 sets, not specifically limited) of configuration solutions with the fewest configuration modification points. After simulation in the digital twin environment, it sorts and selects the configuration with the best KPI target achievement rate and pushes the modification points to the user. It will only be distributed to the live network after the user confirms.

[0051] The functions of the network management system are as follows:

[0052] 1. Alarm Management: Responsible for network alarm management and pushing alarms to relevant systems when KPIs are abnormal.

[0053] 2. Configuration Management: Responsible for managing, distributing, and synchronizing all configuration information of network devices.

[0054] 3. Performance Management: Provides regular monitoring and storage of all KPI metrics of network devices.

[0055] 4. Network Element Management: Responsible for network element information management, network element link establishment, model package management, etc.

[0056] The functions of the AI platform are as follows:

[0057] 1. Model Training: Based on the GPU, perform training and parameter tuning of the configuration pre-training LM, configuration generation model, and intent recognition model.

[0058] 2. Model reasoning: Based on GPU, call the existing model and return the reasoning result based on the input.

[0059] 3. Digital twins use digital simulation technology to deploy network configurations in a simulation environment. Through channel simulation, geographic simulation, physical simulation, etc., 3D algorithms are used for real-time calculation to obtain the simulated impact of network configurations on the network.

[0060] 4. GPU driver, the underlying hardware driver module, sends algorithm training and reasoning to the GPU deployed on PAAS for parallel computing, accelerating the training and reasoning process.

[0061] Specifically, the execution process of each module in the system is as follows:

[0062] 1. Intent user input intent: natural language intent input, mainly including scope, target KPI, time, and operation type.

[0063] 2. Intent type identification: Identify the type of intent. If it is an existing type, directly identify the five elements and generate configuration after user confirmation, and directly enter the effect evaluation stage. If it is not an existing type, start the new intent process after user confirmation.

[0064] 3. Dispatching work orders: If the environmental perception module finds that the intent is not achieved effectively, or an alarm occurs, or a new intent process is started, the work order management dispatches the intent work order, which includes the intent input and the trigger information provided by the environmental perception module, and the operation and maintenance expert accepts the order.

[0065] 4. View associated KPIs: Operation and maintenance experts view intent-related KPIs and set these KPIs to intent-related KPIs for configuring the input of the generation model and also as the key focus KPIs for environmental perception.

[0066] 5. The model generates multiple sets of recommended configurations: The KPIs added by the experts are weighted and input into the configuration pre-training LM or configuration generation model together with the existing configuration snapshot and other KPIs, and the model generates multiple sets of recommended configurations.

[0067] 6. Simulation and optimization: Use digital twin technology to simulate the generated configuration, sort and select according to the degree of achievement. If none of them can be achieved, report to the expert for intervention, modify the recommended configuration as the new configuration, and feedback that the configuration modification strategy has been formulated.

[0068] 7. Issue configuration: The configuration generation module pushes the configuration modification strategy to the user for confirmation. After the user confirms, the configuration is issued and the policy sample is recorded, including the configuration and KPI snapshots.

[0069] 8. Intent achievement monitoring: The work order system continuously monitors the intent achievement rate. If the intent is not achieved, a work order is dispatched again, and the policy sample is marked as an invalid modification sample. If the intent is achieved, it is marked as a valid modification sample and passed to the sample collection module.

[0070] 9. Work order closure: The work order is closed until the intent is achieved.

[0071] Among them, Figure 6 is a detailed process schematic diagram of S10 in the first embodiment. As Figure 6 shown, the step S10 of using the intent engine to process the user intent to obtain the first intent configuration includes:

[0072] Step S11, based on the intent engine, identify the intent type corresponding to the user intent.

[0073] It should be noted that in the intent engine, a preset intent recognition model is used to identify the user's natural language into several key elements of the intent, including scope, KPI target, time, and operation type. After obtaining these elements, the intent type corresponding to the user intent can be determined.

[0074] Step S12, if it is determined that the intent type corresponding to the user intent is a new intent type using the existing intent type, input the initial network metrics associated with the user intent into the preset configuration generation component.

[0075] It should be noted that when it is determined that the existing intent type does not include the currently received user intent and the intent type corresponding to the user intent is determined to be a new intent type, it is necessary to input the initial network metrics associated with the user into the preset configuration generation component for processing to obtain the required intent configuration.

[0076] It should be noted that the initial network metrics include KPI metrics to be input into the preset configuration model. When operating on the initial network metrics using the preset configuration model, it is also necessary to pay attention to task configuration information such as network scope, environmental information, and operation type required for creating the intent.

[0077] It should be understood that the preset configuration generation component is a component that processes the user intent and outputs the required intent configuration. Among them, the preset configuration generation component mainly includes a preset clustering model, a preset configuration model, a preset intent recognition model, etc. The user intent is clustered and predicted through each model to determine the intent configuration to be output.

