CN113315655A

CN113315655A - Information configuration method of intelligent networking environment and intelligent networking system

Info

Publication number: CN113315655A
Application number: CN202110563356.6A
Authority: CN
Inventors: 孙国意; 陈厚山; 徐蔷薇
Original assignee: Henglong Communication Technology Co ltd
Current assignee: Henglong Communication Technology Co ltd
Priority date: 2021-05-24
Filing date: 2021-05-24
Publication date: 2021-08-27

Abstract

The application provides an information configuration method of an intelligent networking environment and an intelligent networking system, wherein the method comprises the following steps: the management equipment obtains a first structural equipment feature vector of each network equipment according to the equipment description information of each network equipment; the management equipment determines the topological relation among the network equipment according to the first structural equipment characteristic vector of the network equipment; the management equipment issues networking configuration information to each network equipment according to the topological relation; and each network device executes corresponding network parameter configuration according to the received networking configuration information. The management equipment determines the topological relation of each network equipment according to the equipment description information of each network equipment, and then sends corresponding networking configuration information according to the topological relation, so that automatic intelligent networking controlled by the management equipment is realized, user configuration actions are reduced, and the networking configuration efficiency is improved.

Description

Information configuration method of intelligent networking environment and intelligent networking system

Technical Field

The application relates to the technical field of intelligent networking, in particular to an information configuration method of an intelligent networking environment and an intelligent networking system.

Background

With the continuous development of information technology, network communication is very important for various industries. In some scenarios, terminals located in different geographic locations are required to operate in the same intranet, for example, network nodes, branch offices, business offices, etc. located in different locations throughout the country need to be connected to the same intranet.

In order to enable devices in different geographic locations to access the same intranet, in the prior art, a Virtual Private Network (VPN) based on Internet Protocol Security (IPSec) or a Multi-Protocol Label Switching (MPLS) dedicated line of a leased operator is often used. However, these implementations all require a network administrator with high network maintenance capability to perform special network parameter setting on local network devices to complete networking, which is difficult to implement, inconvenient for users to use, and prone to error.

Disclosure of Invention

In order to overcome the above disadvantages in the prior art, an object of the present application is to provide an information configuration method for an intelligent networking environment, which is applied to an intelligent networking system including a management device and a network device, the method including:

the management equipment obtains a first structural equipment feature vector of each network equipment according to the equipment description information of each network equipment; the device description information includes at least one of a device name, a device type, and an installation location of the network device recorded in a natural language form;

the management equipment determines a topological relation between the network equipment according to the first structural equipment characteristic vector of the network equipment;

the management equipment issues networking configuration information to each network equipment according to the topological relation;

and each network device executes corresponding network parameter configuration according to the received networking configuration information.

In a possible implementation manner, the step of determining, by the management device, a topological relation between the network devices according to the first structured device feature vector of each of the network devices includes:

for target network equipment in the network equipment, inputting a first structural equipment feature vector of the target network equipment into a trained feature extraction model to obtain a topological relation text representation vector of the target network equipment, and respectively inputting a first structural equipment feature vector of each first candidate equipment except the target network equipment in the network equipment into the trained feature extraction model to obtain a topological relation text representation vector of each first candidate equipment; the trained feature extraction model is obtained according to a first sample set marked with an upper-level relation and a lower-level relation, and the first sample set comprises network equipment samples with the same-level relation and first structured equipment feature vectors of network equipment samples with non-same-level relation;

determining the probability that each first candidate device is a superior device of the target network device according to the topological relation text representation vector of each first candidate device and the topological relation text representation vector of the target network device;

determining superior equipment of the target network equipment according to the probability that each first candidate equipment is superior equipment of the target network equipment;

and determining the topological relation of each network device according to the superior-inferior relation of each network device.

In a possible implementation manner, the step of determining, according to the topological relation text expression vector of each of the first candidate devices and the topological relation text expression vector of the target network device, a probability that each of the first candidate devices is a higher-level device of the target network device includes:

for each first candidate device, inputting the topological relation text representation vector and the device type of the target network device, the topological relation text representation vector and the device type of the first candidate device, and the installation position relation between the target network device and the first candidate device into a trained relation prediction model, and obtaining the probability that each first candidate device is a superior device of the target network device;

the trained relation prediction model is obtained according to a second sample set with a labeled probability, and the second sample set comprises a topological relation text representation vector of the network equipment samples, equipment types and installation position relations among the network equipment samples.

