Parameter configuration method and device and electronic equipment

Abstract

The invention discloses a parameter configuration method and device and electronic equipment. The method comprises the steps that configuration requirement information of a target object for network slices is converted into a slice parameter set, and the slice parameter set comprises slice parameters of multiple dimensions; the slice parameter set is verified through an association rule base, a verification result is obtained, and the association rule base is used for quantitatively representing association information and conflict information between slice parameters of different dimensions; under the condition that the verification result indicates that the verification is successful, determining a parameter configuration strategy corresponding to the slice parameter set; and configuring the network slices by adopting the parameter configuration strategy. According to the method and the device, the technical problem that the service quality after slice deployment does not reach the standard due to conflicts among parameters in a configuration strategy because only the compliance of the parameters is considered when the network slice parameters are configured in the prior art is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electronic device parameter configuration, in particular to a parameter configuration method and device and electronic device. BACKGROUND

[0002] With the rapid development and wide application of 5G network technology, network slicing as a key technology to realize "one network for multiple uses" and personalized services, the accuracy and automation level of its parameter configuration directly affect the utilization efficiency of network resources and the service experience of users. However, the related art only considers the compliance of the parameters themselves when configuring the network slicing parameters, resulting in conflicts between parameters in the configuration strategy, which leads to substandard service quality after slice deployment.

[0003] To address the above problems, no effective solutions have been proposed so far. SUMMARY

[0004] The embodiments of the present application provide a parameter configuration method, device and electronic device to at least solve the technical problem that the related art only considers the compliance of the parameters themselves when configuring the network slicing parameters, resulting in conflicts between parameters in the configuration strategy, which leads to substandard service quality after slice deployment.

[0005] According to an aspect of an embodiment of the present application, a parameter configuration method is provided, including: converting configuration requirement information of a target object for a network slice into a set of slice parameters, wherein the set of slice parameters includes slice parameters of multiple dimensions; using an association rule library to verify the set of slice parameters to obtain a verification result, wherein the association rule library is used to quantitatively represent association information and conflict information between slice parameters of different dimensions; in the case that the verification result indicates that the verification is successful, determining a parameter configuration strategy corresponding to the set of slice parameters; and using the parameter configuration strategy to configure the network slice.

[0006] In some embodiments of the present application, the association rule library is determined by: obtaining historical parameter configuration data corresponding to a historical network slice; determining a knowledge graph corresponding to the historical parameter configuration data, wherein the nodes in the knowledge graph are historical slice parameters in the historical parameter configuration data, and the edges of the knowledge graph are the association strengths between the nodes; determining attention weights between the nodes according to the vector representation of each node, wherein the attention weights are used to quantitatively represent the importance of the association rules corresponding to two nodes; and determining the association rule library according to the attention weights.

[0007] In some embodiments of the present application, the knowledge graph corresponding to the historical parameter configuration data is determined by extracting historical slice parameters of multiple dimensions from the historical parameter configuration data and taking the historical slice parameters as nodes of the knowledge graph, determining the correlation strength between any two nodes and taking the correlation strength as the weight of the edge between any two nodes, and removing edges with a correlation strength less than a preset correlation strength from the knowledge graph to obtain the knowledge graph.

[0008] In some embodiments of the present application, the correlation strength is determined by determining the co-occurrence frequency corresponding to the first node and the second node in the knowledge graph, wherein the co-occurrence frequency includes the proportion of the number of times that the first node and the second node appear simultaneously to the total configuration times of the historical parameter configuration data, obtaining a preset weight corresponding to the first node and the second node, wherein the preset weight is used to reflect the association importance between the first node and the second node determined based on the business characteristics, and performing weighted summation on the co-occurrence frequency and the preset weight to obtain the correlation strength.

[0009] In some embodiments of the present application, the attention weight between nodes is determined according to the vector representation of each node, including processing the knowledge graph by using a target model to obtain the vector representation of each node, wherein the target model determines the vector representation of each node by aggregating the neighbor information of the nodes in the knowledge graph, and determining the attention weight between nodes according to the vector representation of each node.

[0010] In some embodiments of the present application, the target model is obtained by training in the following manner: obtaining training data, wherein the training data includes the correlation strength of a pair of training nodes in the training knowledge graph; processing the training data by using an initial target model to obtain the predicted vector representation of each training node, and determining the predicted attention weight of the pair of training nodes according to the predicted vector representation; determining the loss corresponding to the predicted attention weight and the correlation strength; adjusting the model parameters of the initial target model according to the loss, and continuing to train until a stop condition is met to obtain the target model.

[0011] In some embodiments of the present application, the attention weight between nodes is determined according to the vector representation of each node, including obtaining the first vector representation and the second vector representation corresponding to the third node and the fourth node in the knowledge graph respectively, determining the first similarity between the first vector representation and the second vector representation, determining the second similarity commonly corresponding to the third node and all neighbor nodes, and determining the attention weight according to the first similarity and the second similarity.

[0012] In some embodiments of the present application, the association rule library is determined according to the attention weight, comprising: comparing the attention weight with a first attention weight threshold to obtain a first comparison result; in the case that the first comparison result indicates that the attention weight is greater than or equal to the first attention weight threshold, determining that the first node pair corresponding to the attention weight belongs to a strong association rule; comparing the attention weight with a second attention weight threshold to obtain a second comparison result; in the case that the second comparison result indicates that the attention weight is less than or equal to the second attention weight threshold, determining that the second node pair corresponding to the attention weight belongs to a weak association rule; and determining the association rule library according to the strong association rule and the weak association rule.

[0013] In some embodiments of the present application, the slice parameter set is verified by using the association rule library to obtain a verification result, comprising: matching the slice parameters in the slice parameter set with the association rules in the association rule library to obtain a matching result; in the case that the matching result indicates that there is a slice parameter combination in the slice parameters that satisfies the weak association rule in the association rule, determining a conflict score corresponding to the slice parameter combination, wherein the weak association rule is used to reflect a parameter combination that should not appear at the same time, and the conflict score is used to quantitatively represent the conflict degree of the slice parameter combination; and in the case that the conflict score is greater than or equal to a preset score, determining that the verification result is a verification failure.

[0014] In some embodiments of the present application, the conflict score corresponding to the slice parameter combination is determined, comprising: determining a parameter dimension corresponding to each slice parameter combination, wherein the parameter dimension includes a scene parameter dimension, a function parameter dimension, and a type parameter dimension; determining a sub-conflict score corresponding to each slice parameter combination; and performing weighted summation on the sub-conflict scores by using the weight corresponding to the parameter dimension to obtain the conflict score.

[0015] In some embodiments of the present application, the parameter configuration strategy corresponding to the slice parameter set is determined, comprising: obtaining a user priority corresponding to a target object, wherein each user priority corresponds to a preset configuration strategy; and determining the parameter configuration strategy corresponding to the slice parameter set according to the preset configuration strategy corresponding to the user priority.

[0016] In some embodiments of the present application, the slice parameters of multiple dimensions include scene slice parameters used to reflect business scenarios in the configuration requirement information, function slice parameters used to reflect network performance and service priority in the configuration requirement information, and type slice parameters used to reflect parameter attributes in the configuration requirement information.

