Information processing method, network recommendation method and model training method

Abstract

The embodiments of this specification provide an information processing method, a networking recommendation method, and a model training method, wherein the information processing method includes: using a heterogeneous graph network model to process the networking topology data associated with a cloud computing project into a networking topology vector, and storing the networking topology vector in a topology vector database; upon receiving networking description information, inputting the networking description information into a networking recommendation model, wherein the networking recommendation model includes an information processing unit and a retrieval unit; processing the networking description information by the information processing unit to obtain a networking description vector, and using the retrieval unit to retrieve a target networking topology vector that matches the networking description vector in the topology vector database; determining a target networking structure based on the target networking topology vector, and constructing networking structure graphic information corresponding to the target networking structure.

Description

Technical Field

[0001] The embodiments of this specification relate to the field of information processing technology, and in particular to information processing methods, networking recommendation methods, and model training methods. Background Art

[0002] With the development of cloud computing technology, the demand for enterprises and individuals to build and manage networks in cloud environments is growing, and cloud networking has become crucial. However, choosing a networking solution that meets expectations is a difficult challenge. Most of the existing automated tools complete networking recommendations based on fixed templates or simple algorithms. However, if the networking intent, specific service scenarios, and limitations brought about by security-related factors are not fully considered in the actual application stage, the recommended user networking configuration may be inaccurate. In addition, the complex cloud service and device topology structure often causes the recommendation algorithm to ignore actual constraints, the recommendation system is not real-time enough, and cannot quickly adapt to the changing environment. Moreover, this implementation basis is mostly based on historical networking data, which makes the recommendation process opaque and seriously affects the user's trust. Therefore, an effective solution is urgently needed to solve the above problems. Summary of the invention

[0003] In view of this, an embodiment of this specification provides an information processing method. One or more embodiments of this specification also relate to a networking recommendation method, a model training method, an information processing device, a networking recommendation device, a model training device, a computing device, a computer-readable storage medium, and a computer program product to solve the technical defects existing in the prior art.

[0004] According to a first aspect of an embodiment of this specification, there is provided an information processing method, including:

[0005] Using a heterogeneous graph network model, the networking topology data associated with the cloud computing project is processed into a networking topology vector, and the networking topology vector is stored in a topology vector database;

[0006] When receiving the network description information, inputting the network description information into a network recommendation model, wherein the network recommendation model includes an information processing unit and a retrieval unit;

[0007] Processing the networking description information by the information processing unit to obtain a networking description vector, and using the retrieval unit to retrieve a target networking topology vector matching the networking description vector in the topology vector database;

[0008] A target networking structure is determined according to the target networking topology vector, and networking structure graphic information corresponding to the target networking structure is constructed.

[0009] According to a second aspect of an embodiment of this specification, a networking recommendation method is provided, which is applied to a cloud-side device, including:

[0010] Receiving the network description information submitted by the terminal side device, and inputting the network description information into the network recommendation model, wherein the network recommendation model includes an information processing unit and a retrieval unit;

[0011] Processing the networking description information by the information processing unit to obtain a networking description vector, and using the retrieval unit to retrieve a target networking topology vector matching the networking description vector in a topology vector database;

[0012] Determine a target networking structure according to the target networking topology vector, and construct networking structure graphic information corresponding to the target networking structure;

[0013] The networking structure graphic information is sent to the terminal side device.

[0014] According to a third aspect of an embodiment of this specification, a model training method is provided, including:

[0015] Obtain the network configuration information corresponding to the users participating in the cloud computing project;

[0016] Determine the network topology data corresponding to the user according to the network configuration information, and store the network topology data in a graph database;

[0017] Preprocessing the networking topology data stored in the graph database, and constructing a networking knowledge graph according to the preprocessing results;

[0018] The networked knowledge graph is used to train an initial heterogeneous graph network model including an adaptive attention mechanism until a heterogeneous graph network model that meets the training stop condition is obtained, wherein the heterogeneous graph network model is applied to the above-mentioned information processing method.

[0019] According to a fourth aspect of the embodiments of this specification, there is provided an information processing device, including:

[0020] A processing module is configured to process the networking topology data associated with the cloud computing project into a networking topology vector using a heterogeneous graph network model, and store the networking topology vector in a topology vector database;

[0021] An input module, configured to input the networking description information into a networking recommendation model when receiving the networking description information, wherein the networking recommendation model includes an information processing unit and a retrieval unit;

[0022] A retrieval module is configured to process the networking description information through the information processing unit to obtain a networking description vector, and use the retrieval unit to retrieve a target networking topology vector matching the networking description vector in the topology vector database;

[0023] The construction module is configured to determine the target networking structure according to the target networking topology vector, and construct the networking structure graphic information corresponding to the target networking structure.

[0024] According to a fifth aspect of an embodiment of this specification, a networking recommendation device is provided, which is applied to a cloud-side device, including:

[0025] An information receiving module is configured to receive the network description information submitted by the terminal side device and input the network description information into the network recommendation model, wherein the network recommendation model includes an information processing unit and a retrieval unit;

[0026] an information processing module, configured to process the networking description information through the information processing unit to obtain a networking description vector, and use the retrieval unit to retrieve a target networking topology vector matching the networking description vector in a topology vector database;

[0027] An information construction module is configured to determine a target networking structure according to the target networking topology vector, and to construct networking structure graphic information corresponding to the target networking structure;

[0028] The information sending module is configured to send the networking structure graphic information to the terminal side device.

[0029] According to a sixth aspect of the embodiments of this specification, a model training device is provided, including:

[0030] An information acquisition module is configured to acquire network configuration information corresponding to users participating in the cloud computing project;

[0031] A data determination module is configured to determine the network topology data corresponding to the user according to the network configuration information, and store the network topology data in a graph database;

[0032] A data processing module is configured to preprocess the networking topology data stored in the graph database and construct a networking knowledge graph according to the preprocessing result;

[0033] The training model module is configured to use the networked knowledge graph to train an initial heterogeneous graph network model including an adaptive attention mechanism until a heterogeneous graph network model that meets the training stop condition is obtained, wherein the heterogeneous graph network model is applied to the above-mentioned information processing method.

[0034] According to a seventh aspect of an embodiment of this specification, there is provided a computing device, including:

[0035] A memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the above-mentioned information processing method, networking recommendation method, or model training method.

[0036] According to an eighth aspect of the embodiments of this specification, a computer-readable storage medium is provided, which stores computer-executable instructions, which, when executed by a processor, implement the steps of the above-mentioned information processing method or networking recommendation method or model training method.

[0037] According to the ninth aspect of the embodiments of this specification, a computer program product is provided, including a computer program or instructions, which, when executed by a processor, implement the steps of the above-mentioned information processing method or networking recommendation method or model training method.

[0038] The information processing method provided in this embodiment can firstly process the network topology data associated with the cloud computing project into a network topology vector by using a heterogeneous graph network model, and store the network topology vector in a topology vector database in order to more accurately recommend the network structure to the user, and to be more explanatory and convenient for the user to understand the network distribution. The method is used to record the vector expression of all network topology data through the topology vector database, so that the user can choose from it when recommending the network structure to the user. Then, when the network description information is received, in order to improve the efficiency and accuracy of the network recommendation, the network description information can be first input into the network recommendation model, wherein the network recommendation model includes an information processing unit and a retrieval unit; on this basis, the network description information can be processed by the information processing unit to obtain the network description vector, and after the network description information is converted into a vector expression, the retrieval unit can be used to retrieve the target network topology vector matching the network description vector in the topology vector database; the target network topology vector can be considered as the vector expression corresponding to the network structure that better matches the network description information. Therefore, the target networking structure can be determined according to the target networking topology vector, and the networking structure graphic information corresponding to the target networking structure can be constructed, so as to improve the interpretability of the currently recommended target networking structure, thereby increasing the user's willingness to accept it. At the same time, reflecting the target networking structure through graphic information can effectively reduce the complexity of network understanding and accelerate the popularization of simple and easy-to-use intelligent cloud networks. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is a schematic diagram of an information processing method provided by an embodiment of this specification;

[0040] Figure 2is a flow chart of an information processing method provided by an embodiment of this specification;

[0041] Figure 3 is a flowchart of a recommended networking method provided by an embodiment of this specification;

[0042] Figure 4 is a flow chart of a model training method provided by an embodiment of this specification;

[0043] Figure 5 It is a processing flow chart of a recommended networking method provided by an embodiment of this specification;

[0044] Figure 6 is a structural diagram of an information processing device provided by an embodiment of this specification;

[0045] Figure 7 It is a structural diagram of a recommended networking device provided by an embodiment of this specification;

[0046] Figure 8 It is a structural schematic diagram of a model training device provided by an embodiment of this specification;

[0047] Fig. 9 It is a structural block diagram of a computing device provided by an embodiment of this specification. DETAILED DESCRIPTION

[0048] Many specific details are described in the following description to facilitate a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of this specification, so this specification is not limited to the specific implementation disclosed below.

[0049] The terms used in one or more embodiments of this specification are only for the purpose of describing specific embodiments, and are not intended to limit one or more embodiments of this specification. The singular forms of "a", "said" and "the" used in one or more embodiments of this specification and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0050] It should be understood that although the terms first, second, etc. may be used to describe various information in one or more embodiments of this specification, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of this specification, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0051] In addition, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0052] In one or more embodiments of this specification, a large model refers to a deep learning model with large-scale model parameters, which usually contains hundreds of millions, tens of billions, hundreds of billions, trillions, or even more than 10 trillion model parameters. A large model can also be called a foundation model / foundation model. The large model is pre-trained with large-scale unlabeled corpus to produce a pre-trained model with more than 100 million parameters. This model can adapt to a wide range of downstream tasks, and the model has good generalization ability, such as a large-scale language model (LLM), a multi-modal pre-training model, etc.