[0078] Step S13, based on the preset configuration generation component and the initial network metrics, determine the first intent configuration.

[0079] It should be noted that before inputting the initial network metrics into the preset configuration model, it is also necessary to determine network metric elements such as the network scope, time, and operation type related to the network metrics. Inputting the initial network metrics into the preset configuration model for prediction processing can automatically generate multiple sets of intent configurations and the recommended sorting of these intent configurations. Among them, the preset configuration model is a model based on deep reinforcement learning.

[0080] Among them, Figure 7 is a detailed process schematic diagram of S11 in the first embodiment. As Figure 7 shown, the step S11 of identifying the intent type corresponding to the user intent based on the intent engine includes:

[0081] Step S111, perform clustering processing on the user intent to obtain intent clustering data, and input the intent clustering data into a preset intent recognition model.

[0082] It should be noted that the way to perform clustering processing on the user intent can be to cluster the user intent with the existing intents, which can classify the user intent in advance to make it different from the existing intent types.

[0083] It should be noted that the intent clustering data can be the data obtained after clustering the user intent. In the intent clustering data, it also includes new intent types, user intents of existing intent types, and outliers that do not belong to any cluster, strengthening the discrimination of user intents.

[0084] Step S112, based on the preset intent recognition model, identify the intent clustering data to determine the intent type corresponding to the user intent.

[0085] It should be noted that the preset intent recognition model can be a text intent recognition model, which recognizes the natural language of the user intent into several key elements of the intent, so as to determine the intent type to which the user intent belongs.

[0086] Among them, the preset configuration generation component includes a preset configuration model; Figure 8 is a detailed process schematic diagram of S13 in the first embodiment. As Figure 8 shown, the step S13 of determining the first intent configuration based on the preset configuration generation component and the initial network metrics includes:

[0087] Step S131, input the initial network metrics into the preset configuration model.

[0088] Step S132, based on the preset configuration model, perform prediction processing on the initial network metrics to obtain the first intent configuration.

[0089] It should be noted that when using the preset configuration model to predict and process the initial network metrics, it is also necessary to determine the intended optimization scenario. Among them, the intended optimization scenario can be various network metrics that need to be optimized. For example, when the metric to be optimized is network latency, the intended optimization scenario is Scenario 1; when the metric to be optimized is the number of data packets transmitted per unit time, the intended optimization scenario is Scenario 2.

[0090] It should be noted that the prediction process is the same as that of the input parameter prediction process of the deep learning-based model in the related technology. When determining the intended optimization scenario and the initial network metrics corresponding to the new intention type, the preset configuration model performs prediction processing on the intended optimization scenario and the initial network metrics corresponding to the new intention type to obtain the first intention configuration.

[0091] Step S20: Optimize the current network through the first intention configuration so that the target network metrics of the current network reach the intention target corresponding to the user intention.

[0092] It should be noted that the target network metrics can be the network metrics reached by the current network metrics. For example, the target network metrics can be that the network latency is less than 50 ms. When the intention target is that the network latency is less than 50 ms, the target network metrics are that the network latency is equal to 40 ms. At this time, it can be determined that the target network metrics of the current network reach the intention target corresponding to the user intention.

[0093] Specifically, Figure 4 is a schematic diagram of the execution process involved in this application. As Figure 4 shown, this application first parses the user input intention through semantic recognition, identifies the intention type of the user intention. Among them, the user intention also includes multiple intention target names and target values. The intention target is all KPIs supported by the system. Using the machine learning classification algorithm, the intention type is identified. First, it is judged whether it is an existing intention type. If the intention type input by the user is in the existing intention types, the configuration generation can be directly performed. If it is not in the currently supported existing intention types, the intention configuration corresponding to the new intention type is generated through the configuration generation model.

[0094] After determining that the user intent is a new intent, the user's input intent target and the current target value are sent to the intent engine. The intent engine dispatches a work order, and then through the corresponding network management operation, the indicator is checked according to the target. According to the KPI associated with the user intent, the initial network target is determined, and the initial network target is input into the previously trained configuration generation model for configuration generation. Then, through simulation selection, the optimal intent configuration is selected and transmitted to the client for confirmation. After receiving the issuance instruction from the user interface, it is finally issued to the existing network and feedback is fed back that the work order has been completed; according to the achievement of the intent target after issuance, the sample data is marked as a valid modification, and the invalid modification sample or the valid modification sample is used for model training.