[0017] According to another aspect of the embodiments of this application, a parameter configuration apparatus is also provided, comprising: a conversion module, configured to convert configuration requirement information of a target object for a network slice into a slice parameter set, wherein the slice parameter set includes slice parameters of multiple dimensions; a verification module, configured to verify the slice parameter set using an association rule base to obtain a verification result, wherein the association rule base is used to quantify the association information and conflict information between slice parameters of different dimensions; a determination module, configured to determine a parameter configuration strategy corresponding to the slice parameter set if the verification result indicates successful verification; and a configuration module, configured to configure the network slice using the parameter configuration strategy.

[0018] According to another aspect of the embodiments of this application, an electronic device is also provided, including: a memory and a processor, wherein the memory is used to store program instructions; the processor is connected to the memory and is used to execute a method for implementing the above-described parameter configuration.

[0019] According to another aspect of the embodiments of this application, a non-volatile storage medium is also provided, the non-volatile storage medium including a stored computer program, wherein the device where the non-volatile storage medium is located executes the above-mentioned parameter configuration method by running the computer program.

[0020] According to another aspect of the embodiments of this application, a computer program product is also provided, including computer instructions that, when executed by a processor, implement the above-described parameter configuration method.

[0021] In this embodiment, a multi-dimensional analysis approach is adopted. By transforming and extracting the network slice configuration requirements of the target object, a multi-dimensional slice parameter set is formed. Then, an association rule base is invoked to perform in-depth verification of the slice parameter set, quantitatively analyze the implicit associations and potential conflicts between slice parameters of different dimensions, and then determine the parameter configuration strategy corresponding to the slice parameter set and automatically configure and distribute the network slice. This achieves the purpose of pre-identifying and avoiding parameter configuration conflicts, thereby improving the technical effect of network slice service quality. It also solves the technical problem that related technologies only consider the compliance of the parameters themselves when configuring network slice parameters, resulting in conflicts between parameters in the configuration strategy and causing the service quality to fail to meet the standards after slice deployment. Attached Figure Description

[0022] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0023] Figure 1 This is a hardware structure block diagram of a computer terminal according to a parameter configuration method according to an embodiment of this application;

[0024] Figure 2 This is a flowchart of a parameter configuration method according to an embodiment of this application;

[0025] Figure 3 This is a flowchart illustrating a method for configuring parameters according to an embodiment of this application, specifically a network slice configuration and distribution process.

[0026] Figure 4 This is a schematic diagram of a multi-parameter auxiliary configuration device for a parameter configuration method according to an embodiment of this application;

[0027] Figure 5 This is a schematic diagram of a parameter configuration device according to an embodiment of this application. Detailed Implementation

[0028] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0029] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0030] To better understand the embodiments of this application, the technical terms involved in the embodiments of this application are explained below:

[0031] Enhanced Mobile Broadband (eMBB): A service scenario in 5G communication technology designed to significantly improve the data transmission rate and bandwidth capabilities of mobile networks to meet the high data rate demands of applications such as high-definition video streaming, virtual reality (VR), and augmented reality (AR). In this embodiment, eMBB, as one of the core scenarios, participates in the construction of a knowledge graph. Its parameter configuration rules are learned through a Generative Neural Network (GNN) to achieve a high degree of matching with user service needs, ensuring the service quality of high-speed data services.

[0032] Natural Language Processing (NLP) is an important branch of computer science and artificial intelligence that studies how to enable computers to understand, parse, and generate human natural language (including text and speech) and achieve interaction with human language. In this application embodiment, NLP technology is used to parse network slice request information submitted by users in the form of text or speech, and convert it into structured scene, function, and type parameters for input into subsequent parameter configuration processes.

[0033] Quality of Service (QoS) is a set of metrics used in network communication to measure data transmission performance and ensure differentiated service requirements for various businesses, such as bandwidth, latency, jitter, and packet loss rate. In this embodiment, QoS metrics are a key component of network slicing parameter configuration. Through cross-dimensional correlation modeling, they can be intelligently adjusted to adapt to different scenarios and user needs, ensuring the stability of slice services and user satisfaction.

[0034] Graph Sampling and Aggregation (GraphSAGE) is a graph neural network (GNN) architecture that learns low-dimensional vector representations of nodes in a graph by sampling and aggregating features from the nodes' neighbors, thereby capturing the relationships between nodes. In this embodiment, the GraphSAGE model is used to construct cross-dimensional relationships between slice parameters. By learning implicit association rules between scene parameters, function parameters, and type parameters, it enables automatic identification and verification of parameter configuration conflicts.

[0035] With the continuous evolution of network technology, enterprises and users are increasingly demanding higher quality communication services. Coupled with the diversification of customer groups, this has spurred a strong demand for personalized network service packages. Furthermore, with the accelerated commercialization of 5G networks, network slicing, as a core technology for achieving "one network, multiple uses," requires configuring parameters (such as bandwidth, latency, and QoS levels) according to the differentiated needs of different users (e.g., VIP enterprise users, ordinary individual users, and industry clients). However, the slicing parameter configurations used in related technologies suffer from the following pain points:

[0036] (1) Fragmented parameter parsing: Traditional configuration relies on manual extraction of user needs (such as text / voice consultation), and cannot use NLP technology to systematically parse the three-dimensional parameters of scenario, function and type, which easily overlooks core needs such as "use in different locations" and "low latency".

[0037] (2) Lack of correlation verification: The parameter configuration only focuses on single-dimensional compliance (such as whether the value exceeds the threshold) and does not consider cross-dimensional correlation conflicts (such as configuring the "ultra-low latency" parameter in the eMBB scenario), resulting in QoS failure after slice deployment and an error rate of over 15%.

[0038] (3) Fixed matching strategy: The same matching logic is used for VIP exclusive users, shared users and users with special needs, without dynamically adjusting the threshold based on priority (e.g., the priority difference of shared users is not reflected in the parameter matching), resulting in low resource utilization and uneven user experience.

[0039] (4) Insufficient iteration capability: The association rules rely on manual updates and cannot adapt to emerging scenarios such as 6G evolution and XR mixed reality. When new business needs emerge, the configuration efficiency drops by more than 50%.

[0040] To address the aforementioned technical problems, this application provides corresponding solutions, which are detailed below.

[0041] The parameter configuration method provided in this application can be executed on a mobile terminal, computer terminal, or similar computing device. Figure 1 A hardware block diagram of a computer terminal for implementing a parameter configuration method is shown. Figure 1 As shown, the computer terminal 10 may include one or more processors (shown as 102a, 102b, ..., 102n in the figure) (the processor may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 104 for storing data, and a transmission module 106 for communication functions connected via wired and / or wireless networks. In addition, it may also include: a display, a keyboard, a cursor control device, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, and a BUS bus. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0042] It should be noted that the aforementioned one or more processors and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10. As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0043] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the parameter configuration method in the embodiments of this application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby implementing the above-mentioned parameter configuration method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor, and these remote memories can be connected to the computer terminal 10 via a network. Examples of the above-mentioned networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0044] The transmission module 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission module 106 includes a network interface controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission module 106 may be a radio frequency (RF) module, used for wireless communication with the Internet.

[0045] The display can be, for example, a touchscreen liquid crystal display (LCD) that allows the user to interact with the user interface of the computer terminal 10.