[0053] When the big model is used in practice, only a small number of samples are needed to fine-tune the pre-trained model and it can be applied to different tasks. The big model can be widely used in natural language processing (NLP), computer vision and other fields. Specifically, it can be applied to computer vision tasks such as visual question answering (VQA), image caption (IC), image generation, as well as natural language processing tasks such as text-based sentiment classification, text summary generation, and machine translation. The main application scenarios of the big model include digital assistants, intelligent robots, search, online education, office software, e-commerce, intelligent design, etc.

[0054] First, the terms involved in one or more embodiments of this specification are explained.

[0055] HetGraph (Heterogeneous Graph): A graph data structure in which nodes and edges can represent a variety of different entities and relationships. For example, in a cloud network topology, nodes can be servers, storage devices, network devices, etc., and edges can represent connections, dependencies, traffic, and other relationships.

[0056] GAT (Graph Attention Network): Graph Attention Network introduces the attention mechanism into the processing of graph data, allowing the model to dynamically assign weights based on the node's neighbor information to improve the efficiency and accuracy of information aggregation.

[0057] HetGAT (Heterogeneous Graph Attention Network): A graph neural network model that processes heterogeneous graph data. It considers the adjacency of nodes and edges, as well as their types and attributes. The model can better understand and model different types of entities and their interactions.

[0058] AAM (Adaptive Attention Mechanism): is a mechanism used to process sequence data in deep learning models. It allows the model to dynamically adjust its focus of attention based on the contextual information of the input data, helping the model to assign weights according to the importance of nodes.

[0059] Large vertical models in the cloud network field: large pre-trained models customized for the unique knowledge and tasks in the cloud network field to better understand and process complex information and professional terms in this field.

[0060] In this specification, an information processing method is provided. One or more embodiments of this specification also involve a networking recommendation method, a model training method, an information processing device, a networking recommendation device, a model training device, a computing device, a computer-readable storage medium, and a computer program product, which are described in detail one by one in the following embodiments.

[0061] See also Figure 1As shown in the schematic diagram, the information processing method provided in this embodiment can firstly use the heterogeneous graph network model to process the network topology data associated with the cloud computing project into a network topology vector, and store the network topology vector in a topology vector database in order to more accurately recommend the network structure to the user and ensure that the recommendation result is more explanatory and more convenient for the user to understand the network distribution. The network topology data associated with the cloud computing project can be processed into a network topology vector, and the network topology vector can be stored in a topology vector database; the vector expression of all network topology data can be recorded through the topology vector database, so that the network structure can be selected when the user is recommended. Then, when the network description information is received, in order to improve the efficiency and accuracy of the network recommendation, the network description information can be first input into the network recommendation model, wherein the network recommendation model includes an information processing unit and a retrieval unit; on this basis, the network description information can be processed by the information processing unit to obtain the network description vector, and after the network description information is converted into a vector expression, the retrieval unit can be used to retrieve the target network topology vector matching the network description vector in the topology vector database; the target network topology vector can be considered as the vector expression corresponding to the network structure that better matches the network description information. Therefore, the target networking structure can be determined according to the target networking topology vector, and the networking structure graphic information corresponding to the target networking structure can be constructed, so as to improve the interpretability of the currently recommended target networking structure, thereby increasing the user's willingness to accept it. At the same time, reflecting the target networking structure through graphic information can effectively reduce the complexity of network understanding and accelerate the popularization of simple and easy-to-use intelligent cloud networks.

[0062] See also Figure 2 , Figure 2 A flowchart of an information processing method provided according to an embodiment of the present specification is shown, which specifically includes the following steps.

[0063] Step S202: Using a heterogeneous graph network model, the networking topology data associated with the cloud computing project is processed into a networking topology vector, and the networking topology vector is stored in a topology vector database.

[0064] The information processing method provided in this embodiment is applied to cloud computing scenarios, and is used to recommend a target networking structure based on the networking requirements submitted by the user, so that the user can connect the computing resources, storage resources and network resources in different geographical locations, different cloud service providers or different resource pools through a virtual network based on cloud computing technology according to the networking structure recommendation results, so as to form a logically unified and physically dispersed virtual network environment, so that users can apply it to operations such as enterprise cloud migration, data center interconnection, and cross-regional service deployment.

[0065] Computing resources mainly include servers, which are the core of data processing and computing capabilities on the cloud. Servers can be in various forms, including but not limited to rack servers, blade servers, virtualization technology, etc. Storage resources are used to store and manage data and are an important part of cloud networking. They mainly include hard disk arrays (such as RAID), network attached storage (NAS), storage switches, and storage devices (such as SSD, HDD). Network resources are the basis of data exchange and information services on the cloud, and mainly include routers, switches, firewalls, network adapters, network management software, etc.

[0066] Specifically, the heterogeneous graph network model specifically refers to a model for converting the networking topology data associated with the cloud computing project into a networking topology vector. The model can use the graph neural network model HetGAT for processing heterogeneous graph data, so that the networking topology data corresponding to different users in the cloud computing project can be converted into vector expressions and stored in the topology vector database, so that in the network structure recommendation stage, the vector can be directly read from the database to complete the calculation, thereby improving the efficiency of network structure recommendation. Correspondingly, the networking topology data specifically refers to the networking structure data corresponding to different users in the cloud computing project, which is determined according to the network configuration information corresponding to the user, and is used to reflect the network structure established by the user when participating in the cloud computing project. Since the service provider of the cloud computing project has a large number of served users, different users have different networking topology data. Therefore, by constructing vectors of the networking topology data corresponding to different users, it is used for matching and recommending when a new user is networking, so that it is easier to select a networking structure that matches the networking needs of the new user. Correspondingly, the networking topology vector specifically refers to the vector expression of the corresponding networking structure level obtained after the network topology data is vector-converted, which corresponds to the vector expression of the entire networking structure of the user. Correspondingly, the topology vector database specifically refers to a database for storing networking topology vectors corresponding to different users, which can be directly read by the vectors for calculation in the networking recommendation phase.

[0067] Based on this, in order to recommend the network structure to users more accurately and with stronger explanation so that users can understand the network distribution more conveniently, the powerful association representation ability of the heterogeneous graph network model can be used to perform high-dimensional, comprehensive, and deep-level feature learning and network modeling on the network topology data corresponding to real users, so as to establish the network topology vector and store it in the topology vector database. When the network description information is received, the intention and demand of the network description information can be identified based on the powerful understanding ability of the large language model, and then similarity matching can be performed in the unified vector space, and the networking process can be explained and visualized using the fine-tuned large language model, so as to obtain the network structure graphic information corresponding to the network description information, thereby effectively improving the user's willingness to accept. At the same time, the target network structure can be reflected through graphic information, which can effectively reduce the complexity of network understanding and accelerate the popularization of simple and easy-to-use intelligent cloud networks. Among them, the large language model is a network recommendation model.

[0068] Furthermore, when constructing a networking topology vector using a heterogeneous graph network model, in order to enable the networking topology data to be used as a processed object of the heterogeneous graph network model, it is necessary to convert the networking configuration information corresponding to the user into the networking topology data of the corresponding graph database. In this embodiment, the specific implementation method is as follows:

[0069] Obtain network configuration information corresponding to users participating in a cloud computing project; determine network topology data corresponding to the user based on the network configuration information, and store the network topology data in a graph database; use a heterogeneous graph network model to process the network topology data stored in the graph database to obtain a network topology vector corresponding to the network topology data.

[0070] Specifically, the network configuration information refers to the configuration information related to the network structure corresponding to the users participating in the cloud computing project. The configuration information is used to record the usage of the computing resources, storage resources and network resources corresponding to the users, the network topology relationship, the network scale, etc. Correspondingly, the graph database refers to the Neo4J graph database used to store the network topology data, which is convenient for efficient management and query. Among them, the nodes in the graph database can represent resources, such as computing resources, storage resources and network resources; the edges can represent the relationship between resources and additional attributes, such as configuration details, traffic data, etc.

[0071] Based on this, in order to realize the conversion of all network topology data associated with the cloud computing project into vectors and store them in the topology vector database for network recommendation in the application stage, the network configuration information corresponding to the users participating in the cloud computing project can be obtained first; the network configuration information can reflect the configuration details of the user's corresponding network structure, and then the network topology data corresponding to the user can be determined according to the network configuration information, so as to realize the conversion of the network configuration information into the user's actual network topology data, and in order to facilitate the capture of the network relationship in the network topology data, the network topology data can be stored in the graph database; the management efficiency and query efficiency can be improved through the characteristics of the graph database. On this basis, the heterogeneous graph network model with the ability to process graph data can be used to process the network topology data stored in the graph database to obtain the network topology vector corresponding to the network topology data, and then store it in the topology vector database, which can be used in the network recommendation stage.

[0072] In practical applications, in order to extract the user's networking configuration information from the cloud computing project and complete the construction of the networking topology vector, the cloud computing project service provider can provide a cloud network dedicated network topology interface to display and integrate the networking configuration information of all users in the cloud computing project. This interface can effectively capture the topological structure, device configuration and connection relationship of various user networks, and then obtain the networking topology data of each user's corresponding networking structure. On this basis, the networking topology data corresponding to each user can be uniformly stored in the Neo4J graph database for subsequent efficient management and query.

[0073] In summary, by using the user's networking configuration information in the cloud computing project to build networking topology data, and then using it to construct the networking topology vector, it is possible to build networking topology data based on real data, which can effectively improve the accuracy of networking recommendations in the application stage.