[0095] Then, after collecting sample data, new samples are regularly summarized and the model is tuned. First, a clustering algorithm is used to cluster the differences before and after the optimization of the KPI to identify whether the sample is a new category of sample data. Through outlier detection, samples that do not belong to any known aggregation are identified, and these outliers are classified into a new category. Before the sample data is used for fine-tuning training, the proportion of high-quality samples is increased and the proportion of harmful samples is reduced through sample similarity calculation and fine-tuning data set optimization to obtain the optimized sample set. Finally, incremental supervised fine-tuning training is performed on the optimized sample set through incremental SFT (Supervised Fine-Tun-ing) to fine-tune the existing model weights so that it can generate more accurate recommendation results in the future, and the fine-tuned configuration generation model is used for the configuration generation of the next new intent type.

[0096] The next time the same type of intent is issued, the intent engine will automatically obtain multiple sets of model recommendation values ​​and rankings, and then evaluate and select the appropriate one to issue directly, or feed the result back to the intent engine after modification. The intent engine records the changes and uses the output intent configuration as sample data to continue training.

[0097] The existing system has different requirements for network performance indicators when encountering different network conditions. When local residents have different habits of using the network, or when encountering intent types that the system does not support, it is necessary to collect manual samples, identify intent, implement algorithm development, and release versions to solve these differentiation problems.

[0098] The present application provides a network optimization method, device, and storage medium. Compared with the related art, where using an intent network to optimize a communication network requires support for new intent types from intent collection, to solution design, and then to R & D delivery, which slows down the speed of intent type expansion and results in low efficiency of communication network optimization, in the present application, by obtaining user intents and using an intent engine to process the user intents to obtain a first intent configuration. Since the first intent configuration includes intent optimization strategies corresponding to new intent types, and the intent optimization strategies are obtained by predicting initial network metrics associated with the new intent types based on a preset configuration model, therefore, the first intent configuration is more in line with the optimization strategies corresponding to the new intent types. Furthermore, corresponding intent configurations can be generated to optimize the current network in a timely manner, thereby improving the network optimization efficiency.

[0099] Further, based on the first embodiment of the present application, another embodiment of the present application is provided. Figure 9 It is a schematic flowchart of the second embodiment, as Figure 9 shown. In this embodiment, after the step S111 of clustering the user intents to obtain intent clustering data, the following is further included:

[0100] Step S1110: Send the intent clustering data to the user terminal and receive the new intent types obtained by naming the intent clustering data by the user terminal at preset time intervals.

[0101] It should be noted that, in addition to using a language model for processing, the method for determining new intents in the present application can also be processed through the user terminal. The specific execution process is as follows: After clustering the sample vector space, if the user sets it to automatic classification, there is no need for manual access, and the language model is directly used to automatically generate category names for it; if the user sets it to require user confirmation, the new categories generated in each clustering will be pushed to the user, and the user can manually name the new categories regularly.

[0102] In this embodiment, by sending the intent clustering data to the user terminal and allowing the user to name the intent clustering data, the new intent types are made more in line with user needs, and through the processing of the user terminal, it is also easier for the model to process the input intent data.

[0103] Further, based on the first embodiment and the second embodiment of the present application, another embodiment of the present application is provided. Figure 10 It is a schematic flowchart of the third embodiment, as Figure 10 shown. In this embodiment, the first intent configuration further includes a plurality of parameter intent configurations; after the step S10 of using the intent engine to process the user intents to obtain the first intent configuration, the following is further included:

[0104] Step S110: Based on the first intention configuration, perform real-time calculation on the current network to obtain multiple network simulation results.

[0105] It should be noted that the parameter intention configuration can be a parameter optimization strategy for the network. For example, the parameter intention configuration can be that after adjusting parameter A, the current network can reach the intention target.

[0106] It should be noted that after obtaining the first intention configuration, deploy the current network in the simulation environment and perform real-time calculation on the current network with the first intention configuration, so as to obtain multiple network simulation results. Among them, one network simulation result corresponds to the result obtained after implementing with a certain parameter intention configuration.

[0107] Step S120: Send multiple network simulation results to the user terminal. When receiving the confirmation and distribution instruction fed back by the user terminal based on the network simulation results, determine to distribute the parameter intention configuration corresponding to the optimal value among the multiple network simulation results to the current network.

[0108] It should be noted that after obtaining the network simulation results, they need to be sent to the user terminal for confirmation. When the user confirms that the network simulation results meet the expected effect, the user will send a confirmation and distribution instruction through the user terminal, so as to select the parameter intention configuration corresponding to the optimal value of the network simulation results to the current network, thereby optimizing the current network.

[0109] Step S130: When receiving the rejection and distribution instruction fed back by the user terminal based on the network simulation results, determine to feed back the first intention configuration to the operation and maintenance interface where the operation and maintenance expert is located, so that the operation and maintenance expert can modify the first intention configuration based on the network simulation results.