[0046] It should be noted here that, in some optional embodiments, the above... Figure 1 The computer terminal shown may include hardware elements (including circuitry), software elements (including computer code stored on a computer-readable medium), or a combination of both hardware and software elements. It should be noted that... Figure 1 This is only one instance of a specific particular instance, and is intended to illustrate the types of components that may exist in the aforementioned computer terminal.

[0047] In the above operating environment, this application provides a method embodiment for parameter configuration. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0048] Figure 2 This is a flowchart of a parameter configuration method according to an embodiment of this application, such as... Figure 2 As shown, the method includes the following steps:

[0049] Step S202: Convert the target object's configuration requirements for network slices into a slice parameter set, wherein the slice parameter set includes slice parameters of multiple dimensions.

[0050] In step S202 above, the configuration requirement information refers to the specific performance indicators, functional requirements and scenario preferences of network slices (such as 5G network slices) proposed by the target object (such as enterprises, individual users, industry customers, etc.) based on their own business characteristics and network service requirements.

[0051] The slice parameter set is a set of standardized parameters extracted and transformed from configuration requirement information with the help of NLP technology. In some embodiments of this application, the multi-dimensional slice parameters include scenario slice parameters that reflect the business scenarios in the configuration requirement information, function slice parameters that reflect the network performance and service priority in the configuration requirement information, and type slice parameters that reflect the parameter attributes in the configuration requirement information.

[0052] In other words, these parameters cover scenario dimensions (such as eMBB, mMTC, uRLLC), functional dimensions (such as bandwidth, latency, QoS level), and type dimensions (such as numerical, option, and text types), which together constitute a complete configuration requirement description of the target network slice. Each parameter represents a user's requirements or attributes for a certain aspect of the network slice.

[0053] In some specific embodiments of this application, the text / voice requests input by the user through the UI can be processed by an NLP model and then parsed into a slice parameter set S according to the following three dimensions:

[0054] (1) Scene parameter labels: Match the three major 5G service scenarios and mark eMBB (enhanced mobile broadband), mMTC (massive machine-type communication), and uRLLC (ultra-reliable ultra-low latency communication) labels. For example, the user requirement "high-definition live broadcast bandwidth guarantee" corresponds to the eMBB label.

[0055] (2) Functional parameter tags, which may include the following two types of sub-tags:

[0056] Network function tags: Extract attributes such as latency (e.g., "latency ≤ 10ms"), QoS level (e.g., "Gold QoS"), and connection count (e.g., "millions of connections");

[0057] Priority auxiliary tags: labeled A (highest, VIP exclusive), B, C, D, E (lowest) according to user level, used for subsequent matching strategy selection.

[0058] (3) Parameter type label: Calibrate parameter attributes to fit threshold selection:

[0059] Numerical values: such as bandwidth "100Mbps", latency "5ms";

[0060] Options include: such as QoS level "Gold / Silver / Bronze" and encryption method "SSL / TLS";

[0061] Text-based: such as custom requirements "Support industrial Ethernet protocol".

[0062] By using NLP technology for parsing and type recognition, step S202 transforms the fuzzy configuration requirement information into a structured set of slice parameters, solving the problems of fragmented parameter parsing and lack of type recognition, and providing clear and comprehensive data input for cross-dimensional correlation modeling and multi-scenario matching.

[0063] Step S204: The set of slice parameters is validated using an association rule base to obtain validation results. The association rule base is used to quantify the association and conflict information between slice parameters of different dimensions.

[0064] In step S204 above, the association rule base is a storage system constructed from historical configuration data and expert experience. It contains quantified representations of positive association information (i.e., strong association rules) and conflict information (i.e., weak association rules / conflicting rules) between cross-dimensional slice parameters. For example, this rule base can be built based on knowledge graphs and graph neural networks (GNNs) to perform real-time consistency checks during parameter configuration, ensuring that the configured slice parameter set is coordinated across multiple dimensions such as scenario, function, and type, and conforms to business and technical specifications.

[0065] The verification result refers to the conclusion drawn after verifying the slice parameter set using an association rule base. It reflects whether the parameter set meets the consistency requirements. In some embodiments of this application, if the parameter set conforms to strong association rules and does not trigger any conflicting rules, the verification result indicates successful verification; conversely, if non-compliant associations or conflicts are detected, the verification result will indicate the specific conflicting parameter pair and trigger the exception handling process.

[0066] To address the lack of cross-dimensional parameter validation in traditional configurations, a method based on knowledge graphs and attention weights can be used for association rule validation. Specifically, the knowledge graph structure and the attention weights trained on the GraphSAGE model (the target model) are used for association rule validation. The knowledge graph pre-stores scene parameters, function parameters, type parameters, and their inter-parameter association strengths. The GraphSAGE model learns these deep-level relationships, represented by attention weights between nodes. When validating a slice of parameters, the system queries and calculates the attention weights of each parameter pair in the knowledge graph to assess the association between parameters. If the attention weight of a parameter pair is lower than a set threshold θ (indicating a conflicting rule), or higher than θ but lower than another preset threshold (indicating a weak positive association), the system determines that the parameter pair is conflicting or inconsistent, and the validation fails.

[0067] Taking a configuration containing "eMBB" scenario parameters and "ultra-low latency" function parameters as an example, since "eMBB" (enhanced mobile broadband) and "ultra-low latency" are marked as conflict rules in the knowledge graph (for example, the attention weight is 0.25), when these two parameters appear in the configuration, the verification algorithm will determine that there is a conflict based on the attention weight calculation result, thus producing a verification failure result, prompting the configuration personnel or the system to adjust the parameters.

[0068] In some embodiments of this application, the association rule base is determined by: obtaining historical parameter configuration data corresponding to historical network slices; determining a knowledge graph corresponding to the historical parameter configuration data, wherein the nodes in the knowledge graph are historical slice parameters in the historical parameter configuration data, and the edges of the knowledge graph are the association strength between nodes; determining the attention weight between nodes based on the vector representation of each node, wherein the attention weight is used to quantify the importance of the association rules corresponding to two nodes; and determining the association rule base based on the attention weight.

[0069] It should be noted that historical parameter configuration data refers to the actual record dataset of network operators configuring parameters for various network slices over a period of time in the past. It includes slice parameter instances of various scenarios (such as eMBB, mMTC, uRLLC), functions (such as bandwidth, latency), and types (numerical, option, text).

[0070] A knowledge graph is a structured database used to represent and store the relationships between slice parameters. Nodes in the graph represent different slice parameters, while edges represent the strength of the relationship between nodes, quantifying the positive association or conflict between parameters, such as the strength of the relationship between "eMBB scene parameters" and "high bandwidth function parameters".

[0071] Association strength is an indicator that quantifies the degree of association between two slice parameters. It is usually between 0 and 1. The larger the value, the stronger the association between the two parameters, that is, the higher the probability of them appearing at the same time in a configuration. In the embodiments of this application, association strength can be calculated by weighted summation of parameter co-occurrence frequency and expert weight in historical parameter configuration data. It directly determines the weight of the edge between nodes.

[0072] Vector representation refers to converting each slice parameter into a low-dimensional vector using node features learned through a graph neural network (GNN). This captures the intrinsic properties of the parameters and their interrelationships. In this embodiment, the GraphSAGE model can be used to embed and encode nodes, enabling each slice parameter to be represented not only by its own attributes but also by incorporating features from its neighboring parameters in the knowledge graph, resulting in a richer representation.