[0074] Furthermore, when the heterogeneous graph network model processes the networking topology data stored in the graph database, it actually uses the heterogeneous graph network model to perform vector fusion of nodes and edges in the networking topology data, thereby obtaining the networking topology vector of the corresponding networking structure level. In this embodiment, the specific implementation method is as follows:

[0075] The networking topology data stored in the graph database is input into the heterogeneous graph network model; the target node and the target edge are extracted from the networking topology data through the heterogeneous graph network model, and a target node embedding vector corresponding to the target node and a target edge embedding vector corresponding to the target edge are constructed; the target node embedding vector and the target edge embedding vector are normalized, and a networking topology vector corresponding to the networking topology data is generated according to the normalization result.

[0076] Specifically, the target node and the target edge specifically refer to the nodes and edges contained in the networking topology data, wherein the node is the resource in the networking structure, such as computing resources, storage resources and network resources, and the edge is the relationship between resources, such as traffic data, configuration details, etc. Correspondingly, the target node embedding vector specifically refers to the embedding vector expression corresponding to the target node, and the target edge embedding vector specifically refers to the embedding vector expression corresponding to the target edge.

[0077] Based on this, after storing the network topology data in the graph database, the network topology relationship in each user's network structure can be reflected through the graph structure relationship. On this basis, it is necessary to combine the model with graph data processing capabilities to process the network topology data in the graph database, so a heterogeneous graph network model can be used. In this process, the network topology data stored in the graph database can be input into the heterogeneous graph network model; all target nodes and target edges contained in the network topology data can be extracted through the heterogeneous graph network model.

[0078] On this basis, the target node embedding vector corresponding to the target node and the target edge embedding vector corresponding to the target edge can be constructed; the characteristics of all nodes and edges in the networking topology data can be expressed through vectors, and then the target node embedding vector and the target edge embedding vector can be normalized to achieve splicing and fusion of the vector expressions corresponding to the nodes and edges involved in the networking topology data through normalization, and then the networking topology vector corresponding to the networking topology data can be generated according to the normalization results, so as to facilitate subsequent networking recommendations.

[0079] In summary, by using a heterogeneous graph network model with graph data processing capabilities to process the networking topology data contained in the graph database, it is possible to convert the graph data into a vector expression, and the vector expression can fully reflect the structural characteristics and attribute characteristics of each networking topology data, thereby facilitating the improvement of the accuracy of networking structure recommendations in the recommendation stage.

[0080] In this process, when constructing the target node embedding vector corresponding to the target node in the networking topology data, in order to ensure that the vector expression can better reflect the attributes of the node, so that the final networking topology vector represents the networking topology data, the node vector fusion method can be used for construction. In this embodiment, the specific implementation method is as follows:

[0081] Determine the neighbor nodes corresponding to the target node, and construct an initial node embedding vector corresponding to the target node and a neighbor node embedding vector corresponding to the neighbor node; calculate the attention score between the target node and the neighbor node according to the initial node embedding vector and the neighbor node embedding vector through the dot product attention mechanism in the heterogeneous graph network model; generate the target node weight of the target node and the neighbor node weight of the neighbor node according to the attention score; perform weighted sum calculation on the initial node embedding vector and the target node weight, as well as the neighbor node embedding vector and the neighbor node weight, to obtain the target node embedding vector corresponding to the target node.

[0082] Specifically, neighbor nodes specifically refer to other nodes in the network topology data that have a direct connectivity relationship with the currently processed target node. Correspondingly, the initial node embedding vector specifically refers to the vector expression corresponding to the target node without integrating other topological structure information, that is, the initial node embedding vector is the vector expression of the corresponding node level. Correspondingly, the neighbor node embedding vector specifically refers to the vector expression of the neighbor node corresponding to the node level. Correspondingly, the dot-product attention mechanism specifically refers to the attention mechanism used to calculate the attention score between the neighbor node and the target node in the heterogeneous graph network model.

[0083] Based on this, in order to enable the target node embedding vector corresponding to the target node to fully represent the vector of the target node relative to the networking structure level, and then integrate the edge embedding vector to reflect the networking topology vector of the networking topology data, the neighbor nodes corresponding to the target node can be determined in the networking topology data during the vector construction stage. At this time, the initial node embedding vector corresponding to the target node and the neighbor node embedding vector corresponding to the neighbor node can be constructed first.

[0084] On this basis, the dot product attention mechanism in the heterogeneous graph network model can be used to calculate the attention score between the target node and the neighbor node according to the initial node embedding vector and the neighbor node embedding vector; the attention score can be used to reflect the strength of the attachment relationship between the target node and the neighbor node, and then the target node weight of the target node and the neighbor node weight of the neighbor node can be generated according to the attention score, and the influence relationship between each other can be reflected by the size of the node weight. Then, the initial node embedding vector and the target node weight, as well as the neighbor node embedding vector and the neighbor node weight can be weighted and calculated, and the target node embedding vector corresponding to the target node can be obtained through the weighted and calculated, thereby obtaining the vector expression of the target node relative to the networking structure level. The construction of the networking topology vector can be completed on this basis in the future.

[0085] In practical applications, in order to realize the use of heterogeneous graph network models to convert network topology data into network topology vectors, so as to fully reflect the characteristics of network topology data through the vector expression of the network structure level, the heterogeneous graph network model can be constructed to include multiple attention mechanism sublayers, so that it can calculate the weighted sum of the target node and the neighboring nodes, and obtain the vector expression of the network structure level corresponding to the node. Among them, the weight calculation can be obtained by the feature vectors of the two nodes and the learnable attention coefficient matrix. The attention scores between the nodes are calculated by the dot product attention mechanism, and then normalized by the softmax function to obtain the final attention weight.

[0086] For all nodes of each type, the node embedding vector corresponding to the node can be calculated using the above method, and then the node embedding vectors corresponding to different types of nodes in the networking topology data are fused to obtain the networking topology vector corresponding to the corresponding networking topology data.

[0087] Furthermore, skip-connection can be used to add the output of the previous layer to the output of the current layer, and multiple HetGAT layers can be stacked. Each layer performs the above processing operations to gradually extract deeper graph features. On this basis, in order to improve the stability of the vector space and the accuracy of the similarity calculation, the embedding vector of each node and edge can be normalized by L2 norm. Then, a networking topology vector that fully characterizes the characteristic attributes of the networking topology data can be obtained, and then it can be stored in the topology vector database, so that it can be directly read in the application stage for networking structure recommendation.

[0088] In summary, when constructing a node embedding vector for the target node, by combining it with the processing method of the networking structure hierarchy to vectorize it, the final target node embedding vector can fully represent the characteristic attributes of the node in the networking topology data, and the networking topology vector constructed on this basis is more accurate, which is convenient for use in the application stage.

[0089] In addition, in order to enable the heterogeneous graph network model to effectively complete the mapping of networking topology data to networking topology vectors, after storing the networking topology data in the graph database, the data stored in the graph database can be used to construct a knowledge graph, and then the knowledge graph is used to train the initial heterogeneous graph network model, so that the model has vector conversion capabilities, and then subsequent predictions can be used. In this embodiment, the specific implementation method is as follows:

[0090] Preprocessing is performed on the networking topology data stored in the graph database, and graph nodes, graph edges and networking attributes are determined according to the preprocessing results; a networking knowledge graph is constructed based on the graph nodes, the graph edges and the networking attributes; and an initial heterogeneous graph network model including an adaptive attention mechanism is trained using the networking knowledge graph until a heterogeneous graph network model that meets the training stop conditions is obtained.

[0091] Specifically, preprocessing specifically refers to the operations of data cleaning, standardization, and outlier processing of the networking topology data stored in the graph database, which is used to normalize the networking topology data in the graph database and unify the standards. Correspondingly, the graph node specifically refers to the graph node constructed in combination with the node data contained in the networking topology data. For example, the networking topology data is the networking structure corresponding to user A, which includes computing node 1 and storage node 1. Then computing node 1 and storage node 1 can both be used as graph nodes, and their corresponding edges are graph edges, and the networking attributes are the additional attributes between nodes. Correspondingly, the training stop condition specifically refers to the condition for stopping the training of the initial heterogeneous graph network model, which includes but is not limited to the loss value comparison condition, the verification set verification condition, or the number of iterations condition; in actual applications, it can be selected according to actual needs, and this embodiment does not make any restrictions here.

[0092] Based on this, in order to enable the heterogeneous graph network model to more accurately construct the networking topology vector corresponding to the networking topology data, the initial heterogeneous graph network model can be trained to fully learn the relationship between the nodes in the networking topology data. Specifically, the networking topology data stored in the graph database can be preprocessed first, and the graph nodes, graph edges and networking attributes can be determined according to the preprocessing results; on this basis, the networking knowledge graph can be constructed based on the graph nodes, graph edges and networking attributes; the relationship and attributes between various resources in the cloud computing project can be reflected through the networking knowledge graph. Then, the initial heterogeneous graph network model containing the adaptive attention mechanism can be trained using the networking knowledge graph, so that the model can fully learn the relationship between the nodes, and on this basis, the mapping relationship learning from data to vectors can be completed to obtain a heterogeneous graph network model that meets the training stop conditions, and it can be deployed in the project scenario for the recommendation of the networking structure.

[0093] In practical applications, after obtaining the graph database that stores networking graph data, considering that the stored data is stored based on the networking structure, and in order to enable the heterogeneous graph network model to have the ability to perform vector conversion on the networking topology data in the graph database, the networking topology data stored in the graph database can be used to construct a cloud network knowledge graph, so as to reflect the resource relationships and attributes at a fine-grained level through the knowledge graph, where nodes can represent resources, edges represent the relationships between resources and additional attributes, such as configuration details and traffic data.