[0110] It should be noted that when receiving the rejection and distribution instruction fed back by the user terminal based on the network simulation results, it means that the network simulation results do not meet the expected effect or the network indicators do not reach the intention target. At this time, the first intention configuration needs to be fed back to the operation and maintenance interface where the operation and maintenance expert is located, and after the operation and maintenance expert modifies / improves the first intention configuration, receive the adjusted intention configuration sent by the operation and maintenance expert.

[0111] Among them, Figure 11 is a schematic diagram of the refined process of S110 in the third embodiment. As Figure 11 shown, the step S110 of performing real-time calculation on the current network based on the first intention configuration to obtain multiple network simulation results includes:

[0112] Step S1101, determining the number of parameter modifications of each of the intended parameter configurations, wherein the intended parameter configurations are arranged in order from small to large based on the number of parameter modifications.

[0113] It should be noted that the number of parameter modifications corresponds to the number of parameter pairs that need to be modified for each parameter intention configuration. The number of parameter modifications can be 1, 2, 3, etc., and there is no specific limitation. The fewer parameters that need to be modified for the parameter intention configuration, the higher the ranking.

[0114] Step S1102: Based on the number of parameter modifications, select a preset number of parameter intention configurations in the first intention configuration before sorting.

[0115] It should be noted that in the process of obtaining the network simulation results, the parameter intention configuration with the front of the first intention configuration is selected, and the calculation is performed according to the recommended ranking of the parameter intention configuration. The parameter intention configuration with the lower ranking is considered to be the intention configuration with poor calculation effect and is not considered. The preset number can be 3, 5, 7, etc., and there is no specific limitation.

[0116] Step S1103, performing real-time calculations on the current network one by one with the parameter intention configuration to obtain multiple network simulation results.

[0117] It should be noted that the current network is calculated in real time according to the order of parameter intention configuration to obtain various network simulation results.

[0118] In this embodiment, before the first intention configuration is sent to the current network and put into use, the first intention configuration is calculated in real time on the network arranged in the simulation environment, so as to select the optimal intention configuration output by the preset configuration model according to the network simulation results.

[0119] Furthermore, based on the first embodiment, the second embodiment and the third embodiment of the present application, another embodiment of the present application is provided. Figure 12 is a flow chart of the fourth embodiment, as shown in Figure 12 As shown, in this embodiment, after the step S20 of optimizing the current network through the first intention configuration so that the target network indicator of the current network reaches the intention target corresponding to the user intention, it also includes:

[0120] Step S30: If the target network indicator does not reach the intention target corresponding to the user intention, the first intention configuration is marked as an invalid modification sample.

[0121] It should be noted that when the first intent configuration is applied to the current network, if the target network indicator does not reach the intent target corresponding to the user intent, it means that the currently generated first intent configuration is invalid and is marked as an invalid modification sample.

[0122] Step S40, if the target network metric reaches the intent target corresponding to the user intent, mark the first intent configuration as a valid modification sample.

[0123] It should be noted that after the first intent configuration is applied to the current network and the target network metric reaches the intent target corresponding to the user intent, it indicates that the currently generated first intent configuration is valid and is marked as a valid modification sample.

[0124] Step S50, input the invalid modification sample or the valid modification sample into a preset configuration generation component for iterative training of the preset configuration model.

[0125] It should be noted that after determining the invalid modification sample or the valid modification sample, these two types of samples are input into the preset configuration generation component to fine-tune and iteratively train the preset clustering model and the preset configuration model.

[0126] Among them, Figure 13 is a schematic diagram of the model training process. As Figure 13 shown, after step S50 of inputting the invalid modification sample or the valid modification sample into the preset configuration generation component, the following steps are further included:

[0127] Step S510, use the invalid modification sample or the valid modification sample as new sample data, cluster the new sample data and the existing sample data to obtain clustering data;

[0128] It should be noted that the invalid modification sample or the valid modification sample is used as sample data and input into the model. At the same time, the new sample data and the existing sample data are clustered. After determining the clustering data, the preset configuration model is fine-tuned using the clustering data to obtain a fine-tuned clustering model.

[0129] It should be noted that the methods for pre-training and fine-tuning training of the preset configuration model can be:

[0130] 1. Obtain a large amount of configuration and KPI snapshot information: Obtain a large amount of exported XML configuration information and the KPI value snapshots at that time from the field or laboratory for unsupervised training. Among them, the KPI can be used as an XML node and strung to the end of the configuration XML;

[0131] 2. Generate a configuration pre-trained LM: Based on the GPU, perform unsupervised training on the configuration and KPI snapshots after random MASK;

[0132] 3. The sample data includes the MASKed configuration matrix and KPI matrix, which are used as the input of the neural network. The Transformer or LSTM can be used in the middle layer, and the output is the configuration matrix;

[0133] 4. Enable the model to generate complete configuration parameters based on partial configuration and KPI information in the input.