In a possible implementation manner, the obtaining a first structured device feature vector of each network device includes:

extracting the structural description information of each network device from the device description information of each network device;

inputting the structural description information of each network device into a trained semantic feature extraction model to obtain a first structural device feature vector of each network device; the first structural device feature vector contains an attribute value of each piece of structural description information in the device description information of the network device;

the trained feature extraction model is a knowledge representation learning network; the step of inputting the first structural device feature vector of the target network device into the trained feature extraction model to obtain the topological relation text representation vector of the target network device, and inputting the first structural device feature vector of each first candidate device in the network device except the target network device into the trained feature extraction model to obtain the topological relation text representation vector of each first candidate device includes:

based on a knowledge acquisition layer of the knowledge representation learning network, performing weighted summation on each attribute value contained in a first structured device feature vector of the target network device to obtain a second structured device feature vector of the target network device, and performing weighted summation on each attribute value contained in the first structured device feature vector of each first candidate device to obtain a second structured device feature vector of each first candidate device;

and based on the representation learning layer of the knowledge representation learning network, performing feature extraction on the topological relation text features in the second structural device feature vector of the target network device to obtain a topological relation text representation vector of the target network device, and performing feature extraction on the topological relation text features in the second structural device feature vector of each first candidate device to obtain a topological relation text representation vector of each first candidate device.

In one possible implementation manner, the trained relation prediction model comprises a semantic similarity detection network, a first knowledge graph embedding network and a second knowledge graph embedding network;

the step of inputting, for each of the first candidate devices, the topological relation text expression vector and the device type of the target network device, the topological relation text expression vector and the device type of the first candidate device, and the installation position relation between the target network device and the first candidate device into a trained relation prediction model, and obtaining the probability that each of the first candidate devices is a superior device of the target network device, includes:

obtaining description similarity characteristics between the topological relation text expression vectors of the target network device and each first candidate device through the semantic similarity detection network;

acquiring device type difference characteristics between the device types of the target network device and the first candidate devices through the first knowledge graph embedded network;

embedding the second knowledge graph into a network to obtain topological difference characteristics between the topological relation text characteristics of the target network equipment and the topological relation text characteristics of each first candidate equipment;

and determining the probability that each first candidate device is a superior device of the target network device according to the acquired description similarity feature, the device type difference feature, the installation position relationship between the target network device and each first candidate device, and the topology difference feature.

In a possible implementation manner, before the step of determining, according to the obtained description similarity feature, the device type difference feature, the installation location relationship between the target network device and each of the first candidate devices, and the topology difference feature, a probability that each of the first candidate devices is a higher-level device of the target network device, the method further includes:

inputting the description information length of each first candidate device into the trained relation prediction model;

the step of determining, according to the obtained description similarity feature, the device type difference feature, the installation position relationship between the target network device and each of the first candidate devices, and the topology difference feature, a probability that each of the first candidate devices is a higher-level device of the target network device includes:

and determining the probability that each first candidate device is a superior device of the target network device according to the acquired description similarity feature, the device type difference feature, the installation position relationship between the target network device and each first candidate device, the device description information length of each first candidate device, and the topology difference feature.

In one possible implementation, the topological relationship textual feature includes a topological relationship graph; the method further comprises the following steps:

determining first candidate equipment with similarity between topological relation text expression vectors of the first candidate equipment and the target network equipment larger than a first set value; establishing a connection relation between the first candidate device and the target network device to obtain a topological relation graph of the target network device;

for each first candidate device, determining a second candidate device, in which the similarity between each of the network devices and the topological relation text expression vector of the first candidate device is greater than a first set value, wherein each second candidate device is a candidate superior device corresponding to each first candidate device; and establishing the connection relation between the second candidate device and the first candidate device to obtain a topological relation graph of the first candidate device.

In one possible implementation, the topological relation text representation vector is a topological relation text representation vector; the method further comprises the following steps:

after training is completed, selecting a first anchor device, a positive sample device having a peer relationship with the first anchor device, a negative sample device having a non-peer relationship with the first anchor device, and a double negative sample device having a non-peer relationship with the first anchor device and a non-peer relationship with the negative sample device from the first sample set;

inputting the first structural device feature vectors of the first anchor device, the positive sample device, the negative sample device and the double negative sample device into an untrained feature extraction model respectively to obtain topological relation text representation vectors of the first anchor device, the positive sample device, the negative sample device and the double negative sample device;

adjusting parameters in the untrained feature extraction model through back propagation until the variation of the loss function is within a set threshold range, so that the distance between the topological relation text representation vectors of the sample network equipment with the same level relation is smaller than the distance between the topological relation text representation vectors of the sample network equipment with the non-same level relation, and the maximum distance between the topological relation text representation vectors of the sample network equipment with the same level relation is not larger than the minimum distance between the topological relation text representation vectors of the sample network equipment with the non-same level relation, thereby obtaining the trained feature extraction model