[0073] Attention weights are learned through the GraphSAGE model and are used to quantify the importance of the association rules between node u and node v. They reflect the necessity or rationality of two parameters appearing simultaneously in a network slice configuration. High attention weights mean that there is a strong correlation between the parameters, which is suitable for configuration in the same scenario; conversely, it may indicate that there is a conflict between the parameters, which should not appear at the same time.

[0074] Specifically, the first step in building a knowledge graph is to extract slice parameters from the historical parameter configuration dataset and define them as nodes in the graph. Next, the co-occurrence frequency between parameters in the historical configuration is statistically analyzed to calculate the association strength between each pair of parameters. The GraphSAGE model is used to perform deep learning on the slice parameter nodes in the knowledge graph. By aggregating the information of the nodes and their neighbors, a vector representation of each node is generated, and then the attention weight between nodes is calculated. Through the calculation of attention weights, it is possible to identify which parameter combinations are strong association rules that frequently appear in the historical configuration and meet business requirements, and which are conflicting rules that are theoretically difficult to coexist.

[0075] In some embodiments of this application, the knowledge graph corresponding to the historical parameter configuration data can be determined in the following way: extract historical slice parameters of multiple dimensions from the historical parameter configuration data, and use the historical slice parameters as nodes of the knowledge graph; determine the association strength between any two nodes, and use the association strength as the weight of the edge between any two nodes; remove the edges in the knowledge graph whose association strength is less than the preset association strength to obtain the knowledge graph.

[0076] The association strength is determined as follows: the co-occurrence frequency of the first node and the second node in the knowledge graph is determined, wherein the co-occurrence frequency includes the proportion of the number of times the first node and the second node appear simultaneously to the total number of configurations in the historical parameter configuration data; the preset weights corresponding to the first node and the second node are obtained, wherein the preset weights are used to reflect the association importance between the first node and the second node based on business characteristics; and the association strength is obtained by weighted summation of the co-occurrence frequency and the preset weights.

[0077] Specifically, knowledge graphs are the foundation for carrying cross-dimensional relationships, and it is necessary to clarify node definitions, rules for calculating edge association strength, and the method for constructing the initial graph, including:

[0078] (1) Node definition (entity design).

[0079] The parameters of the three dimensions (for example, the number of dimensions is not limited) are abstracted into nodes of the knowledge graph, specifically divided into:

[0080] 1) Scene parameter nodes: three types of core nodes: eMBB (enhanced mobile broadband), mMTC (massive machine-type communication), and uRLLC (ultra-reliable low-latency communication).

[0081] 2) Functional parameter nodes: further divided into two types of sub-nodes:

[0082] Network function tags: such as "high bandwidth", "low latency", "ultra-high reliability", "high QoS level", etc. (12 categories in total, covering the core functional requirements of 5G).

[0083] Priority auxiliary labels: such as "A (highest)", "B", "C", "D" and "E (lowest)" (5 categories, corresponding to user priority).

[0084] 3) Parameter type nodes: Numerical (such as bandwidth, latency values), Option (such as QoS level options "Gold / Silver / Bronze"), and Text (such as custom description "Supports industrial control protocols") are the three types.

[0085] (2) Definition of edge (association strength calculation).

[0086] Edges represent the relationships between nodes. The degree of association is quantified by the association strength value (0-1). The calculation method is: association strength = α × co-occurrence frequency + (1-α) × expert weight.

[0087] Co-occurrence frequency: The proportion of times two nodes (such as "eMBB" and "high bandwidth") appear simultaneously in the total number of configurations in the historical parameter configuration data (range 0-1).

[0088] Expert weight: The importance of association is marked by 5G network experts according to the characteristics of the service (range 0-1, such as "eMBB→high bandwidth" is marked as 0.9, "eMBB→ultra-low latency" is marked as 0.1).

[0089] α is the balance coefficient: it is set to 0.7 by default (based on actual data and supplemented by expert experience), and can be dynamically adjusted according to the amount of data (α is reduced to 0.5 when the amount of data is small).

[0090] For example, in 1000 historical configurations, "eMBB" and "high bandwidth" co-occurred 800 times, with a co-occurrence frequency of 0.8; and an expert weight of 0.9, then the association strength is 0.7 × 0.8 + 0.3 × 0.9 = 0.79.

[0091] (3) Initial knowledge graph construction process.

[0092] Input: Historical 5G slice parameter configuration data (100,000+ records, including scenario, function, and type parameters), and expert-annotated association rules (200+ records).

[0093] Processing: Calculate the edge weights of all node pairs using the above association strength formula, retain edges with a strength greater than or equal to the preset association strength (e.g., 0.3) (filtering out weak associations), and form an initial knowledge graph (containing 18 types of nodes and 30+ core association edges).

[0094] In some embodiments of this application, the attention weights between nodes can be determined in the following way: the knowledge graph is processed by a target model to obtain a vector representation of each node, wherein the target model determines the vector representation of each node by aggregating the neighbor information of the nodes in the knowledge graph; and the attention weights between nodes are determined based on the vector representation of each node.

[0095] The target model is trained as follows: training data is acquired, which includes the association strength of training node pairs in the training knowledge graph; the initial target model is used to process the training data to obtain the prediction vector representation of each training node, and the prediction attention weight of the training node pair is determined based on the prediction vector representation; the loss corresponding to the prediction attention weight and the association strength is determined; the model parameters of the initial target model are adjusted based on the loss, and training continues until the stopping condition is met to obtain the target model.

[0096] Specifically, GraphSAGE (graph sampling generation model, i.e. target model) can be used as the core GNN architecture. By aggregating the information of the node's neighbors, the vector representation of the node can be learned, thereby mining implicit association rules.

[0097] The model structure (3-layer architecture) can be designed as follows:

[0098] (1) Input layer: Initialize the embedding for each node (convert the node ID into a low-dimensional vector), with embedding dimension d0=32 (adaptively adjusted according to the number of nodes, taking 32 when the number of nodes is ≤50).

[0099] (2) Hidden layer: 2 GraphSAGE aggregation layers:

[0100] Layer 1: Features of the first-order neighbors (directly associated nodes) of the aggregated node, with an output dimension d1=64, using the mean aggregator function.

[0101] Layer 2: Aggregate the features of the second-order neighbors (neighbors of neighbors) of the aggregation node, with an output dimension d2=128, using the pooling aggregation function (max-poolingaggregator).

[0102] (3) Output layer: The final vector representation of the nodes (dimension 128) is output through the fully connected layer, and the attention weight between nodes is calculated to quantify the importance of the association.

[0103] Model training includes:

[0104] Training data: Association strength of all node pairs in the knowledge graph (as labels), totaling 5000+ samples.

[0105] Loss function: Mean Squared Error (MSE), the optimization objective is to minimize the error between the predicted attention weights and the actual association strength.

[0106] Optimizer: Adam, learning rate η=0.001, decay rate β1=0.9, β2=0.999.

[0107] Training rounds: 50 rounds (earlystopping strategy, stopping if the validation set loss does not decrease for 5 consecutive rounds).

[0108] Regularization: L2 regularization (coefficient λ=1e-5) to prevent overfitting.