[0094] In the process of building the knowledge graph, the network topology data can be preprocessed, including but not limited to data cleaning, standardization, outlier processing, etc., to ensure the accuracy and consistency of the graph. After that, the perfect knowledge graph can be used to build a heterogeneous graph network model.

[0095] In specific implementation, a heterogeneous graph network model that introduces the adaptive attention mechanism AAM can be constructed. According to the collected networking topology data, the node types of the heterogeneous graph, such as servers, routers, switches, etc., are defined, and the edge types, such as physical connections, logical connections, subordinate relationships, etc., and the corresponding weights represent the strength and importance of the connection. A feature vector is created for each node type, which contains the key attribute information of the device, and the feature can be numerical or categorical, such as device model, configuration parameters, etc. This allows the model to accurately complete the vector mapping process.

[0096] The embedded vectors of all nodes and edges can be learned through the trained heterogeneous graph network model. By inserting the embedded vectors into the corresponding vector database, an initial topological vector database can be constructed, so that the topological vector data recorded in the database can be directly used in the network structure recommendation stage.

[0097] In the specific implementation, it is considered that the network topology data associated with the cloud computing project will change over time. For example, a new user has established a new network structure. On this basis, in order to ensure that the subsequent network structure recommendation for other users can be completed in combination with the network topology data with better real-time performance, the topology vector database can be updated when the new network topology data is added. In this embodiment, the specific implementation method is as follows:

[0098] At a preset time node, detect whether the cloud computing project has new networking topology data; if so, use the heterogeneous graph network model to process the new networking topology data to obtain a new networking topology vector; and incrementally store the new networking topology vector in the topology vector database.

[0099] Specifically, the preset time node refers to a node for detecting whether there is newly added networking topology data for the cloud computing vector, and the time node can also be used as a time node for reusing the heterogeneous graph network model to construct the networking topology vector for the networking topology data, so that the networking topology vector in the topology vector database can ensure its real-time performance, thereby improving the accuracy of the network structure recommendation. Correspondingly, the newly added networking topology data specifically refers to the newly added networking topology data in the cloud computing project, and the networking topology vector has not yet been constructed. Correspondingly, the newly added networking topology vector specifically refers to the vector expression obtained after processing the newly added networking topology data using the heterogeneous graph network model.

[0100] Based on this, considering that cloud computing projects provide cloud computing services to a large number of users, and network topology data may be generated at any time, in order to achieve a stronger real-time performance of the topology vector database, it is possible to detect whether there is new network topology data in the cloud computing project at a preset time node; if so, it means that there is network topology data that has not been added to the topology vector database, so the heterogeneous graph network model can be used to process the new network topology data to obtain the new network topology vector; then the new network topology vector is incrementally stored in the topology vector database to achieve real-time updates.

[0101] In practical applications, the heterogeneous graph network model can calculate the embedding vectors of all nodes and edges at a set time to update the topological vector database. Through this operation, the topological vector database can more accurately reflect the dynamic changes of the network environment and provide real-time, embedding-based network features for the recommendation system, thereby achieving more accurate networking recommendations.

[0102] For example, in order to recommend a network structure that better meets the service needs of users using cloud computing services, cloud computing project party A will collect the network topology data of all users, and then store the network topology data in a unified manner in the Neo4J graph database. By defining nodes to represent resources, edges to represent the relationship between resources and additional attributes, a cloud network knowledge graph is constructed. In the process of building the cloud network knowledge graph, the network topology data will be cleaned, standardized, and processed for outliers to obtain a cloud network knowledge graph with higher quality. After that, the cloud network knowledge graph can be used to train the HetGAT model, and during the training process, the node types of the heterogeneous graph, such as servers, routers, switches, etc., are defined, and the edge types, such as physical connections, logical connections, etc., are defined. After the HetGAT model is trained, the network topology data corresponding to each user can be input into the model for processing, and then the network topology vector corresponding to each user can be obtained. After obtaining the network topology vectors corresponding to the network structures of all users, by storing them in the topology vector database, the network structure recommendation for other users can be realized in the application stage.

[0103] In addition, cloud computing project party A can also set a specified time node and use the HetGAT model to recalculate the embedding vectors for all nodes and edges. For example, at 1 a.m. every day, the network topology vector corresponding to the network topology data can be rebuilt, so that the network topology vector in the database can fully reflect the dynamic changes of the network environment, so that cloud computing project party A can make more accurate network structure recommendations to other users.

[0104] In summary, by regularly detecting and constructing new networking topology data, the real-time performance of the networking topology vectors in the topology vector database can be made stronger, and thus high-precision networking structure recommendations can be completed in the application stage.

[0105] Step S204: when the networking description information is received, input the networking description information into a networking recommendation model, wherein the networking recommendation model includes an information processing unit and a retrieval unit.

[0106] Specifically, when the network description information is received, it means that there is a user who needs to participate in the cloud computing project and has put forward his own corresponding networking requirements. On this basis, in order to be able to recommend a more accurate networking structure to the user and meet the user's networking expectations, the network description information can be input into a networking recommendation model including an information processing unit and a retrieval unit, so that the network recommendation model can be used to complete the user's networking demand intention understanding and networking recommendation operations, thereby improving the interpretability of the network recommendation and ensuring the recommendation accuracy.

[0107] The network description information specifically refers to the network requirement description information submitted by the user when the user has a network requirement, which may include the user's networking intention and service requirements, etc., and is used for the network recommendation model to understand and complete the network recommendation operation. Correspondingly, the network recommendation model specifically refers to a large language model including an information processing unit and a retrieval unit, wherein the information processing unit is used to understand the network description information, and the retrieval unit is used to retrieve the network topology vector according to the information understanding result, and is used to find the network topology vector that is closer to the user's needs from the topology vector database for recommendation.

[0108] Furthermore, after the networking description information is input into the networking recommendation model, considering that the networking description information may not clearly explain the networking requirements or is incomplete, and in order to ensure that the networking recommendation model has a stronger recommendation effect, multiple rounds of interaction can be used to supplement the description information, and based on this, the target networking description information is formed for model processing. In this embodiment, the specific implementation method is as follows:

[0109] When it is detected that the networking description information does not meet the networking recommendation conditions, a multi-round interaction method is used to obtain supplementary description information associated with the networking description information; target networking description information is constructed based on the networking description information and the supplementary description information, the target networking description information is used as the networking description information, and the step of inputting the networking description information into the networking recommendation model is executed.

[0110] Specifically, the network recommendation condition refers to the condition of detecting whether the network description information is complete and being able to recommend the network structure to the user based on the network description information. Correspondingly, the multi-round interaction refers to a mechanism for collecting more network demand information in a question-and-answer manner to obtain supplementary description information. Correspondingly, the supplementary description information refers to the user continuing to submit auxiliary information about the network demand description on the basis of submitting the network description information. Correspondingly, the target network description information is the global network description information obtained by combining the network description information and the supplementary network description information.

[0111] Based on this, after receiving the networking description information, considering that the networking description information may not fully reflect the user's networking needs, the networking description information can be checked each time it is received. If it is detected that the networking description information does not meet the networking recommendation conditions, it means that the currently submitted networking description information is incomplete. Therefore, multiple rounds of interaction can be used to obtain supplementary description information of the associated networking description information; thereafter, the target networking description information can be constructed based on the networking description information and the supplementary description information, and then the target networking description information can be used as the networking description information, and the step of inputting the networking description information into the networking recommendation model can be executed.

[0112] In summary, by adopting a multi-round interactive method to determine the target networking description information, it can be ensured that the target networking description information fully represents the user's networking needs. After inputting it into the networking recommendation model, the networking recommendation model can understand the information from multiple dimensions to achieve accurate networking structure recommendation.

[0113] Step S206: Process the networking description information through the information processing unit to obtain a networking description vector, and use the retrieval unit to retrieve a target networking topology vector that matches the networking description vector in the topology vector database.

[0114] Specifically, after the network description information is input into the network recommendation model, in order to complete the matching network structure screening in the vector space, the network description information can be first processed by the information processing unit to obtain the network description vector corresponding to the network description information. After obtaining the network description vector, since the network topology vector in the topology vector database is also expressed as a vector, the retrieval unit can be used to directly retrieve the target network topology vector that matches the network description vector in the topology vector database, and then determine the target network topology vector that matches the user's networking needs. Subsequently, the target network topology vector can be processed to obtain more interpretable graphic information that is more convenient for users to understand the network structure, so as to improve the user's willingness to accept.

[0115] The network description vector specifically refers to the vector expression obtained after semantic analysis of the network description information. Correspondingly, the target network topology vector specifically refers to the network topology vector that is retrieved from the topology vector database by the retrieval unit and has a higher matching degree with the network description information. Its representation is closer to the user's networking needs. After determining its corresponding network structure, it is recommended to the user, which can make it easier for the user to understand whether the current network structure meets their own usage needs.

[0116] Furthermore, when converting the networking description information into the networking description vector, in order to fully understand the networking intention and networking requirements, the networking description information may be split first and then semantic analysis may be performed. In this embodiment, the specific implementation is as follows:

[0117] The information processing unit is used to split the networking description information to obtain an information sequence; the information embedding layer in the information processing unit is used to process the information sequence to obtain an information embedding vector; and semantic analysis is performed on the information embedding vector to obtain a networking description vector corresponding to the networking description information.

[0118] Specifically, the information sequence refers to a sequence of words, short sentences, and character units obtained after text splitting of the network description information. By processing the information sequence, the information embedding vectors corresponding to the words, short sentences, and character units can be obtained. Then, semantic analysis is performed based on this to obtain the network description vector corresponding to the network description information.