[0134] 5. Regularly summarize new samples: Regularly collect samples from the sample collection module for fine-tuning training.

[0135] 6. Sample aggregation: Extract the snapshot differences of KPIs before and after for the new and existing samples, and perform clustering.

[0136] 7. Identify new categories: Through outlier detection, identify samples that do not belong to any known aggregation, and classify these outliers into a new category. If the system is set to automatic addition mode, use the LLM (Large Language Model, used to generate and understand natural language) to summarize and name them, otherwise regularly push them to experts for naming or adjustment of classification.

[0137] In step S520, perform vectorization conversion on the clustering data and the old sample data to obtain clustering sample vectors and old sample vectors.

[0138] It should be noted that after performing doc2vec (an algorithm that converts document samples into fixed-length vectors) vectorization conversion on the clustering data and the old sample data, clustering sample vectors and old sample vectors are obtained. The purpose of the vectorization conversion is to calculate the cosine distance between each vector, thereby calculating the similarity of the two types of data.

[0139] In step S530, calculate the similarity between the clustering sample vectors and the old sample vectors, and based on the similarity, determine the optimized sample dataset.

[0140] It should be noted that the method for determining the optimized sample dataset is: traverse the category to which each new sample belongs, delete the sample with the closest cosine distance to the invalid sample, and directly add the valid samples to the fine-tuning dataset to increase the proportion of high-quality samples and reduce the proportion of harmful samples.

[0141] In step S540, fine-tune the preset configuration model through the optimized sample dataset to obtain the fine-tuned preset configuration model.

[0142] It should be noted that perform incremental SFT (Supervised Fine-Tuning, using labeled sample data to train and adjust model parameters, and back up and synchronize the changed sample set) on the optimized sample dataset to fine-tune the weights of the existing preset configuration model, so that it can generate more accurate recommendation results subsequently.

[0143] It should be noted that the training frequency of the model can be once a day or once every half day, and there is no specific limit.

[0144] In this embodiment, by performing sample statistics on the intent configurations output by the preset configuration model and respectively labeling them as invalid modification samples and valid modification samples, and then, through these sample data, iteratively training the preset configuration model, the model can output the optimal intent configuration that better conforms to the user's intent.

[0145] The scenarios used in this application are as follows:

[0146] 1. Differences in local network requirements: For example, the concert network pays more attention to improving user capacity, while the conference room network pays more attention to the quality of user communication.

[0147] 2. Differences in user usage habits: Users in region A prefer to visit text-based websites, while users in region B prefer to watch short videos, and their bandwidth requirements for the network are different.

[0148] 3. Unsupported intent types: Users use KPI targets created by the local site, which are not within the scope of the system-built-in intent types.

[0149] If these scenarios are encountered, the marketing team needs to submit requirements to the R & D center. Then, the planning team determines the local site specificity and the intent new version release strategy. Then, driven by the requirements, the R & D team collects intent sample data, intent type definitions, intent recognition algorithm research and development, intent implementation algorithm research and development, testing and delivery, version release, field verification, etc. The entire process involves the entire process of system delivery and all roles. Generally, the cycle is two or three months, with huge human consumption, and many intent types may be applicable to fewer local sites, and the R & D cost is not proportional to the usage frequency.

[0150] In scenario 1, the original solution might be to solve it through a differential version strategy, resulting in a very complex version strategy, or to provide the full set of requirements for all local sites for the local side to select a solution, resulting in a very complex function that is difficult to maintain.

[0151] In scenario 2, it may be necessary to collect the user habits in different regions and then develop different intent implementation strategies for different user habits. The local differences among users are very common and numerous, and it is not feasible to rely entirely on R & D to collect and implement.

[0152] In scenario 3, only the new requirement R & D process can be followed, which may lead to the R & D process not being able to keep up with the addition of new intent requirements.

[0153] By means of the above-mentioned intent addition process and model training process, when the above scenarios are detected, the field intent system can be automatically triggered to optimize the process. Through system work order operation, the intervention of field operation and maintenance experts, simulation, sample collection, and model training, the entire intent addition and optimization process can be completed, reducing the R & D cycle of several months to one or two weeks, and the optimization strategy is more suitable for the local network conditions and user habits.

[0154] Examples of application scenarios involved in this application are as follows:

[0155] Application Example 1:

[0156] The user hopes to reduce the average UE delay in Shanghai New International Expo Center, so he says to the system: "Shanghai New International Expo Center, average UE delay < 100ms."

[0157] The intent recognition module identifies that the KPI name is "average UE delay", the KPI target value is less than 100ms, and the scope is "Shanghai New International Expo Center". Combining the current status of 110ms of this KPI and the configured current value, it conducts intent classification and finds that the matching degree with the existing intent types is relatively low. Therefore, the system prompts the user: "This intent type is not currently supported. Do you need to enable the new intent type mode?"