[0109] After obtaining the vector representations, the attention weights between nodes can be determined as follows: obtain the first and second vector representations corresponding to the third and fourth nodes in the knowledge graph, respectively; determine the first similarity between the first and second vector representations; determine the second similarity between the third node and all its neighboring nodes; and determine the attention weights based on the first and second similarities.

[0110] Specifically, for any two nodes u (eMBB) and v (high bandwidth), their attention weights a u,v The calculation formula is: au,v =exp(cos(h u ,h k )) / (∑ k∈N(u) exp(cos(h u ,h v ))).

[0111] in, h u , h k The vector representation of nodes u and v output by the GNN; cos( h u , h v ) represents the vector cosine similarity, which measures the similarity of node features; N ( u ) is a node u The set of all neighboring nodes, attention weights a u,v The higher the value (range 0-1), the more likely it is to be considered as a high value. u and v The more significant the association (stronger the rule), the stronger the correlation.

[0112] To automatically identify strong association rules between cross-dimensional parameters based on historical data to guide configuration and promptly detect and prevent conflicting parameters in the configuration, the association rule base can be determined as follows: Compare the attention weight with a first attention weight threshold to obtain a first comparison result; if the first comparison result indicates that the attention weight is greater than or equal to the first attention weight threshold, determine that the first node pair corresponding to the attention weight belongs to a strong association rule; compare the attention weight with a second attention weight threshold to obtain a second comparison result; if the second comparison result indicates that the attention weight is less than or equal to the second attention weight threshold, determine that the second node pair corresponding to the attention weight belongs to a weak association rule; determine the association rule base based on the strong and weak association rules.

[0113] Specifically, after the GNN model is trained, "strong association rules" are extracted and stored as a rule base for subsequent consistency verification.

[0114] (1) Threshold for strong association rule determination.

[0115] Set the attention weight threshold θ = 0.7 (as an example), and validate based on historical data: when a u,v When the value is ≥0.7, the association accuracy is ≥95%, satisfying a. u,v Node pairs with a value greater than or equal to θ are considered "strong association rules".

[0116] Example of a strong association rule:

[0117] eMBB → High bandwidth (a=0.85), eMBB → Numerical (a=0.82);

[0118] uRLLC → Ultra-low latency (a=0.90), uRLLC → High reliability (a=0.88);

[0119] mMTC → High connection count (a=0.80), mMTC → Option type (a=0.75).

[0120] (2) Construction of conflict rule base.

[0121] Simultaneously extract "weak association rules" (a u,v ≤0.3) is a conflict rule, which means that theoretically, parameter combinations should not occur at the same time.

[0122] Example conflict rules (i.e., weak association rules):

[0123] eMBB → Ultra-low latency (a=0.25), uRLLC → Low bandwidth (a=0.18).

[0124] In the above steps, by automatically identifying strong correlation rules in historical configuration data, the accuracy and efficiency of parameter configuration are improved. By automatically identifying weak correlation rules, the system can perform cross-dimensional conflict detection, effectively avoiding the limitations of single-dimensional verification and reducing the error rate of parameter configuration.

[0125] After constructing the association rule base, the slice parameter set can be validated in the following way to obtain the validation results: match the slice parameters in the slice parameter set with the association rules in the association rule base to obtain the matching results; if the matching results indicate that there are slice parameter combinations that satisfy weak association rules in the association rules, determine the conflict score corresponding to the slice parameter combination. Here, weak association rules are used to reflect parameter combinations that should not appear at the same time, and conflict scores are used to quantify the degree of conflict of slice parameter combinations; if the conflict score is greater than or equal to the preset score, the validation result is determined as validation failure.

[0126] The process involves determining the conflict score corresponding to each slice parameter combination: determining the parameter dimension corresponding to each slice parameter combination, where the parameter dimension includes the scene parameter dimension, the function parameter dimension, and the type parameter dimension; determining the sub-conflict score corresponding to each slice parameter combination; and using the weights corresponding to the parameter dimensions to perform a weighted summation of the sub-conflict scores to obtain the conflict score.

[0127] To facilitate understanding of the above verification process, some specific embodiments are explained below.

[0128] Specifically, when parsing a new parameter set S (i.e., the slice parameter set), the association rule base is called to automatically verify the consistency between parameters across multiple dimensions such as scenario, function, and type. The specific process is as follows:

[0129] (1) Explanation of input parameters and their sources.

[0130] The sources of parameters in the slice parameter set can be divided into three main categories:

[0131] User requirement parameters: These are submitted by the user through the interface or API during the business application stage, specifying the business use case and performance requirements (such as "eMBB" scenario parameters).

[0132] Network slice template parameters: a library of predefined slice function attributes by operators, containing standard function configurations for various scenarios (such as "high bandwidth" and "high QoS" function parameters).

[0133] System adaptation parameters: Parameters automatically generated based on network resource status, such as type parameters (numerical / enumerated), which are automatically allocated by the system according to the resource pool configuration;

[0134] (2) Assume the newly parsed parameters are:

[0135] Scenario parameter: eMBB (user-specified service scenario as enhanced mobile broadband);

[0136] Functional parameters: {"Ultra-low latency", "Priority B"} (User-defined functional requirements + system default priority configuration);

[0137] Type parameter: Numeric (automatically assigned by the system based on resource pool characteristics);

[0138] (3) Cross-dimensional association query and relationship mining.

[0139] By constructing a parameter association graph (i.e., a knowledge graph), we can perform compatibility and incompatibility analysis between requirements and slice parameters.

[0140] Positive association rules (i.e. strong association rules): Based on historical successful business configuration data, a strong association model between scenarios and functions is established (e.g., eMBB → high bandwidth, confidence level 0.95).

[0141] Conflict rules (i.e. weak association rules): Integrate operator operation and maintenance experience and fault cases to define conflict parameter pairs (e.g., {eMBB, ultra-low latency} conflict coefficient 0.25).

[0142] The specific verification steps include:

[0143] Step 1: Query the strongly related functional parameters of the scenario parameter "eMBB" (which should be "high bandwidth", "high QoS", etc.) and find that the current functional parameter contains "ultra-low latency" (which belongs to the conflict rule).

[0144] Step 2: Calculate the inconsistency score (Sinc, i.e., the conflict score):

[0145]

[0146] Where C is the set of node pairs that belong to the conflict rule in the current parameters; is the compatibility coefficient for the parameter pair (u,v) (values ​​range from 0 to 1, provided by the correlation map).

[0147] In this example, C contains only (eMBB, ultra-low latency), so Sinc = 1 - 0.25 = 0.75.

[0148] (4) Abnormal trigger threshold and dynamic optimization.

[0149] Threshold setting mechanism (i.e., preset score):

[0150] Static threshold: T=0.5 (based on historical error case statistics: when Sinc≥0.5, the parameter configuration error rate is ≥80%).

[0151] Dynamic threshold: By monitoring the business success rate after parameter configuration in real time, the threshold coefficient is updated every hour to ensure detection sensitivity.

[0152] If Sinc ≥ T, an error message will be triggered, which reads: "Conflicting parameter detected: eMBB scenario does not match the 'ultra-low latency' function. It is recommended to replace it with 'high bandwidth'." A list of similar parameter recommendations will also be provided.

[0153] (5) Multi-dimensional joint verification and core capability construction.

[0154] When there are multiple conflicting dimensions, calculate the inconsistency score for each dimension separately. (i.e., sub-conflict score).