Selecting a second anchor device and at least one target candidate superior device corresponding to the second anchor device from the second sample set, wherein the target candidate superior device is marked with a probability that the target candidate superior device is an superior device of the second anchor device;

for each target candidate superior device, inputting the topological relation text of the second anchor device into an untrained relation prediction model to obtain the probability that each target candidate superior device is the superior device of the second anchor device, wherein the topological relation text of the second anchor device represents the vector and the device type, the topological relation text of each target candidate superior device represents the vector and the device type, and the installation position relation between each target candidate superior device and the second anchor device;

adjusting parameters in the untrained relation prediction model through an adjustment algorithm, so that the difference value between the probability of labeling each target candidate superior device and the probability obtained through the untrained relation prediction model is within a set threshold range, and obtaining the trained relation prediction model;

wherein, the step of inputting, by each target candidate superior device, the topological relation text of the second anchor device to the untrained relation prediction model, the topological relation text of each target candidate superior device to the vector and the device type, and the installation position relation between each target candidate superior device and the second anchor device to obtain the probability that each target candidate superior device is the superior device of the second anchor device specifically includes:

inputting the topological relation text of the second anchor device to the untrained relation prediction model to obtain the probability that each target candidate superior device is the superior device of the second anchor device, wherein the topological relation text of the second anchor device represents the vector and the device type, the topological relation text of each target candidate superior device represents the vector and the device type, the installation position relation of each target candidate superior device and the second anchor device, and preset adjustment parameters; the preset adjustment parameters include one or more of a topological relation text feature of the second anchor device, a topological relation text feature of each target candidate superior device, and a device description information length of each target candidate superior device.

In a possible implementation manner, the determining, according to the probability that each of the first candidate devices is an upper device of the target network device, the target network device and the upper device thereof includes:

selecting the first candidate equipment with the maximum probability in the first candidate equipment;

if the probability corresponding to the selected first candidate device is greater than a first preset probability threshold, the selected first candidate device is a superior device of the target network device;

and determining the topological relation of each network device according to the superior-inferior relation between each network device.

Another object of the present application is to provide an intelligent networking system, which includes a management device and a network device;

the management device is configured to obtain a first structured device feature vector of each network device, where the first structured device feature vector includes at least one of a device identity, a device installation location, and a device attribute; determining a topological relation between the network devices according to the first structured device feature vector of each network device; issuing networking configuration information to each network device according to the topological relation;

and the network equipment is used for executing corresponding network parameter configuration according to the received networking configuration information.

Compared with the prior art, the method has the following beneficial effects:

the embodiment of the application provides an information configuration method and an intelligent networking system for an intelligent networking environment, wherein a management device determines the topological relation of each network device according to the device description information of each network device, and then sends corresponding networking configuration information according to the topological relation, so that automatic intelligent networking controlled by the management device is realized, user configuration actions are reduced, and the networking configuration efficiency is improved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are required to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained from the drawings without inventive effort.

Fig. 1 is a schematic diagram of an intelligent networking system provided in an embodiment of the present application;

fig. 2 is a schematic flowchart illustrating steps of an information configuration method for an intelligent networking environment according to an embodiment of the present application;

fig. 3 is a schematic flowchart of a sub-step of step S120 according to an embodiment of the present disclosure.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some embodiments of the present application, but not all embodiments. The components of the embodiments of the present application, generally described and illustrated in the figures herein, can be arranged and designed in a wide variety of different configurations.

Thus, the following detailed description of the embodiments of the present application, presented in the accompanying drawings, is not intended to limit the scope of the claimed application, but is merely representative of selected embodiments of the application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined and explained in subsequent figures.

In the description of the present application, it is noted that the terms "first", "second", "third", and the like are used merely for distinguishing between descriptions and are not intended to indicate or imply relative importance.

In the description of the present application, it is further noted that, unless expressly stated or limited otherwise, the terms "disposed," "mounted," "connected," and "connected" are to be construed broadly, e.g., as meaning either a fixed connection, a removable connection, or an integral connection; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meaning of the above terms in the present application can be understood in a specific case by those of ordinary skill in the art.