[0158] The user selects "Yes", and the system automatically dispatches a work order to the operation and maintenance expert. At the same time, the intent is sent down and the status shows that the work order has been dispatched.

[0159] Based on the user's description, the expert views the associated KPIs according to experience and marks these viewed KPIs as the key KPIs to be focused on for this type of intent. According to the expert's marks, the system automatically generates multiple sets of configuration parameters through the large model. The first set recommends "increasing the single-batch message processing volume by 10", and the second set recommends "reducing the data message processing interval by 10ms", and presents them to the expert.

[0160] Through simulation, the simulation effect of the second set of configurations is better. Therefore, "reducing the data message processing interval by 10ms" is configured in the MOC, and it is reported that the configuration strategy has been formulated.

[0161] The second set of recommended configurations is sent down to the network elements within the specified intent range.

[0162] After 5 minutes, the intent is achieved, the work order is automatically closed, and the recommended configuration is marked as an effective modification. The snapshots of the configuration before and after the modification and the snapshots of the KPI before and after are recorded as effective modification samples.

[0163] That night, the system automatically aggregated all samples collected on the same day and found that "at the Shanghai New International Expo Center, the average UE latency < 100 ms" was a new type, and then pushed it to the operation and maintenance experts for naming. After the operation and maintenance experts confirmed it, they named it "Optimization of Average UE Latency".

[0164] According to the result of the intention implementation, after adding the second set of recommended configurations to the effective sample pool, the system performed fine-tuning training and saved the configuration generation model after fine-tuning.

[0165] The next day, the user input "Zhangjiang Hi-Tech Metro Station, average UE latency < 80 ms". The system judged it as an intention of the "Optimization of Average UE Latency" category based on clustering, and judged the current status of the KPI as 110 ms. Then, a set of parameters closer to the original second set of configuration plans was generated, with the "data packet processing interval" reduced by 30 ms, and it was sent for user confirmation and then issued.

[0166] Five minutes later, the intention was achieved, and the system automatically marked this configuration as a valid modification and recorded the sample.

[0167] Application Example 2:

[0168] In step (11) of Application Example 1, the intention achievement rate was still not good five minutes later.

[0169] The system automatically marked this configuration as an invalid modification sample and dispatched a work order, and the expert received the work order;

[0170] After investigation, it was found that the high latency in the network elements around Zhangjiang Hi-Tech Metro Station was not caused by the "data packet processing interval", but by the high KPI "downlink utilization rate". Therefore, the weight of the associated KPI "downlink utilization rate" was increased for this intention type and input into the model for re-configuration generation.

[0171] In the newly generated configuration, the third set of configuration plans was to increase the "cell bandwidth" within the network element.

[0172] After selecting the third set of plans based on the simulation results, the feedback modification strategy was completed.

[0173] After user confirmation, the configuration was issued.

[0174] Five minutes later, the intention was achieved and recorded as a valid modification sample.

[0175] Application Example 3:

[0176] In step (5) of Application Example 2, the expert felt that none of the existing plans met the requirements, so a new plan "increase the message cache count by 100" was manually created, and the feedback modification strategy was completed.

[0177] After user confirmation, the configuration was issued;

[0178] After 5 minutes, the intention is achieved and recorded as a valid modified sample.

[0179] Application Example 4:

[0180] In the scenario of Application Example 2, when the system was performing clustering on the same night, the user found that in this scenario, "Zhangjiang Hi-Tech Metro Station, UE average delay < 80ms". Combining with the current status of the KPI snapshot, the system determined that the intention of "UE average delay optimization" was not appropriate according to the clustering, but should belong to the intention of "bandwidth utilization optimization".

[0181] The expert modifies the sample attribution and optimizes the clustering model.

[0182] The next time the downlink utilization rate is insufficient, it will be more likely to be recognized as the "bandwidth utilization optimization" type, and the KPIs associated with this type will be concerned.

[0183] Refer to Figure 3 , Figure 3 is a schematic diagram of the device structure of the hardware operating environment involved in the solution of the embodiment of the present application.

[0184] As Figure 3 shown, the network optimization device may include: a processor 1001, a memory 1005, and a communication bus 1002. The communication bus 1002 is used to realize the connection and communication between the processor 1001 and the memory 1005.

[0185] Optionally, the network optimization device may further include a user interface, a network interface, a camera, an RF (Radio Frequency) circuit, sensors, a WiFi module, etc. The user interface may include a display screen (Display) and an input sub-module such as a keyboard (Keyboard). Optionally, the user interface may further include a standard wired interface and a wireless interface. The network interface may include a standard wired interface and a wireless interface (such as a WI-FI interface).