[0119] Based on this, after the networking description information is input into the networking recommendation model, in order to achieve a more accurate component recommendation operation in the vector space, the networking description information can be first split through the information processing unit to obtain an information sequence composed of words, short sentences, and character units; then the information sequence can be processed using the information embedding layer in the information processing unit to obtain the corresponding information embedding vectors; on this basis, by performing semantic analysis on the information embedding vectors, the networking description vector corresponding to the networking description information can be obtained for subsequent retrieval and use.

[0120] In summary, by performing semantic analysis on the networking description information and constructing a networking description vector, it is possible to fully reflect the networking requirements through the networking description vector, thereby facilitating improved retrieval accuracy in subsequent retrieval.

[0121] Furthermore, after the network description information is converted into a network description vector, the structure alignment of the network topology data and the network description information in the vector space can be achieved, and then the retrieval of the target network topology vector can be completed by calculating the similarity through the retrieval unit. In this embodiment, the specific implementation method is as follows:

[0122] The retrieval unit calculates the networking similarity between the networking description vector and the networking topology vector in the topology vector database; a set number of candidate networking topology vectors are screened in the topology vector database according to the networking similarity, and a candidate networking topology vector list is formed; the candidate networking topology vector list is filtered based on the recommendation rules preset by the cloud computing project, and a target networking topology vector is determined according to the filtering result.

[0123] Specifically, the network similarity specifically refers to calculating the similarity between the network description vector and the network topology vector, and the similarity calculation can be completed using cosine similarity. Correspondingly, the candidate network topology vector specifically refers to a plurality of network topology vectors with higher network similarity screened out from the topology vector database. Correspondingly, the candidate network topology vector list specifically refers to a list obtained by sorting a plurality of candidate network topology vectors according to the similarity with the network description vector. Correspondingly, the recommendation rule specifically refers to a rule composed of multi-dimensional settings such as service rules, cost-effectiveness, and performance indicators, which is used to screen out target network topology vectors that meet the needs from the candidate network topology vector list. In practical applications, the target network topology vector can be one or more, so that users can choose according to their needs, thereby improving the diversity of recommendations, and the probability of covering user networking needs is higher.

[0124] Based on this, after obtaining the network description vector corresponding to the network description information, the network similarity between the network description vector and each network topology vector in the topology vector database can be calculated by the retrieval unit; at this time, the network similarity is compared with the preset similarity threshold, so as to select the network topology vector greater than the similarity threshold as the candidate network topology vector according to the comparison result, or directly select a set number of network topology vectors as the candidate network topology vector according to the similarity. Then, the network topology vectors can be sorted according to the similarity and a list of candidate network topology vectors can be formed.

[0125] On this basis, in order to make the final target networking topology vector more in line with the user's networking needs, the candidate networking topology vector list can be filtered based on the recommendation rules preset by the cloud computing project to eliminate unnecessary networking topology vectors and obtain one or more target networking topology vectors to facilitate subsequent networking recommendation operations.

[0126] In practical applications, the network recommendation model can use Tokenization technology to convert network description information into words and phrases, and then convert words and phrases into network description vectors through the Word2Vec text embedding model, so that the network recommendation model can understand semantics.

[0127] Furthermore, the network recommendation model can analyze the network description vector and identify key intentions and requirements, such as service scale, type, data flow, security, and cost budget. In this process, if the network description information entered by the user is incomplete or vague, the network recommendation model can use multiple rounds of interaction to ask more specific questions, such as the expected number of concurrent users, geographical distribution, and whether specific network services (such as load balancing or backup) are required, to clarify the user's networking needs.

[0128] Furthermore, after the networking recommendation model understands the user's networking needs, it can extract key points directly related to the network, and this information can be used as the basis for subsequent matching and recommendation. At this time, the user's networking needs and networking topology data have been converted into vector expressions, which allows the two to be compared and matched using the same algorithm. Therefore, the networking recommendation model can search in the topology vector database, calculate the cosine similarity between the networking description vector and each networking topology vector, and generate a sorted list based on the similarity score. It is used to record the networking topology vectors that are close to the user's networking needs. Thereafter, the networking recommendation model can also combine preset recommendation rules, such as cost-effectiveness and performance indicators to conduct a secondary evaluation of the networking topology vectors contained in the list, and finally, one or more networking topology vectors that meet the user's networking needs can be screened out according to the evaluation results to achieve the networking solution that recommends the user to his needs.

[0129] Using the above example, after receiving the network description information submitted by the user, the cloud computing project party A can input the network requirement description information into the LLM model for network recommendation. The network requirement description information can be processed by the LLM model to obtain its corresponding vector expression VT. On this basis, the LLM model can be used to calculate the cosine similarity between the vector expression VT of the network requirement description information and the network topology vector stored in the topology vector database. According to the calculation results, a list of candidate network topology vectors sorted from large to small by cosine similarity is obtained, which contains n candidate network topology vectors. On this basis, the LLM model is used to perform a secondary evaluation of the n candidate network topology vectors according to service rules, cost-effectiveness, and performance indicators. According to the evaluation results, it is determined that the network topology vector 1 better matches the user's networking needs, so its corresponding networking solution can be used as the networking solution recommended to the user.

[0130] In summary, by screening the target networking topology vector by calculating the vector similarity, it can be ensured that the screened target networking topology vector better matches the user's networking needs, thereby effectively improving the user's participation experience in cloud computing projects.

[0131] Step S208: determining a target networking structure according to the target networking topology vector, and constructing networking structure graphic information corresponding to the target networking structure.

[0132] Specifically, after obtaining the target networking topology vector as described above, the target networking structure can be determined according to the target networking topology vector to determine the networking structure solution recommended to the user. On this basis, in order to make the recommendation of the target networking structure more explainable and easier to be accepted by the user, the networking structure graphic information corresponding to the target networking structure can be reconstructed. Through the networking structure graphic information, the user can understand the reason why the target networking structure is recommended through text description and image expression, thereby improving the user's acceptance of the networking recommendation.

[0133] The target network structure specifically refers to the network structure corresponding to the target network topology vector. Correspondingly, the network structure graphic information specifically refers to the information constructed by combining the description text and topology diagram of the target network structure, which is used to explain and express the target network structure through text and image representation.

[0134] Furthermore, when constructing the network structure graphic information corresponding to the target network structure, in order to improve the interpretability of the target network structure and thus improve the user's willingness to accept, the network structure description information and the network structure topology diagram can be combined for construction. In this embodiment, the specific implementation method is as follows:

[0135] The network structure description information corresponding to the target network structure is generated by the information generation unit in the network recommendation model, and a preset visualization component is called to construct a network structure topology map for the target network structure; based on the network structure description information and the network structure topology map, the network structure graphic information corresponding to the target network structure is constructed.

[0136] Specifically, the information generation unit specifically refers to a unit in the network recommendation model that generates text description information for the target network structure. Accordingly, the network structure description information specifically refers to information that describes the target network structure in natural language. Accordingly, the visualization component specifically refers to a component that constructs a network structure topology diagram for the target network structure, and accordingly, the network structure topology diagram is the network topology diagram corresponding to the target network structure.

[0137] Based on this, in order to be able to represent the overall details of the target networking structure from multi-dimensional information, the network structure description information corresponding to the target networking structure can be generated by the information generation unit in the networking recommendation model, and the preset visualization component can be called to construct a network structure topology map for the target networking structure; thereafter, the network structure graphic information corresponding to the target networking structure can be constructed based on the network structure description information and the network structure topology map, which can be used for feedback to users for viewing.

[0138] In summary, by combining the network structure description information and the network structure topology diagram to construct the network structure graphic information, it is possible to reflect the performance characteristics of the target network structure and other information through text description and network topology diagram, thereby making the target network structure more interpretable.

[0139] In addition, considering that there is a deviation between the networking structure graphic information directly recommended to the user and the user's networking intention, in order to correct the deviation, the user can also be supported to provide feedback on adjustment information, and then further adjustments can be made based on the user's adjustment feedback information. In this embodiment, the specific implementation method is as follows:

[0140] Upon receiving adjustment feedback information submitted for the networking structure graphic information, the adjustment feedback information is input into the networking recommendation model; the adjustment feedback information is processed by the information processing unit to obtain an adjustment feedback vector, and the networking description vector is updated to a target networking description vector using the adjustment feedback vector; the target networking description vector is used as the networking description vector, and the step of using the retrieval unit to retrieve a target networking topology vector matching the networking description vector in the topology vector database is performed.

[0141] Specifically, the adjustment feedback information specifically refers to the adjustment information submitted again by the user after the target networking structure is recommended to the user. Correspondingly, the adjustment feedback vector specifically refers to the vector expression of the corresponding adjustment feedback information. Correspondingly, the target networking description vector specifically refers to the vector obtained by integrating the vector expression corresponding to the adjustment feedback information on the basis of the networking description vector, and is used to realize the re-recommendation of the networking structure according to the user's adjustment requirements. Among them, the generation of the adjustment feedback vector can refer to the generation of the networking description vector in the above embodiment, and this embodiment does not make any limitation here.

[0142] Based on this, after feeding back the network structure graphic information to the user, if the adjustment feedback information submitted for the network structure graphic information is received, it means that the currently matched target network structure does not meet the user's networking needs, so a new solution can be screened according to the user's adjustment feedback information. At this time, the adjustment feedback information can be input into the network recommendation model; the adjustment feedback information is processed by the information processing unit to obtain the adjustment feedback vector, and then it can be fused to the network description vector to obtain the target network description vector according to the fusion result; at this time, the target network description vector can be used as the network description vector, and the step of using the retrieval unit to retrieve the target network topology vector matching the network description vector in the topology vector database is performed. A new round of network recommendation operations is implemented. Thereby ensuring that the final network structure is more in line with the user's networking needs.