Referring to fig. 1, fig. 1 is a schematic diagram of an intelligent networking system according to this embodiment, where the intelligent networking system may include a management device 100 and a plurality of network devices 200. The intelligent networking system may adopt an architecture based on a Software-Defined Wide Area Network (SD-WAN) technology, the management device 100 may be an SD-WAN controller, and each of the Network devices 200 may be Customer Premises Equipment (CPE).

Each of the network devices 200 may adopt various connection modes, such as an intranet, an MPLS network connection, an Internet (Internet) connection, a Long Term Evolution (LTE) network, and a hybrid connection (also referred to as a hybrid-link) of MPLS and Internet. For example, the network devices 200 may be disposed at different locations, the network devices 200 may communicate with each other through an intranet, the internet, or MPLS private lines provided by an operator, and the network devices 200 may be further communicatively connected to the management device 100 through a network. The network management device 100 may obtain information from each network device 200 and issue networking configuration information to each network device 200 to instruct each network device 200 to complete intelligent networking.

In this embodiment, the management device 100 is configured to obtain a first structured device feature vector of each network device 200, where the first structured device feature vector includes at least one of a device identity, a device installation location, and a device attribute; determining a topological relation between the network devices 200 according to the first structured device feature vectors of the network devices 200; and issuing networking configuration information to each network device 200 according to the topological relation.

The network device 200 is configured to execute corresponding network parameter configuration according to the received networking configuration information. After the configuration is completed, each of the network devices 200 may form an overlay internal network with a multi-level hierarchical relationship on an underlay network formed by an MPLS private network or the internet, so that a user terminal accessing to the network devices 200 performs data or information interaction in the internal network.

Specifically, referring to fig. 2, fig. 2 is a schematic flowchart illustrating steps of an information configuration method applied to an intelligent networking environment of the intelligent networking system shown in fig. 1 according to this embodiment, and each step of the method is explained in detail below.

Step S110, the management device obtains a first structured device feature vector of each network device according to the device description information of each network device.

In this embodiment, the device description information includes at least one of a device name, a device type, and an installation location of the network device recorded in a natural language. For example, the device description information of a certain network device may be "router of street C, city, a, B, D, company, F, department, a, B, etc. The management device may obtain a first structured device feature vector of the network device by analyzing the device description information.

Step S120, the management device determines a topological relation between the network devices according to the first structured device feature vectors of the network devices.

In this embodiment, the management device may determine a topological relation of each network device according to the first structured device feature vector of each network device, where the topological relation may be used to represent an upper-lower level relation of each connected network device.

For example, according to the installation position recorded in the description information of the network device, the superior-inferior relationship between the network devices installed in different administrative level areas (province, city, district) can be determined; according to the department to which the description information of the network equipment belongs, the superior-inferior relation of different network equipment installed at the same approximate position can be determined.

Step S130, the management device issues networking configuration information to each network device according to the topological relation.

In this embodiment, the management device may determine, according to the topological relation, parameters required for networking of each network device. For example, a router of a provincial company is a superior router of a plurality of urban companies, and the router of the provincial company can be used as a gateway of the router of the urban companies in an intranet when the router is intelligently networked. Therefore, the management device can enable the router of the provincial company and the router of the city company to establish a tunnel on the network of the underlay level through issuing configuration, and configure the router of the provincial company as a gateway of the router of the city company in the intranet of the overlay level.

The management device may be configured with a networking rule in advance, the networking rule may record different networking requirements corresponding to different topological relations, and the management device may match the corresponding networking rule according to the determined topological relation and determine corresponding networking configuration information.

Step S140, each network device executes corresponding network parameter configuration according to the received networking configuration information.

In this embodiment, each network device may automatically complete networking according to the networking configuration information received by itself, and a user does not need to perform a special configuration operation, so that configuration actions of the user may be reduced, configuration efficiency may be improved, and the possibility of errors in manual configuration may be reduced.

The management device may send the networking configuration information to the corresponding network device after establishing communication with each of the network devices. The management device can also generate links, mails, configuration files and the like with networking configuration information, send the links, mails, configuration files and the like to corresponding network devices through other communication modes, and then obtain the networking configuration information and complete network parameter configuration by each network device according to the received connections, mails, configuration files and the like.

Optionally, in this embodiment, the management device may determine networking configuration information required for completing networking for each network device, and may also determine a corresponding security policy, speed limit policy, internal network forwarding policy, and the like for each network device according to the topology information, so as to automatically manage a data forwarding action of the entire intelligent networking system on an overlay level according to the determined topology information.