[0186] Those skilled in the art can understand that Figure 3 the network optimization device structure shown in

[0187] does not constitute a limitation to the network optimization device, and may include more or fewer components than shown, or combine some components, or have different component arrangements. Figure 3As shown in the figure, the memory 1005 as a storage medium may include an operating system, a network communication module, and a network optimization program. The operating system is a program that manages and controls the hardware and software resources of the network optimization device and supports the operation of the network optimization program and other software and / or programs. The network communication module is used to implement communication between components within the memory 1005 and communication between other hardware and software in the network optimization system.

[0188] In Figure 3 the network optimization device shown in the figure, the processor 1001 is used to execute the network optimization program stored in the memory 1005 to implement the steps of the network optimization method described in any one of the above.

[0189] The specific implementation manner of the network optimization device in this application is basically the same as that of each embodiment of the above network optimization method, and will not be elaborated here.

[0190] This application also provides a network optimization device. Referring to Figure 5 the figure, the network optimization device includes:

[0191] An acquisition module, configured to acquire a user intention and process the user intention by using an intention engine to obtain a first intention configuration, where the first intention configuration includes an intention optimization strategy corresponding to a new intention type, and the intention optimization strategy is predicted based on initial network metrics;

[0192] An optimization module, configured to optimize the current network through the first intention configuration so that the target network metrics of the current network reach the intention target corresponding to the user intention.

[0193] In a possible implementation manner of this application, the acquisition module includes:

[0194] An identification unit, configured to identify the intention type corresponding to the user intention based on the intention engine;

[0195] A first determination unit, configured to determine that the intention type corresponding to the user intention is a new intention type by using an existing intention type, and then input the initial network metrics associated with the user intention into a preset configuration generation component;

[0196] A processing unit, configured to determine a first intention configuration based on the preset configuration generation component and the initial network metrics ，

[0197] In a possible implementation manner of this application, the processing unit includes:

[0198] An input subunit, configured to input the initial network metrics into a preset configuration model;

[0199] ​A processing subunit, configured to perform prediction processing on the initial network metrics based on the preset configuration model to obtain a first intent configuration.

[0200] In a possible implementation manner of the present application, the device further includes:

[0201] A calculation module, configured to perform real-time calculation on the current network based on the first intent configuration to obtain multiple network simulation results;

[0202] A distribution module, configured to send the multiple network simulation results to the user side, and determine to distribute the parameter intent configuration corresponding to the optimal value among the multiple network simulation results to the current network when receiving a confirmation distribution instruction fed back by the user side based on the network simulation results;

[0203] A receiving module, configured to determine to feed back the first intent configuration to the operation and maintenance interface where the operation and maintenance expert is located for the operation and maintenance expert to modify the first intent configuration based on the network simulation results when receiving a rejection distribution instruction fed back by the user side based on the network simulation results.

[0204] In a possible implementation manner of the present application, the calculation module includes:

[0205] A second determination unit, configured to determine the number of parameter modifications for each parameter intent configuration, where the parameter intent configurations are arranged in ascending order based on the number of parameter modifications;

[0206] A selection unit, configured to select the parameter intent configurations with the top preset number in the first intent configuration based on the number of parameter modifications;

[0207] A real-time calculation unit, configured to perform real-time calculation on the current network one by one with the parameter intent configurations to obtain multiple network simulation results.

[0208] In a possible implementation manner of the present application, the recognition unit includes:

[0209] A third processing subunit, configured to perform clustering processing on the user intent to obtain intent clustering data, and input the intent clustering data into a preset intent recognition model;

[0210] A recognition subunit, configured to recognize the intent clustering data based on the preset intent recognition model to determine the intent type corresponding to the user intent.

[0211] In a possible implementation manner of the present application, the device further includes:

[0212] A first marking module, configured to mark the first intention configuration as an invalid modification sample if the target network metric fails to reach the intention target corresponding to the user intention;

[0213] A second marking module, configured to mark the first intention configuration as a valid modification sample if the target network metric reaches the intention target corresponding to the user intention;

[0214] An input module, configured to input the invalid modification sample or the valid modification sample into a preset configuration generation component for iteratively training the preset configuration model.

[0215] In a possible implementation manner of this application, the device further includes:

[0216] A training module, configured to use the invalid modification sample or the valid modification sample as new sample data, cluster the new sample data and existing sample data to obtain clustering data;

[0217] A conversion module, configured to perform vectorization conversion on the clustering data and old sample data to obtain a clustering sample vector and an old sample vector;

[0218] A determination module, configured to calculate the similarity between the clustering sample vector and the old sample vector, and determine an optimized sample data set based on the similarity;

[0219] A fine-tuning module, configured to fine-tune the preset configuration model through the optimized sample data set to obtain a fine-tuned preset configuration model.

[0220] It should be noted that in this article, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or system. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of another identical element in the process, method, article or system including that element.