[0143] In actual applications, the network recommendation model can call pre-designed topology visualization tools to convert the recommended target network structure into an easy-to-understand graphical representation. The various components in the target network structure, such as servers, routers, switches, etc., as well as the connection paths and logical relationships between the above components are represented through graphical representations. This makes it easier for users to understand and evaluate the recommended networking solutions.

[0144] Furthermore, when users submit feedback, considering that the user's feedback may be a change in the demand for a specific component, or the emphasis on a specific function, such as the user's need to add redundant backup, or adjust the network's load balancing strategy, the networking recommendation model can convert it into a new or adjusted adjustment vector based on the feedback, and then match it with the information in the topology vector database again. In this way, a new round of networking structure matching and recommendation is carried out based on the user's feedback, so that the recommended structure is more in line with the user's networking needs.

[0145] Continuing with the above example, after determining that networking topology vector 1 better matches the user's networking needs, networking solution 1 corresponding to networking topology vector 1 can be determined. On this basis, the networking solution description information and the network topology map corresponding to networking solution 1 can be generated through the LLM model, so that the networking solution description information and the network topology map are fed back to the user at the same time, so that the user can understand the characteristics of the networking solution.

[0146] The information processing method provided in this embodiment can firstly process the network topology data associated with the cloud computing project into a network topology vector by using a heterogeneous graph network model, and store the network topology vector in a topology vector database in order to more accurately recommend the network structure to the user, and to be more explanatory and convenient for the user to understand the network distribution. The method is used to record the vector expression of all network topology data through the topology vector database, so that the user can choose from it when recommending the network structure to the user. Then, when the network description information is received, in order to improve the efficiency and accuracy of the network recommendation, the network description information can be first input into the network recommendation model, wherein the network recommendation model includes an information processing unit and a retrieval unit; on this basis, the network description information can be processed by the information processing unit to obtain the network description vector, and after the network description information is converted into a vector expression, the retrieval unit can be used to retrieve the target network topology vector matching the network description vector in the topology vector database; the target network topology vector can be considered as the vector expression corresponding to the network structure that better matches the network description information. Therefore, the target networking structure can be determined according to the target networking topology vector, and the networking structure graphic information corresponding to the target networking structure can be constructed, so as to improve the interpretability of the currently recommended target networking structure, thereby increasing the user's willingness to accept it. At the same time, reflecting the target networking structure through graphic information can effectively reduce the complexity of network understanding and accelerate the popularization of simple and easy-to-use intelligent cloud networks.

[0147] See also Figure 3 , Figure 3 A flowchart of a networking recommendation method provided according to an embodiment of the present specification is shown. The method is applied to a cloud-side device and specifically includes the following steps.

[0148] Step S302, receiving the network description information submitted by the end-side device, and inputting the network description information into a network recommendation model, wherein the network recommendation model includes an information processing unit and a retrieval unit;

[0149] Step S304, processing the networking description information by the information processing unit to obtain a networking description vector, and using the retrieval unit to retrieve a target networking topology vector matching the networking description vector in a topology vector database;

[0150] Step S306, determining a target networking structure according to the target networking topology vector, and constructing networking structure graphic information corresponding to the target networking structure;

[0151] Step S308: Send the networking structure graphic information to the terminal side device.

[0152] The above is a schematic scheme of a networking recommendation method of this embodiment. It should be noted that the technical scheme of the networking recommendation method and the technical scheme of the above information processing method belong to the same concept, and the details not described in detail in the technical scheme of the networking recommendation method can be referred to the description of the technical scheme of the above information processing method.

[0153] See also Figure 4 , Figure 4 A flowchart of a model training method provided according to an embodiment of the present specification is shown, which specifically includes the following steps.

[0154] Step S402, obtaining network configuration information corresponding to users participating in the cloud computing project;

[0155] Step S404, determining the network topology data corresponding to the user according to the network configuration information, and storing the network topology data in a graph database;

[0156] Step S406, preprocessing the networking topology data stored in the graph database, and constructing a networking knowledge graph according to the preprocessing result;

[0157] Step S408, using the networked knowledge graph to train an initial heterogeneous graph network model including an adaptive attention mechanism until a heterogeneous graph network model that meets the training stop condition is obtained, wherein the heterogeneous graph network model is applied to the above-mentioned information processing method.

[0158] The above is a schematic scheme of a model training method of this embodiment. It should be noted that the technical scheme of the model training method and the technical scheme of the above information processing method belong to the same concept, and the details not described in detail in the technical scheme of the model training method can be referred to the description of the technical scheme of the above information processing method.

[0159] The following combination Figure 5 , taking the networking recommendation method provided in this specification as an example of applying the networking recommendation method in a cloud computing scenario, the networking recommendation method is further described. Figure 5 A processing flow chart of a networking recommendation method provided by an embodiment of this specification is shown, which specifically includes the following steps.

[0160] Step S502: obtaining network configuration information corresponding to users participating in the cloud computing project.

[0161] Step S504: determine the network topology data corresponding to the user according to the network configuration information, and store the network topology data in a graph database.

[0162] Step S506: input the networking topology data stored in the graph database into the heterogeneous graph network model.

[0163] Step S508, extracting target nodes and target edges from the networking topology data through a heterogeneous graph network model, and constructing a target node embedding vector corresponding to the target node and a target edge embedding vector corresponding to the target edge.

[0164] Among them, the construction of the target node embedding vector includes: determining the neighbor nodes corresponding to the target node, and constructing the initial node embedding vector corresponding to the target node and the neighbor node embedding vector corresponding to the neighbor node; calculating the attention score between the target node and the neighbor node according to the initial node embedding vector and the neighbor node embedding vector through the dot product attention mechanism in the heterogeneous graph network model; generating the target node weight of the target node and the neighbor node weight of the neighbor node according to the attention score; performing weighted sum calculation on the initial node embedding vector and the target node weight, as well as the neighbor node embedding vector and the neighbor node weight, to obtain the target node embedding vector corresponding to the target node.

[0165] Step S510, normalizing the target node embedding vector and the target edge embedding vector, generating a networking topology vector corresponding to the networking topology data according to the normalization result, and storing the networking topology vector in a topology vector database.

[0166] Step S512: upon receiving the networking description information submitted by the client, input the networking description information into a networking recommendation model, wherein the networking recommendation model includes an information processing unit and a retrieval unit.

[0167] Step S514: split the networking description information by the information processing unit to obtain an information sequence, and process the information sequence by using the information embedding layer in the information processing unit to obtain an information embedding vector.

[0168] Step S516: perform semantic analysis on the information embedding vector to obtain a networking description vector corresponding to the networking description information.

[0169] Step S518: Calculate the networking similarity between the networking description vector and the networking topology vector in the topology vector database through the retrieval unit.

[0170] Step S520 , screening a set number of candidate networking topology vectors in the topology vector database according to the networking similarity, and forming a candidate networking topology vector list.

[0171] Step S522: filtering the candidate networking topology vector list based on the recommendation rules preset in the cloud computing project, and determining the target networking topology vector according to the filtering result.

[0172] Step S524, determine the target networking structure according to the target networking topology vector, generate the networking structure description information corresponding to the target networking structure through the information generation unit in the networking recommendation model, and call the preset visualization component to construct a networking structure topology diagram for the target networking structure.

[0173] Step S526: constructing network structure graphic information corresponding to the target network structure based on the network structure description information and the network structure topology diagram, and sending the network structure graphic information to the client.

[0174] In summary, in order to recommend a more accurate network structure to users, and to make it more explanatory and convenient for users to understand the network distribution, the heterogeneous graph network model can be used to process the network topology data associated with the cloud computing project into a network topology vector, and the network topology vector can be stored in a topology vector database; the vector expression of all network topology data can be recorded through the topology vector database, so that users can choose from it when recommending a network structure to users. Then, when the network description information is received, in order to improve the efficiency and accuracy of the network recommendation, the network description information can be input into the network recommendation model, wherein the network recommendation model includes an information processing unit and a retrieval unit; on this basis, the network description information can be processed by the information processing unit to obtain the network description vector, and after the network description information is converted into a vector expression, the retrieval unit can be used to retrieve the target network topology vector matching the network description vector in the topology vector database; the target network topology vector can be considered as the vector expression corresponding to the network structure that better matches the network description information. Therefore, the target networking structure can be determined according to the target networking topology vector, and the networking structure graphic information corresponding to the target networking structure can be constructed, so as to improve the interpretability of the currently recommended target networking structure, thereby increasing the user's willingness to accept it. At the same time, reflecting the target networking structure through graphic information can effectively reduce the complexity of network understanding and accelerate the popularization of simple and easy-to-use intelligent cloud networks.

[0175] Corresponding to the above method embodiment, this specification also provides an information processing device embodiment, Figure 6 FIG. 1 shows a schematic diagram of the structure of an information processing device provided by an embodiment of the present specification. Figure 6 As shown, the device comprises:

[0176] The processing module 602 is configured to process the networking topology data associated with the cloud computing project into a networking topology vector using a heterogeneous graph network model, and store the networking topology vector in a topology vector database;

[0177] An input module 604 is configured to input the network description information into a network recommendation model when receiving the network description information, wherein the network recommendation model includes an information processing unit and a retrieval unit;

[0178] A retrieval module 606 is configured to process the networking description information through the information processing unit to obtain a networking description vector, and use the retrieval unit to retrieve a target networking topology vector matching the networking description vector in the topology vector database;

[0179] The construction module 608 is configured to determine a target networking structure according to the target networking topology vector, and construct networking structure graphic information corresponding to the target networking structure.