For example, for network devices of different hierarchies, a network device of a higher hierarchy is generally a network device of a more important website or company, and the device thereof is generally better, the amount of data to be born is larger, and the security requirement is higher, so the management device can determine a relatively strict security policy and a relatively strict speed limit policy for a network device of a higher hierarchy, and determine a relatively strict security policy and a relatively strict speed limit policy for a network device of a lower hierarchy.

Referring to fig. 2, fig. 2 is a schematic flow chart of sub-steps of step S120 shown in fig. 1, and the sub-steps of step S120 are explained in detail with reference to fig. 2.

Step S121, for a target network device in the network devices, inputting a first structural device feature vector of the target network device into a trained feature extraction model, obtaining a topological relation text representation vector of the target network device, and inputting a first structural device feature vector of each first candidate device in the network devices except the target network device into the trained feature extraction model, respectively, to obtain a topological relation text representation vector of each first candidate device.

In this embodiment, the trained feature extraction model is obtained from a first sample set labeled with a hierarchical relationship, where the first sample set includes network device samples having a hierarchical relationship and first structured device feature vectors of network device samples having a non-hierarchical relationship.

It should be noted that, in this embodiment, network devices located at a subordinate level of the same network device have a peer relationship, that is, peer network devices; network devices subordinate to different network devices have a non-peer relationship, i.e., no peer relationship.

Step S122, determining a probability that each of the first candidate devices is a superior device of the target network device according to the topological relation text representation vector of each of the first candidate devices and the topological relation text representation vector of the target network device.

Step S123, determining a superior device of the target network device according to the probability that each of the first candidate devices is the superior device of the target network device.

Step S124, determining a topological relation of each network device according to the superior-inferior relation of each network device.

In this embodiment, after the superior relationship of each network device is determined, the superior relationships are associated and fused, so that the hierarchical topological relationship between the network devices in the entire networking system can be determined.

In some possible implementation manners, in step S122, for each first candidate device, the topological Relation text representation vector and the device type of the target Network device, the topological Relation text representation vector and the device type of the first candidate device, and the installation position Relation between the target Network device and the first candidate device may be input into a trained Relation prediction model (relationship Classification Network), so as to obtain a probability that each first candidate device is a superior device of the target Network device.

Further, the trained relation prediction model comprises a semantic similarity detection network, a first knowledge graph embedding network and a second knowledge graph embedding network. Input parameters of the relational prediction model include topological relational text representation vectors, device types, and spatial distances of installation locations. The spatial distance of the installation position can be determined by combining text description information of the installation position with electronic map positioning and the like.

In step S122, a description similarity feature between the target network device and the topological relation text expression vector of each of the first candidate devices may be obtained through the semantic similarity detection network.

And then acquiring device type difference characteristics between the device types of the target network device and the first candidate devices through the first knowledge graph embedded network. And embedding the second knowledge graph into a network to obtain topological difference characteristics between the topological relation text characteristics of the target network equipment and the topological relation text characteristics of each first candidate equipment.

In the above embodiment, when determining the probability that two network devices have a hierarchical relationship, the relationship such as the device type and the distance of the installation position of the network device is further considered in addition to the topological relationship text representation vector of the network device, and these pieces of information closely related to the topological relationship text feature of the network device are combined to determine the probability based on the relationship prediction model, thereby further improving the accuracy of the topological relationship identification.

In some possible implementations, in step S110, the management device may extract the structural description information of each network device from the device description information of each network device. And then, inputting the structural description information of each network device into the trained semantic feature extraction model to obtain a first structural device feature vector of each network device.

The first structured device feature vector contains an attribute value of each piece of structured description information in the device description information of the network device. The semantic feature extraction model may include a shallow semantic feature extraction model, such as a Bidirectional Encoder Representation algorithm transforms (BERT) model, Word vector models such as Word2Vec, and the like. The semantic feature extraction model can be obtained through sample training in massive Chinese corpora such as news.

In this embodiment, the first structured device feature vector of the network device may be a structured vector, and includes attribute values corresponding to each piece of structured description information. After the structured description information vectors such as the name, the installation address, the device type and the like of the network device are expressed, the attribute values are merged and abstracted to obtain a high-dimensional expression vector as the first structured feature vector.

For example, vector representations of key phrases such as network device administrative areas, network device names, types of network devices and the like are extracted based on the trained semantic feature extraction model, and attribute values corresponding to the structured description information are obtained.