[0155] Perform weighted summation:

[0156]

[0157] The weighting is as follows (as an example): Scenario dimension w1=0.4 (user needs-driven), Function dimension w2=0.4 (business performance core), Type dimension w3=0.2 (system adaptation auxiliary).

[0158] When Total ≥ 0.5, an exception is triggered, and a conflict parameter correction scheme is provided to achieve a closed-loop capability from detection to optimization.

[0159] By continuously learning new configuration cases and dynamically updating the parameter association graph and conflict rule base, we gradually build an adaptive parameter verification core capability, thereby improving the accuracy and efficiency of network slice configuration.

[0160] Step S206: If the verification result indicates that the verification was successful, determine the parameter configuration strategy corresponding to the slice parameter set.

[0161] In step S206 above, the parameter configuration strategy can be designed based on the successfully verified slice parameter set and user priority, and is an implementation scheme used to guide the configuration of network slice parameters.

[0162] In some embodiments of this application, a parameter configuration strategy based on user priority can be determined. Specifically, the user priority corresponding to the target object is obtained, wherein each user priority corresponds to a preset configuration strategy; and the parameter configuration strategy corresponding to the slice parameter set is determined according to the preset configuration strategy corresponding to the user priority.

[0163] For example, after confirming that the slice parameter set is compliant and conflict-free, the system needs to generate corresponding parameter configuration strategies based on user priority (such as VIP users, shared users, and users with special needs): For VIP users (Level A), the system directly calls the preset exclusive configuration parameters. These parameters are pre-filled according to user needs and business scenarios to ensure that the slice configuration of the highest priority user meets all high-level service requirements; the configuration strategy for shared users (Levels B and C) is more flexible. The system will adjust the parameter matching threshold according to the user level. For example, Level C users may have their bandwidth threshold increased by 10% to adapt to the resource allocation needs of different priorities; for users with special needs (no preset strategy), a customized strategy is generated through a comprehensive method of scenario anchoring, function aggregation, and error verification to ensure that all types of users can obtain network slice configurations that meet their needs.

[0164] In determining the parameter configuration strategy, the system needs to comprehensively consider scenario parameters, functional parameters, and type parameters to ensure the comprehensiveness and rationality of the slice parameter configuration. For each scenario parameter, the system first matches the most relevant functional parameter group (e.g., matching "high bandwidth" and "high QoS" for the eMBB scenario), then selects the most suitable configuration threshold and options based on the characteristics of the type parameters (numerical, option-based, and text-based), and finally combines the user priority adjustment strategy to generate the final parameter configuration scheme. This process utilizes the strong association rules and threshold adjustment strategies stored in the association rule base to achieve multi-dimensional and accurate matching of slice parameters.

[0165] In some specific embodiments of this application, the parameter configuration strategy can be determined in the following ways:

[0166] Differentiated matching strategies are adopted for different user types in parameter set S:

[0167] (1) VIP Exclusive Slice User Matching (Priority A):

[0168] Auto-fill: Directly calls VIP-exclusive configuration parameters (such as 500Mbps fixed bandwidth for enterprise users);

[0169] Default completion: Default parameters are filled with the highest priority preset value (e.g., the default delay parameter is filled with "10ms").

[0170] Manual backup: After filling in the parameters, check the error requests that exceed the threshold (such as bandwidth > 1000Mbps) and transfer them to manual processing.

[0171] (2) Shared slice user matching (priority BE):

[0172] Threshold calibration: Set dynamic thresholds according to priority, such as B level +5% (bandwidth threshold increased by 5%), C level +10%, D level -5%, E level -10%;

[0173] Layered matching: First, match scenario parameters (such as mMTC scenario), then match option-type (QoS level) and text-type parameters, and finally select numerical parameter policy groups based on thresholds (such as adjusting the bandwidth threshold of Class C users from 100Mbps to 110Mbps).

[0174] (3) Handling of special slice parameter requirements (non-A level + no preset strategy):

[0175] Scene anchoring: First, match similar scenes according to scene parameters (e.g., anchor uRLLC for "Industrial Control");

[0176] Functional aggregation: Group by functional attributes (e.g., Group A = {bandwidth, latency}, Group B = {QoS, number of connections}), calculate the median threshold of each group (e.g., A' = middle{A1-A, A2-A, A3-A}), and take the policy group corresponding to max{A', B'}.

[0177] Error verification: Calculate the deviation rate (a1'-a1) / a1, and if the weighted average is greater than the priority allowable threshold p' (e.g., p'=5% for level B), proceed to manual processing; otherwise, execute the configuration.

[0178] In the above steps, by adjusting the threshold strategy based on user priority, resources are allocated reasonably to users with different priorities, improving the utilization efficiency of network resources. Especially when resources are scarce, the configuration can be dynamically optimized to meet the needs of more users. Secondly, by matching strategies based on comprehensive scenario, function, and type parameters, the comprehensiveness and rationality of slice parameter configuration are achieved, avoiding the limitations of single-dimensional matching. This ensures that the configuration strategy can cover various business scenarios, improving the flexibility and adaptability of network slicing services.

[0179] Step S208: Configure network slices using parameter configuration strategies.

[0180] In step S208 above, network slice configuration refers to the parameter setting process for a specific slice in the 5G network according to the parameter configuration strategy, including but not limited to the adjustment of key indicators such as bandwidth, latency, and QoS level.

[0181] In some implementations of this application, once parameter configuration strategies are generated, the system will automatically apply these strategies to configure network slice parameters. For example, through the interactive design of the UI-Server, users can input their required parameters via the UI. After receiving the user input, the UI-Server calls the NLP model to parse the parameters, then performs automated configuration according to the matching strategy, and sends the configuration information to the network slice management platform through the parameter configuration distribution module, which then performs the specific parameter adjustments.

[0182] Suppose a VIP user (Level A) wants to customize a network slice for their enterprise applications, requiring "high bandwidth" and "low latency" features under the "enhanced mobile broadband" scenario. After NLP analyzes these requirements, the system automatically populates the exclusive configuration parameters (e.g., fixed bandwidth of 500Mbps, latency less than 10ms) and confirms these configuration details with the user via the UI-Server. Once confirmed, the system automatically triggers the slice parameter distribution command. After receiving the command, the network slice management platform executes the parameter adjustments to complete the slice configuration.

[0183] Furthermore, to adapt to new business scenarios (such as 6G evolution and emerging industry demands), a dynamic optimization mechanism is introduced. This involves iterative updates to the rule base and regular updates to the knowledge graph and GNN model. For example:

[0184] Update cycle: once every 3 months, or triggered when the cumulative number of new parameter configuration data reaches 10,000.

[0185] Update methods include: adding new nodes (such as adding the scene parameter "XR Mixed Reality"); recalculating association strength (incorporating new data); fine-tuning the GNN model (using incremental training, freezing the parameters of the previous layer, and training only the output layer); and parameter adaptation, i.e., the thresholds θ and T are automatically adjusted with the update (dynamically optimizing the association accuracy based on the new data).