[0180] In an optional embodiment, the processing module 602 is further configured to: obtain network configuration information corresponding to users participating in the cloud computing project; determine the network topology data corresponding to the user based on the network configuration information, and store the network topology data in a graph database; use a heterogeneous graph network model to process the network topology data stored in the graph database to obtain a network topology vector corresponding to the network topology data.

[0181] In an optional embodiment, the device also includes: a model training module, configured to preprocess the networking topology data stored in the graph database, determine the graph nodes, graph edges and networking attributes based on the preprocessing results; construct a networking knowledge graph based on the graph nodes, the graph edges and the networking attributes; and use the networking knowledge graph to train an initial heterogeneous graph network model containing an adaptive attention mechanism until a heterogeneous graph network model that meets the training stop conditions is obtained.

[0182] In an optional embodiment, the processing module 602 is further configured to: input the networking topology data stored in the graph database into the heterogeneous graph network model; extract target nodes and target edges from the networking topology data through the heterogeneous graph network model, and construct a target node embedding vector corresponding to the target node and a target edge embedding vector corresponding to the target edge; perform normalization processing on the target node embedding vector and the target edge embedding vector, and generate a networking topology vector corresponding to the networking topology data according to the normalization processing result.

[0183] In an optional embodiment, the processing module 602 is further configured to: determine the neighbor nodes corresponding to the target node, and construct an initial node embedding vector corresponding to the target node and a neighbor node embedding vector corresponding to the neighbor node; calculate the attention score between the target node and the neighbor node according to the initial node embedding vector and the neighbor node embedding vector through the dot product attention mechanism in the heterogeneous graph network model; generate the target node weight of the target node and the neighbor node weight of the neighbor node according to the attention score; perform weighted sum calculation on the initial node embedding vector and the target node weight, as well as the neighbor node embedding vector and the neighbor node weight to obtain the target node embedding vector corresponding to the target node.

[0184] In an optional embodiment, the retrieval module 606 is further configured to: perform information splitting on the networking description information through the information processing unit to obtain an information sequence; process the information sequence using the information embedding layer in the information processing unit to obtain an information embedding vector; perform semantic analysis on the information embedding vector to obtain a networking description vector corresponding to the networking description information.

[0185] In an optional embodiment, the device also includes: an interaction module, which is configured to, when it is detected that the networking description information does not meet the networking recommendation conditions, obtain supplementary description information associated with the networking description information through multiple rounds of interaction; construct target networking description information based on the networking description information and the supplementary description information, use the target networking description information as the networking description information, and execute the step of inputting the networking description information into the networking recommendation model.

[0186] In an optional embodiment, the retrieval module 606 is further configured to: calculate the networking similarity between the networking description vector and the networking topology vector in the topology vector database through the retrieval unit; filter a set number of candidate networking topology vectors in the topology vector database according to the networking similarity, and form a list of candidate networking topology vectors; filter the list of candidate networking topology vectors based on the recommendation rules preset by the cloud computing project, and determine the target networking topology vector according to the filtering results.

[0187] In an optional embodiment, the construction module 608 is further configured to: generate network structure description information corresponding to the target network structure through the information generation unit in the network recommendation model, and call a preset visualization component to construct a network structure topology map for the target network structure; based on the network structure description information and the network structure topology map, construct the network structure graphic information corresponding to the target network structure.

[0188] In an optional embodiment, the device also includes: an adjustment module, configured to input the adjustment feedback information into the networking recommendation model upon receiving adjustment feedback information submitted for the networking structure graphic information; process the adjustment feedback information by the information processing unit to obtain an adjustment feedback vector, and use the adjustment feedback vector to update the networking description vector to a target networking description vector; use the target networking description vector as the networking description vector, and execute the step of using the retrieval unit to retrieve the target networking topology vector that matches the networking description vector in the topology vector database.

[0189] In an optional embodiment, the device also includes: an update module, configured to detect whether there is new networking topology data in the cloud computing project at a preset time node; if so, use the heterogeneous graph network model to process the new networking topology data to obtain a new networking topology vector; and store the new networking topology vector incrementally in the topology vector database.

[0190] The information processing device provided in this embodiment can first use a heterogeneous graph network model to process the network topology data associated with the cloud computing project into a network topology vector, and store the network topology vector in a topology vector database in order to more accurately recommend the network structure to the user, and to be more explanatory and convenient for the user to understand the network distribution. The network topology data associated with the cloud computing project can be processed into a network topology vector, and the network topology vector can be stored in a topology vector database; the vector expression of all network topology data can be recorded through the topology vector database, so that the network structure can be selected when the user is recommended. Then, when the network description information is received, in order to improve the efficiency and accuracy of the network recommendation, the network description information can be first input into the network recommendation model, wherein the network recommendation model includes an information processing unit and a retrieval unit; on this basis, the network description information can be processed by the information processing unit to obtain the network description vector, and after the network description information is converted into a vector expression, the retrieval unit can be used to retrieve the target network topology vector matching the network description vector in the topology vector database; the target network topology vector can be considered as the vector expression corresponding to the network structure that better matches the network description information. Therefore, the target networking structure can be determined according to the target networking topology vector, and the networking structure graphic information corresponding to the target networking structure can be constructed, so as to improve the interpretability of the currently recommended target networking structure, thereby increasing the user's willingness to accept it. At the same time, reflecting the target networking structure through graphic information can effectively reduce the complexity of network understanding and accelerate the popularization of simple and easy-to-use intelligent cloud networks.

[0191] The above is a schematic scheme of an information processing device of this embodiment. It should be noted that the technical scheme of the information processing device and the technical scheme of the above information processing method belong to the same concept, and the details not described in detail in the technical scheme of the information processing device can be referred to the description of the technical scheme of the above information processing method.

[0192] Corresponding to the above method embodiment, this specification also provides a networking recommended device embodiment. Figure 7 FIG. 1 is a schematic diagram showing the structure of a network recommendation device provided by an embodiment of the present specification. Figure 7 As shown, the device is applied to a cloud-side device, including:

[0193] The information receiving module 702 is configured to receive the network description information submitted by the terminal device and input the network description information into the network recommendation model, wherein the network recommendation model includes an information processing unit and a retrieval unit;

[0194] The information processing module 704 is configured to process the networking description information through the information processing unit to obtain a networking description vector, and use the retrieval unit to retrieve a target networking topology vector matching the networking description vector in a topology vector database;

[0195] The information construction module 706 is configured to determine a target networking structure according to the target networking topology vector, and construct the networking structure graphic information corresponding to the target networking structure;

[0196] The information sending module 708 is configured to send the networking structure graphic information to the terminal side device.

[0197] The above is a schematic scheme of a networking recommendation device of this embodiment. It should be noted that the technical scheme of the networking recommendation device and the technical scheme of the networking recommendation method described above are of the same concept, and the details not described in detail in the technical scheme of the networking recommendation device can be found in the description of the technical scheme of the networking recommendation method described above.

[0198] Corresponding to the above method embodiment, this specification also provides a model training device embodiment, Figure 8 FIG. 1 shows a schematic diagram of a model training device provided by an embodiment of the present specification. Figure 8 As shown, the device comprises:

[0199] The information acquisition module 802 is configured to acquire the network configuration information corresponding to the user participating in the cloud computing project;

[0200] A data determination module 804 is configured to determine the network topology data corresponding to the user according to the network configuration information, and store the network topology data in a graph database;

[0201] The data processing module 806 is configured to pre-process the networking topology data stored in the graph database and construct a networking knowledge graph according to the pre-processing result;

[0202] The training model module 808 is configured to use the networked knowledge graph to train an initial heterogeneous graph network model including an adaptive attention mechanism until a heterogeneous graph network model that meets the training stop condition is obtained, wherein the heterogeneous graph network model is applied to the above-mentioned information processing method.

[0203] The above is a schematic scheme of a model training device of this embodiment. It should be noted that the technical scheme of the model training device and the technical scheme of the above-mentioned model training method belong to the same concept, and the details not described in detail in the technical scheme of the model training device can be referred to the description of the technical scheme of the above-mentioned model training method.

[0204] Fig. 9 The structure block diagram of a computing device 900 provided according to an embodiment of the present specification is shown. The components of the computing device 900 include but are not limited to a memory 910 and a processor 920. The processor 920 is connected to the memory 910 via a bus 930, and a database 950 is used to store data.

[0205] The computing device 900 also includes an access device 940 that enables the computing device 900 to communicate via one or more networks 960. Examples of these networks include a public switched telephone network (PSTN), a local area network (LAN), a wide area network (WAN), a personal area network (PAN), or a combination of communication networks such as the Internet. The access device 940 may include one or more of any type of network interface (e.g., a network interface card (NIC)) that is wired or wireless, such as an IEEE 802.11 wireless local area network (WLAN) wireless interface, a world-wide interoperability for microwave access (Wi-MAX) interface, an Ethernet interface, a universal serial bus (USB) interface, a cellular network interface, a Bluetooth interface, and a near field communication (NFC).

[0206] In one embodiment of the present specification, the above components of the computing device 900 and Fig. 9 Other components not shown in the figure may also be connected to each other, for example, via a bus. It should be understood that Fig. 9 The computing device structure block diagram shown is only for the purpose of illustration, and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0207] The computing device 900 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., a tablet computer, a personal digital assistant, a laptop computer, a notebook computer, a netbook, etc.), a mobile phone (e.g., a smart phone), a wearable computing device (e.g., a smart watch, smart glasses, etc.), or other types of mobile devices, or a stationary computing device such as a desktop computer or a personal computer (PC). The computing device 900 may also be a mobile or stationary server.