Further, in some possible implementations, each of the first candidate device description information lengths may also be input into the trained relationship prediction model. Then, according to the obtained description similarity feature, the device type difference feature, the installation position relationship between the target network device and each of the first candidate devices, the device description information length of each of the first candidate devices, and the topology difference feature, the probability that each of the first candidate devices is a superior device of the target network device is determined.

Because the importance of each attribute of the network equipment for distinguishing the superior-subordinate relation of the network equipment is different, and the action values of different words in one attribute are also different. Because the first structural device feature vector is obtained based on each structural description information of the network device, the importance degree of each attribute of the network device can be set so as to improve the accuracy of extracting the hierarchical semantic features.

In some possible implementations, therefore, the trained feature extraction model is a knowledge Learning Network (knowledge Learning Network) that includes a knowledge acquisition layer and a Representation Learning layer.

In step S121, a knowledge acquisition layer of the learning network may be represented based on the knowledge, perform weighted summation on each attribute value included in the first structured device feature vector of the target network device to obtain a second structured device feature vector of the target network device, and perform weighted summation on each attribute value included in the first structured device feature vector of each first candidate device to obtain a second structured device feature vector of each first candidate device.

And based on the representation learning layer of the knowledge representation learning network, performing feature extraction on the topological relation text features in the second structural device feature vector of the target network device to obtain a topological relation text representation vector of the target network device, and performing feature extraction on the topological relation text features in the second structural device feature vector of each first candidate device to obtain a topological relation text representation vector of each first candidate device. In this embodiment, the weighted values of different attribute values can be obtained by back propagation network learning.

Based on the design, the knowledge acquisition layer is adopted to perform linear weighted summation on a plurality of attribute values reaching the network equipment, which is equivalent to applying different attention to different attributes, so that the influence degree of different attribute values on subsequent judgment is changed.

In this embodiment, since attributes of the network devices at the same level, such as device names and installation locations, may have higher similarity, structured description information of the network devices, such as device names and installation locations, is extracted based on a trained semantic feature extraction model to obtain vectorization representation, a network device semantic representation learning network is constructed, and a Loss function, such as Quadruplet Loss, is used for adjustment, so that a hierarchical topological relation text representation vector of each network device can be obtained, instead of only text semantic features of the network devices, and a more accurate context relation of the network devices can be obtained based on a subsequent topological relation text representation vector of the network devices.

In some possible implementations, the topological relationship textual feature includes a topological relationship graph. The topological relation graph can be obtained in the following manner.

For the target network device, determining a first candidate device, of the first candidate devices, whose similarity to the text representation vector of the topological relation of the target network device is greater than a first set value; and establishing a connection relation between the first candidate device and the target network device to obtain a topological relation graph of the target network device.

The similarity between the topological relation text representation vectors can be obtained by calculating cosine similarity, Pearson correlation coefficient, Euclidean distance and other methods.

For example, in the network device topology relationship diagram, the more network devices connected to other network devices, the higher the possibility that the network devices are superior devices, and the less network devices connected to other network devices, the lower the possibility that the network devices are superior devices.

Therefore, according to the topological relation text expression vector containing the hierarchical semantic information of each network device extracted by the trained feature extraction model, the cosine similarity between a certain target network device and all first candidate devices in the adjacent range of the target network device is calculated, and the network device pairs with the cosine similarity larger than a first set value are set to be connected with each other.

Similarly, for each first candidate device, a second candidate device whose similarity between each of the network devices and the topological relation text representation vector of the first candidate device is greater than a first set value may be determined, where each second candidate device is a candidate superior device corresponding to each first candidate device; and establishing the connection relation between the second candidate device and the first candidate device to obtain a topological relation graph of the first candidate device.

Further, in determining the second candidate device, the second candidate device may be determined in a network device located within a set range from the first candidate device, for example, setting a formation area, or setting a distance.

In an optional implementation manner, when determining the topology difference feature between two network devices, the similarity of the topology relationship text representation vector of each network device and its corresponding candidate superior device node may be calculated, and the difference between the mean values of the similarities may be used as the topology difference feature of the two network devices.