[0221] The serial numbers of the embodiments of this application above are only for description and do not represent the advantages and disadvantages of the embodiments.

[0222] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-described example methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. This computer software product is stored in a storage medium as described above (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions for causing a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in various embodiments of the present application.

[0223] The above are only the preferred embodiments of the present application, and do not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present application, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present application.

Claims

1. A network optimization method, characterized in that, Including: Obtain the user intention, and process the user intention using an intention engine to obtain a first intention configuration, where the first intention configuration includes an intention optimization strategy corresponding to a new intention type, and the intention optimization strategy is predicted based on initial network metrics; Optimize the current network through the first intention configuration so that the target network metrics of the current network reach the intention target corresponding to the user intention.

2. The network optimization method according to claim 1, characterized in that, The step of processing the user intention using the intention engine to obtain a first intention configuration includes: Based on the intention engine, identify the intention type corresponding to the user intention; If it is determined that the intention type corresponding to the user intention is a new intention type using the existing intention type, input the initial network metrics associated with the user intention into a preset configuration generation component; Generate components and the initial network metrics based on the preset configuration ， Determine the first intent configuration.

3. The network optimization method according to claim 2, characterized in that, The preset configuration generation component includes a preset configuration model; the step of determining a first intention configuration based on the preset configuration generation component and the initial network metrics includes: Input the initial network metrics into the preset configuration model; Based on the preset configuration model, perform prediction processing on the initial network metrics to obtain a first intention configuration.

4. The network optimization method according to claim 1, characterized in that, The first intention configuration includes multiple parameter intention configurations; after the step of processing the user intention using the intention engine to obtain a first intention configuration, it further includes: Perform real-time calculation on the current network based on the first intention configuration to obtain multiple network simulation results; Send the multiple network simulation results to the user side, and when receiving a confirmation and issuance instruction from the user side based on the network simulation results, determine to issue the parameter intention configuration corresponding to the optimal value among the multiple network simulation results to the current network; When receiving a rejection and issuance instruction from the user side based on the network simulation results, determine to feedback the first intention configuration to the operation and maintenance interface where the operation and maintenance expert is located for the operation and maintenance expert to modify the first intention configuration based on the network simulation results.

5. The network optimization method according to claim 4, characterized in that, The step of performing real-time calculation on the current network based on the first intention configuration to obtain multiple network simulation results includes: Determine the number of parameter modifications for each parameter intention configuration, where the parameter intention configurations are arranged in ascending order based on the number of parameter modifications; Based on the number of parameter modifications, select the first preset number of parameter intention configurations in the first intention configuration; Perform real-time calculation on the current network one by one with the parameter intention configurations to obtain multiple network simulation results.

6. The network optimization method according to claim 2, characterized in that, The step of identifying the intention type corresponding to the user intention based on the intention engine includes: Perform clustering processing on the user intention to obtain intention clustering data, and input the intention clustering data into a preset intention recognition model; Based on the preset intention recognition model, identify the intention clustering data to determine the intention type corresponding to the user intention; Or, After the step of performing clustering processing on the user intention to obtain intention clustering data, it further includes: Send the intention clustering data to the client and receive the new intention types obtained by the client naming the intention clustering data every preset time period.

7. The network optimization method according to claim 1, characterized in that, After the step of optimizing the current network through the first intention configuration so that the target network metrics of the current network reach the intention target corresponding to the user intention, the method further includes: If the target network metrics do not reach the intention target corresponding to the user intention, mark the first intention configuration as an invalid modification sample; If the target network metrics reach the intention target corresponding to the user intention, mark the first intention configuration as a valid modification sample; Input the invalid modification sample or the valid modification sample into a preset configuration generation component for iterative training of the preset configuration model.

8. The network optimization method according to claim 7, characterized in that, After the step of inputting the invalid modification sample or the valid modification sample into the preset configuration generation component, the method further includes: Use the invalid modification sample or the valid modification sample as new sample data, cluster the new sample data and the existing sample data to obtain clustering data; Perform vectorization conversion on the clustering data and the old sample data to obtain a clustering sample vector and an old sample vector; Calculate the similarity between the clustering sample vector and the old sample vector, and based on the similarity, determine the optimized sample data set; Fine-tune the preset configuration model through the optimized sample data set to obtain a fine-tuned preset configuration model.

9. A network optimization device, characterized in that, The device includes: a memory, a processor, and a network optimization program stored on the memory and executable on the processor, the network optimization program being configured to implement the steps of the network optimization method according to any one of claims 1 to 8.

10. A computer storage medium, characterized in that, A network optimization program is stored on the computer storage medium, and when the network optimization program is executed by a processor, it implements the steps of the network optimization method according to any one of claims 1 to 8.

Images