[0186] Through steps S202 to S208, a multi-dimensional analysis approach is adopted. By transforming and extracting the configuration requirements information of the target object for network slices, a multi-dimensional set of slice parameters is formed. Then, the association rule base is called to perform in-depth verification of the slice parameter set, quantitatively analyze the implicit associations and potential conflicts between slice parameters of different dimensions, and then determine the parameter configuration strategy corresponding to the slice parameter set and automatically configure and distribute the network slices. This achieves the purpose of pre-identifying and avoiding parameter configuration conflicts, thereby improving the technical effect of network slice service quality. In turn, it solves the technical problem that related technologies only consider the compliance of the parameters themselves when configuring network slice parameters, resulting in conflicts between parameters in the configuration strategy and causing the service quality to fail to meet the standards after slice deployment.

[0187] Figure 3 This is a flowchart illustrating a network slice configuration and distribution process according to an embodiment of this application, specifically a parameter configuration method. Figure 3 As shown, the process includes:

[0188] (1) NLP multi-dimensional parameter extraction module.

[0189] This module is responsible for automatically extracting scene, function, and type parameters related to the slice configuration from the user's natural language description (whether text or voice input), transforming vague requirements into a clear and actionable set of parameters. Specifically, the NLP model can first preprocess the user's input text or voice to remove noise and irrelevant words, and then use techniques such as Named Entity Recognition (NER), syntactic analysis, and semantic understanding to identify scene parameters (such as "eMBB"), function parameters (such as "high bandwidth" and "low latency"), and type parameters (such as "numerical" and "optional").

[0190] (2) Cross-dimensional parameter correlation modeling module and dynamic optimization module.

[0191] The cross-dimensional parameter correlation modeling module is used to create and maintain a knowledge graph. It learns the implicit relationships between parameters through a graph neural network (GNN) to form a rule base for verifying the consistency of parameter configurations. The dynamic optimization module is responsible for automatically updating and optimizing the knowledge graph and GNN model when business scenarios change or when a certain amount of new configuration data accumulates, ensuring the timeliness and accuracy of parameter configuration strategies.

[0192] (3) Multi-scenario parameter matching module.

[0193] This module's task is to select the most suitable parameter group for network slicing configuration based on user priority and specific scenario requirements, while ensuring the correlation between parameters. By adopting differentiated matching strategies for different types of users, the module can effectively improve resource utilization while meeting the service expectations of users with different priorities. Specifically, different matching logics and thresholds can be set according to user type (VIP exclusive users, shared users, users with special needs). For example, VIP users are directly filled with preset exclusive parameters, shared users have their thresholds dynamically adjusted according to priority, and users with special needs find the parameter combination closest to their needs through scenario anchoring, function aggregation, and error verification. In addition, for the parameter matching process, firstly, the module uses scenario parameters to match similar scenario slice templates; then, it groups and sorts according to functional parameters (network functions and priority assistance) and selects the parameters that best meet the needs; finally, based on user priority and resource status, it determines the specific value or type of the parameters and generates the final parameter configuration strategy.

[0194] (4) Parameter configuration distribution module.

[0195] Once the parameter configuration strategy is determined, this module is responsible for translating the strategy into actual network slice parameters and distributing them to the network slice management platform for configuration. Furthermore, the UI design allows users to view and confirm configuration details, ensuring they have a clear understanding of the specific parameters and enhancing the transparency of the configuration process and user experience.

[0196] It should be noted that the above modules can be used to execute... Figure 2 The parameter configuration method shown is therefore Figure 2 The explanations and instructions regarding parameter configuration methods also apply to the modules mentioned above.

[0197] Figure 4 This is a schematic diagram of a multi-parameter auxiliary configuration device according to an embodiment of the present application, illustrating a parameter configuration method. Figure 4 As shown, users submit network slicing configuration requests to the portal module through this module. The portal module, located at the system front end, is responsible for receiving requests submitted by user modules and supporting user parameter input and confirmation through UI design. The multi-parameter auxiliary configuration module is the core of the entire intelligent configuration process. It integrates NLP multi-dimensional parameter extraction, cross-dimensional parameter correlation modeling, multi-scenario parameter matching, and dynamic optimization functions to achieve full automation and intelligence from requirement analysis to parameter distribution.

[0198] In the entire intelligent configuration process of network slice parameters, it starts from the user submitting a request through the user module. The portal module receives the request and performs preliminary processing. Then, the multi-parameter auxiliary configuration module uses NLP extraction, correlation modeling, multi-scenario matching and other technologies to automatically generate a parameter configuration strategy that meets the user's needs and network resource conditions. After the configuration strategy is finally confirmed by the portal module, the portal module executes the distribution of slice parameters to complete the deployment of network slices.

[0199] The embodiments of this application can achieve at least the following technical effects:

[0200] I. Cracking the fragmentation of parameter parsing.

[0201] To address the issue of easily overlooked requirements when manually extracted, a three-dimensional analysis system of "scenario-function-type" is constructed using NLP. This system automatically identifies eMBB / mMTC / uRLLC scenarios, latency / QoS and other functions and parameter types, transforming user requirements into a standardized set of parameters to avoid analysis bias.

[0202] 2. Fill the gaps in the association verification.

[0203] To address the shortcomings of single-dimensional compliance verification, a rule base for association is built using knowledge graphs and the GraphSAGE model to quantify parameter associations and conflicts. Conflicts (such as eMBB + ultra-low latency) are identified through "conflict score + abnormal triggering", reducing the QoS non-compliance rate from 15%+ to below 3%.

[0204] Third, achieve differentiated matching.

[0205] To address the drawbacks of a unified matching logic, a layered design strategy is adopted: VIP users (Level A) are automatically filled with exclusive parameters; shared users (Levels B and C) have their thresholds adjusted according to priority (e.g., Level C + 10%); users with special needs are matched through "scenario anchoring - function aggregation - error verification" to achieve "personalized solutions for each user".

[0206] IV. Adapting to Business Evolution.

[0207] To address the difficulty of manually updating rules, a dynamic optimization mechanism is constructed: every 3 months or when 10,000 new data entries are added, a new scenario node is added, the model is incrementally trained, and parameters (α / θ) are adaptively adjusted to avoid efficiency drops of 50%+ due to scenario iteration, thus laying the foundation for 6G.

[0208] V. Build end-to-end automated processes.

[0209] To address the issues of numerous manual steps and high risks, a "UI interaction-extraction-matching-confirmation-distribution" process was designed, supporting text / voice input and automating the workflow. VIP configuration time was reduced from 2 hours to 5 minutes, resulting in a 50% cost reduction and a 30%+ increase in resource utilization, thus promoting industry implementation.

[0210] Figure 5 This is a structural diagram of a device with parameter configuration according to an embodiment of this application, such as... Figure 5 As shown, the device includes:

[0211] The conversion module 502 is used to convert the configuration requirement information of the target object for network slices into a slice parameter set, wherein the slice parameter set includes slice parameters of multiple dimensions.

[0212] The verification module 504 is used to verify the set of slice parameters using the association rule base and obtain the verification result. The association rule base is used to quantify the association information and conflict information between slice parameters of different dimensions.

[0213] The determination module 506 is used to determine the parameter configuration strategy corresponding to the slice parameter set when the verification result indicates that the verification is successful.

[0214] Configuration module 508 is used to configure network slices using parameter configuration strategies.

[0215] It should be noted that, Figure 5 The device with the parameters shown is used to perform Figure 2 The parameter configuration method shown is therefore Figure 2 The explanations and instructions regarding parameter configuration methods also apply to... Figure 5 The device with the parameters shown will not be described in detail here.