In one possible implementation, the topological relation text representation vector is a topological relation text representation vector. The feature extraction model may be trained by the following.

and adjusting parameters in the untrained feature extraction model through back propagation until the variation of the loss function is within a set threshold range, so that the distance between the topological relation text representation vectors of the sample network equipment with the same level relation is smaller than the distance between the topological relation text representation vectors of the sample network equipment with the non-same level relation, and the maximum distance between the topological relation text representation vectors of the sample network equipment with the same level relation is not larger than the minimum distance between the topological relation text representation vectors of the sample network equipment with the non-same level relation, thereby obtaining the trained feature extraction model.

During adjustment training, 4 network devices are selected from the first sample set, the first anchor device X1 and the positive sample device X2 are subordinate devices of the same network device, and the negative sample device X3 and the double negative sample device X4 are subordinate devices of two other different network devices, respectively.

And carrying out forward propagation on the feature extraction model to obtain topological relation text representation vectors of the 4 network devices, and marking the vectors as FX1, FX2, FX3 and FX 4.

The final training result of the feature extraction model is to make the euclidean distance between the topological relation text representation vectors of subordinate devices of the same network device as close as possible, for example, between FX1 and FX 2; and the euclidean distances between the topological relation textual representation vectors of subordinate devices of different network devices are as far as possible, e.g., between FX1 and FX3, and FX1 and FX 4. In this embodiment, the Loss function may be a quadlet Loss (quadruple Loss function), and the feature extraction model is adjusted by the quadruple Loss function, so that the feature extraction model has better convergence.

In the embodiment of the present application, the Quadruplet Loss includes two items, where an adjustment target of the first item is to make a euclidean distance between topological relation text representation vectors of subordinate devices of the same network device smaller than a euclidean distance between topological relation text representation vectors of subordinate devices of different network devices; the second term is adjusted to make the distance between the topological relation text representation vectors of the first anchor device and the positive direction sample device smaller than the distance between the topological relation text representation vectors of the negative direction sample devices of two different groups, and the term can make the topological relation text representation vectors after representation learning more closely distributed and clustered.

In one possible implementation, the relational prediction model may be trained by the following steps.

The second anchor device is equivalent to a target network device in the using process of the relationship prediction model, the target candidate superior device is equivalent to a first candidate device in the using process of the relationship prediction model, and the target candidate superior device is marked with the probability that the sample network device is the superior device of the second anchor device, wherein the probability can be manually marked.

In the training process of the relation prediction model, a topological relation text representation vector and a device type of a second anchor device in a network device pair consisting of the second anchor device and a target candidate device, the topological relation text representation vector and the device type of the target candidate device, and the installation position relation between the second anchor device and the target candidate device are used as input parameters to be input into the untrained relation prediction model, the probability that the target candidate device output by the relation prediction model is used as a superior device of the second anchor device is obtained and compared with the probability that the target candidate device is marked as the superior device of the second anchor device, if the difference value of the two probabilities is not in a set threshold range, the parameters in the relation prediction model are continuously adjusted, and the probability that the target candidate device is used as the superior device of the second anchor device is recalculated after the parameters are adjusted, and determining whether to adjust the parameters of the relation prediction model or not according to the difference value of the probabilities, and obtaining the finally trained relation prediction model through iterative training.

In some possible implementation manners, when the input parameters of the relationship prediction model further include the device description information length of the network device and the topological relation text feature of the network device, in the training process of the relationship prediction model, the topological relation text of the second anchor device may also represent a vector and a device type, the topological relation text of each target candidate superior device represents a vector and a device type, the installation position relationship between each target candidate superior device and the second anchor device, and a preset adjustment parameter is input into the untrained relationship prediction model, so that the probability that any one target candidate superior device is an superior device of the second anchor device is obtained. The preset adjustment parameters comprise one or more of topological relation text features of the second anchor device, topological relation text features of each target candidate superior device and device description information lengths of each target candidate superior device, and the topological relation text features are determined according to topological relation text representation vectors of the network devices.

In this embodiment, the first sample set includes at least 4 sample network devices, specifically includes at least one first anchor device, at least one positive sample device, and at least two negative sample devices, where the two negative sample devices include at least one dual negative sample device; the second sample set includes at least two sample network devices, specifically includes at least one second anchor device, and at least one target candidate superior device; the samples included in the first sample set and the second sample set may be all the same or may be partially the same.

In some possible implementation manners, in step S123, the first candidate device with the highest probability corresponding to each of the first candidate devices may be selected. And if the probability corresponding to the selected first candidate device is greater than a first preset probability threshold, the selected first candidate device is a superior device of the target network device.