[0208] Among them, the processor 920 is used to execute the following computer-executable instructions, which, when executed by the processor, implement the steps of the above-mentioned information processing method or networking recommendation method or model training method.

[0209] The above is a schematic scheme of a computing device of this embodiment. It should be noted that the technical scheme of the computing device and the technical scheme of the above-mentioned information processing method or networking recommendation method or model training method belong to the same concept, and the details not described in detail in the technical scheme of the computing device can be referred to the description of the technical scheme of the above-mentioned information processing method or networking recommendation method or model training method.

[0210] An embodiment of the present specification also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the above-mentioned information processing method, networking recommendation method, or model training method.

[0211] The above is a schematic scheme of a computer-readable storage medium of this embodiment. It should be noted that the technical scheme of the storage medium and the technical scheme of the above-mentioned information processing method or networking recommendation method or model training method belong to the same concept, and the details not described in detail in the technical scheme of the storage medium can be referred to the description of the technical scheme of the above-mentioned information processing method or networking recommendation method or model training method.

[0212] An embodiment of the present specification also provides a computer program, wherein when the computer program is executed in a computer, the computer is caused to execute the steps of the above-mentioned information processing method or networking recommendation method or model training method.

[0213] The above is a schematic scheme of a computer program of this embodiment. It should be noted that the technical scheme of the computer program and the technical scheme of the above-mentioned information processing method or networking recommendation method or model training method belong to the same concept, and the details not described in detail in the technical scheme of the computer program can be referred to the description of the technical scheme of the above-mentioned information processing method or networking recommendation method or model training method.

[0214] An embodiment of the present specification also provides a computer program product, including a computer program or instructions, which, when executed by a processor, implements the steps of the above-mentioned information processing method or networking recommendation method or model training method.

[0215] The above is a schematic scheme of a computer program product of this embodiment. It should be noted that the technical scheme of the computer program product and the technical scheme of the above-mentioned information processing method or networking recommendation method or model training method belong to the same concept, and the details not described in detail in the technical scheme of the computer program product can be referred to the description of the technical scheme of the above-mentioned information processing method or networking recommendation method or model training method.

[0216] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0217] The computer instructions include computer program codes, which may be in source code form, object code form, executable files or some intermediate forms, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0218] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of this specification are not limited to the described action sequence, because according to the embodiments of this specification, some steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the embodiments of this specification. In the above embodiments, the descriptions of each embodiment have their own emphasis, and for the parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments. The preferred embodiments of this specification disclosed above are only used to help explain this specification. The optional embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, according to the content of the embodiments of this specification, many modifications and changes can be made. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the embodiments of this specification, so that those skilled in the art can understand and use this specification well. This specification is only limited by the claims and their full scope and equivalents.

Claims

1. An information processing method, comprising: Using a heterogeneous graph network model, the networking topology data associated with the cloud computing project is processed into a networking topology vector, and the networking topology vector is stored in a topology vector database; When receiving the network description information, inputting the network description information into a network recommendation model, wherein the network recommendation model includes an information processing unit and a retrieval unit; Processing the networking description information by the information processing unit to obtain a networking description vector, and using the retrieval unit to retrieve a target networking topology vector matching the networking description vector in the topology vector database; A target networking structure is determined according to the target networking topology vector, and networking structure graphic information corresponding to the target networking structure is constructed.

2. The information processing method according to claim 1, wherein the processing of the network topology data associated with the cloud computing project into a network topology vector using a heterogeneous graph network model comprises: Obtain the network configuration information corresponding to the users participating in the cloud computing project; Determine the network topology data corresponding to the user according to the network configuration information, and store the network topology data in a graph database; The networking topology data stored in the graph database is processed using a heterogeneous graph network model to obtain a networking topology vector corresponding to the networking topology data.

3. The information processing method according to claim 2, wherein before the step of processing the networking topology data stored in the graph database using a heterogeneous graph network model to obtain a networking topology vector corresponding to the networking topology data is performed, the method further comprises: Preprocessing the networking topology data stored in the graph database, and determining graph nodes, graph edges, and networking attributes according to the preprocessing results; Constructing a networking knowledge graph based on the graph nodes, the graph edges and the networking attributes; The networked knowledge graph is used to train an initial heterogeneous graph network model including an adaptive attention mechanism until a heterogeneous graph network model that meets the training stop condition is obtained.

4. The information processing method according to claim 2, wherein the using a heterogeneous graph network model to process the networking topology data stored in the graph database to obtain a networking topology vector corresponding to the networking topology data comprises: Inputting the networking topology data stored in the graph database into the heterogeneous graph network model; Extracting a target node and a target edge from the networking topology data by using the heterogeneous graph network model, and constructing a target node embedding vector corresponding to the target node and a target edge embedding vector corresponding to the target edge; Normalization processing is performed on the target node embedding vector and the target edge embedding vector, and a networking topology vector corresponding to the networking topology data is generated according to the normalization processing result.

5. The information processing method according to claim 4, wherein the step of constructing a target node embedding vector corresponding to the target node comprises: Determine a neighbor node corresponding to the target node, and construct an initial node embedding vector corresponding to the target node and a neighbor node embedding vector corresponding to the neighbor node; Calculating the attention score between the target node and the neighbor node according to the initial node embedding vector and the neighbor node embedding vector through the dot product attention mechanism in the heterogeneous graph network model; Generate a target node weight of the target node and a neighbor node weight of the neighbor node according to the attention score; A weighted sum calculation is performed on the initial node embedding vector and the target node weight, as well as the neighbor node embedding vector and the neighbor node weight to obtain a target node embedding vector corresponding to the target node.

6. The information processing method according to claim 1, wherein the information processing unit processes the network description information to obtain a network description vector, comprising: Splitting the networking description information by the information processing unit to obtain an information sequence; Processing the information sequence using the information embedding layer in the information processing unit to obtain an information embedding vector; A semantic analysis is performed on the information embedding vector to obtain a networking description vector corresponding to the networking description information.

7. The information processing method according to claim 1, after the step of inputting the network description information into the network recommendation model when the network description information is received, further comprising: When it is detected that the networking description information does not meet the networking recommendation condition, obtaining supplementary description information associated with the networking description information by means of multiple rounds of interaction; The target networking description information is constructed based on the networking description information and the supplementary description information, the target networking description information is used as the networking description information, and the step of inputting the networking description information into a networking recommendation model is performed.

8. The information processing method according to claim 1, wherein the using the retrieval unit to retrieve the target network topology vector matching the network description vector in the topology vector database comprises: Calculating the networking similarity between the networking description vector and the networking topology vector in the topology vector database by the retrieval unit; According to the networking similarity, a set number of candidate networking topology vectors are selected from the topology vector database, and a list of candidate networking topology vectors is formed; The candidate networking topology vector list is filtered based on the recommendation rule preset by the cloud computing project, and a target networking topology vector is determined according to the filtering result.

9. The information processing method according to claim 1, wherein the step of constructing the network structure graphic information corresponding to the target network structure comprises: Generate network structure description information corresponding to the target network structure through the information generation unit in the network recommendation model, and call a preset visualization component to construct a network structure topology diagram for the target network structure; Based on the network structure description information and the network structure topology diagram, network structure graphic information corresponding to the target network structure is constructed.

10. The information processing method according to any one of claims 1 to 9, after the step of constructing the network structure graphic information corresponding to the target network structure is executed, further comprising: In case of receiving adjustment feedback information submitted for the networking structure graphic information, inputting the adjustment feedback information into the networking recommendation model; Processing the adjustment feedback information by the information processing unit to obtain an adjustment feedback vector, and updating the networking description vector to a target networking description vector by using the adjustment feedback vector; The target networking description vector is used as the networking description vector, and the step of using the retrieval unit to retrieve a target networking topology vector matching the networking description vector in the topology vector database is performed.

11. The information processing method according to any one of claims 1 to 9, wherein before the step of inputting the network description information into the network recommendation model is executed when the network description information is received, the method further comprises: Detecting whether there is newly added network topology data for the cloud computing project at a preset time node; If yes, use the heterogeneous graph network model to process the newly added networking topology data to obtain a newly added networking topology vector; The newly added networking topology vector increment is stored in the topology vector database.

12. A network recommendation method, applied to a cloud-side device, comprising: Receiving the network description information submitted by the terminal side device, and inputting the network description information into the network recommendation model, wherein the network recommendation model includes an information processing unit and a retrieval unit; Processing the networking description information by the information processing unit to obtain a networking description vector, and using the retrieval unit to retrieve a target networking topology vector matching the networking description vector in a topology vector database; Determine a target networking structure according to the target networking topology vector, and construct networking structure graphic information corresponding to the target networking structure; The networking structure graphic information is sent to the terminal side device.

13. A model training method, comprising: Obtain the network configuration information corresponding to the users participating in the cloud computing project; Determine the network topology data corresponding to the user according to the network configuration information, and store the network topology data in a graph database; Preprocessing the networking topology data stored in the graph database, and constructing a networking knowledge graph according to the preprocessing results; The networked knowledge graph is used to train an initial heterogeneous graph network model including an adaptive attention mechanism until a heterogeneous graph network model that meets the training stop condition is obtained, wherein the heterogeneous graph network model is applied to the information processing method described in any one of claims 1 to 11.

14. A computing device comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the method described in any one of claims 1 to 13 are implemented.

15. A computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions, when executed by a processor, implement the steps of the method according to any one of claims 1 to 13.

16. A computer program product, comprising a computer program or instructions, which implement the steps of the method according to any one of claims 1 to 13 when executed by a processor.

Images