[0216] This application also provides an electronic device, which includes a memory and a processor, wherein the memory is used to store program instructions; the processor is connected to the memory and is used to execute the steps of the parameter configuration method implemented in various embodiments of this application.

[0217] This application also provides a non-volatile storage medium including a stored computer program, wherein the device containing the non-volatile storage medium executes the steps of the parameter configuration method in various embodiments of this application by running the computer program.

[0218] This application also provides a computer program product, including computer instructions that, when executed by a processor, implement the steps of the parameter configuration method in various embodiments of this application.

[0219] This application also provides a computer program that, when executed by a processor, implements the steps of the parameter configuration method in various embodiments of this application.

[0220] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0221] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0222] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0223] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0224] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0225] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0226] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for configuring parameters, characterized in that, include: The configuration requirements of the target object for network slices are converted into a slice parameter set, wherein the slice parameter set includes slice parameters of multiple dimensions. The set of slice parameters is validated using an association rule base to obtain validation results. The association rule base is used to quantify the association and conflict information between slice parameters of different dimensions. If the verification result indicates that the verification was successful, determine the parameter configuration strategy corresponding to the slice parameter set; The network slice is configured using the aforementioned parameter configuration strategy.

2. The method according to claim 1, characterized in that, The association rule base is determined in the following way: Obtain historical parameter configuration data corresponding to historical network slices; A knowledge graph corresponding to the historical parameter configuration data is determined, wherein the nodes in the knowledge graph are the historical slice parameters in the historical parameter configuration data, and the edges of the knowledge graph are the association strength between the nodes; The attention weights between nodes are determined based on the vector representation of each node, wherein the attention weights are used to quantify the importance of the association rules corresponding to two nodes; The association rule base is determined based on the attention weights.

3. The method according to claim 2, characterized in that, Determining the knowledge graph corresponding to the historical parameter configuration data includes: Historical slice parameters of multiple dimensions are extracted from the historical parameter configuration data, and the historical slice parameters are used as nodes of the knowledge graph; Determine the association strength between any two nodes, and use the association strength as the weight of the edge between the two nodes; The knowledge graph is obtained by removing edges whose association strength is less than a preset association strength.

4. The method according to claim 2, characterized in that, The correlation strength is determined in the following way: Determine the co-occurrence frequency of the first node and the second node in the knowledge graph, wherein the co-occurrence frequency includes the proportion of the number of times the first node and the second node appear simultaneously to the total number of configurations of the historical parameter configuration data; Obtain the preset weights corresponding to the first node and the second node, wherein the preset weights are used to reflect the importance of the association between the first node and the second node based on business characteristics; The correlation strength is obtained by weighting and summing the co-occurrence frequency and the preset weight.

5. The method according to claim 2, characterized in that, Determining the attention weights between nodes based on the vector representation of each node includes: The knowledge graph is processed using a target model to obtain a vector representation of each node, wherein the target model determines the vector representation of each node by aggregating the neighbor information of the nodes in the knowledge graph; The attention weights between nodes are determined based on the vector representation of each node.

6. The method according to claim 5, characterized in that, The target model is trained in the following way: Acquire training data, wherein the training data includes the association strength of training node pairs in the training knowledge graph; The training data is processed using an initial target model to obtain a prediction vector representation for each training node, and the prediction attention weights for training node pairs are determined based on the prediction vector representations. Determine the loss corresponding to the predicted attention weight and the association strength; The model parameters of the initial target model are adjusted based on the loss, and training continues until the stopping condition is met to obtain the target model.

7. The method according to claim 2, characterized in that, Determining the attention weights between nodes based on the vector representation of each node includes: Obtain the first vector representation and the second vector representation corresponding to the third node and the fourth node in the knowledge graph, respectively; Determine the first similarity between the first vector representation and the second vector representation; Determine the second similarity between the third node and all its neighboring nodes; The attention weight is determined based on the first similarity and the second similarity.

8. The method according to claim 2, characterized in that, Determining the association rule base based on the attention weights includes: The attention weight is compared with the first attention weight threshold to obtain the first comparison result; If the first comparison result indicates that the attention weight is greater than or equal to the first attention weight threshold, it is determined that the first node pair corresponding to the attention weight belongs to a strong association rule. The attention weight is compared with the second attention weight threshold to obtain a second comparison result; If the second comparison result indicates that the attention weight is less than or equal to the second attention weight threshold, it is determined that the second node pair corresponding to the attention weight belongs to a weak association rule. The association rule base is determined based on the strong association rules and the weak association rules.

9. The method according to claim 1, characterized in that, The set of slice parameters is validated using an association rule base, and the validation results are obtained, including: The slice parameters in the slice parameter set are matched with the association rules in the association rule base to obtain the matching results; If the matching result indicates that there is a combination of slice parameters that satisfies the weak association rule in the association rule, the conflict score corresponding to the combination of slice parameters is determined, wherein the weak association rule is used to reflect parameter combinations that should not occur at the same time, and the conflict score is used to quantify the degree of conflict of the combination of slice parameters. If the conflict score is greater than or equal to the preset score, the verification result is determined to be a verification failure.

10. The method according to claim 9, characterized in that, Determining the conflict score corresponding to the combination of slice parameters includes: Determine the parameter dimensions corresponding to each of the slice parameter combinations, wherein the parameter dimensions include scene parameter dimensions, function parameter dimensions, and type parameter dimensions; Determine the sub-conflict score corresponding to each of the slice parameter combinations; The conflict score is obtained by weighting and summing the sub-conflict scores using the weights corresponding to the parameter dimensions.

11. The method according to claim 1, characterized in that, Determining the parameter configuration strategy corresponding to the slice parameter set includes: Obtain the user priority corresponding to the target object, wherein each user priority corresponds to a preset configuration strategy; The parameter configuration strategy corresponding to the slice parameter set is determined based on the preset configuration strategy corresponding to the user priority.

12. The method according to claim 1, characterized in that, The multi-dimensional slicing parameters include scenario slicing parameters that reflect the business scenarios in the configuration requirement information, function slicing parameters that reflect the network performance and service priority in the configuration requirement information, and type slicing parameters that reflect the parameter attributes in the configuration requirement information.

13. A parameter configuration device, characterized in that, include: The conversion module is used to convert the configuration requirements information of the target object for network slices into a slice parameter set, wherein the slice parameter set includes slice parameters of multiple dimensions. The verification module is used to verify the set of slice parameters using an association rule base to obtain verification results. The association rule base is used to quantify the association information and conflict information between slice parameters of different dimensions. The determination module is used to determine the parameter configuration strategy corresponding to the slice parameter set when the verification result indicates that the verification is successful. The configuration module is used to configure the network slice using the parameter configuration strategy.

14. An electronic device, characterized in that, include: A memory and a processor, the memory being used to store program instructions; the processor being connected to the memory and used to execute the method for implementing the parameter configuration according to any one of claims 1 to 12.

15. A non-volatile storage medium, characterized in that, The non-volatile storage medium includes a stored computer program, wherein the device containing the non-volatile storage medium executes the parameter configuration method according to any one of claims 1 to 12 by running the computer program.

16. A computer program product comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the parameter configuration method according to any one of claims 1 to 12.

Images