For example, the target network device corresponds to 3 first candidate devices, and assuming that the corresponding probabilities are 0.3, 0.8, and 0.1, the third first candidate device with the highest probability is selected, and then the probability of the first candidate device is compared with a first preset probability threshold value of 0.7. If the second candidate device is greater than the first candidate device, the third candidate device is the upper-level device of the target network device. If not, it indicates that there is no upper level device of the target network device in the 3 first candidate devices.

Alternatively, if there is no superior device of the target network device in the first candidate devices, the first candidate devices may be peer network devices of the target network device, or may be network devices unrelated to the target network device, or may be subordinate devices of the target network device. Therefore, further determination may be made based on the distance between the topological relation textual representation vectors of the target network device and the first candidate device.

If the distance between the text representation vectors of the topological relations between the target network device and the first candidate device is smaller than the second preset probability threshold, the target network device and the first candidate device may be peer network devices.

If the distance between the text representation vectors of the topological relation between the target network device and the first candidate device is greater than the third preset probability threshold, it may be determined that the target network device and the first candidate device are non-peer network devices, that is, the target network device is different from a superior device of the first candidate device.

In some possible implementation manners, in step S130, the management device may generate a networking configuration with a hierarchical relationship according to the topological relationship of the network device, so that each network device forms an internal network with a hierarchical relationship at an overlay level. The management device may regard the top-level network devices as the same-level network devices, and issue corresponding networking configuration information to enable the top-level network devices to be located in an intranet of the same network segment on an overlay level.

In summary, the embodiments of the present application provide an information configuration method and an intelligent networking system for an intelligent networking environment, where a management device determines a topological relation of each network device according to device description information of each network device, and then issues corresponding networking configuration information according to the topological relation, so as to implement automatic intelligent networking controlled by the management device, reduce user configuration actions, and improve networking configuration efficiency.

In the embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. The apparatus embodiments described above are merely illustrative, and for example, the flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

In addition, functional modules in the embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

The functions, if implemented in the form of software functional modules and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application or portions thereof that substantially contribute to the prior art may be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.

It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.

The above description is only for various embodiments of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art can easily conceive of changes or substitutions within the technical scope of the present application, and all such changes or substitutions are included in the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. An information configuration method of an intelligent networking environment is characterized in that the method is applied to an intelligent networking system comprising a management device and a network device, and comprises the following steps:

2. The method of claim 1, wherein the step of the management device determining the topological relationship between the network devices according to the first structured device feature vector of each network device comprises:

3. The method according to claim 2, wherein the step of determining the probability that each of the first candidate devices is a superior device of the target network device according to the topological relation text representation vector of each of the first candidate devices and the topological relation text representation vector of the target network device comprises:

4. The method according to claim 2, wherein the step of obtaining the first structured device feature vector of each network device according to the device description information of each network device comprises:

5. The method of claim 3, wherein the trained relationship prediction model comprises a semantic similarity detection network, a first knowledge-graph embedding network, and a second knowledge-graph embedding network;

6. The method according to claim 5, wherein before the step of determining the probability that each of the first candidate devices is a superior device of the target network device according to the acquired description similarity feature, the device type difference feature, the installation location relationship between the target network device and each of the first candidate devices, and the topology difference feature, the method further comprises:

7. The method of claim 6, wherein the topological relational text feature comprises a topological relational graph; the method further comprises the following steps:

8. The method according to claim 3, wherein the topological relation text representation vector is a topological relation text representation vector; the method further comprises the following steps:

adjusting parameters in the untrained feature extraction model through back propagation until the variation of the loss function is within a set threshold range, so that the distance between topological relation text representation vectors of sample network equipment with a peer relation is smaller than the distance between topological relation text representation vectors of sample network equipment with a non-peer relation, and the maximum distance between topological relation text representation vectors of sample network equipment with a peer relation is not larger than the minimum distance between topological relation text representation vectors of sample network equipment with a non-peer relation, thereby obtaining the trained feature extraction model;

9. The method of claim 2, wherein the determining the target network device and the superior device according to the probability that each of the first candidate devices is the superior device of the target network device comprises:

and if the probability corresponding to the selected first candidate device is greater than a first preset probability threshold, the selected first candidate device is a superior device of the target network device.

10. An intelligent networking system is characterized by comprising management equipment and network equipment;

the management device is used for acquiring a first structured device feature vector of each network device according to the device description information of each network device; the device description information includes at least one of a device name, a device type, and an installation location of the network device recorded in a natural language form; determining a topological relation between the network devices according to the first structured device feature vector of each network device; issuing networking configuration information to each network device according to the topological relation;