CN116016122A - Prediction method, device, equipment and storage medium for network failure solution - Google Patents

Abstract

The application provides a prediction method, a device, equipment and a storage medium of a network fault solution. The method comprises the following steps: receiving target network fault data corresponding to a target network sent by a user terminal; predicting the solution of the target network according to the fault data of the target network by adopting a solution prediction model trained to be converged so as to obtain the target solution; the solution prediction model trained to be converged is a fusion model of a bi-directional encoder characterization quantity Bert model of a plurality of twin converters which are constructed in advance; the target solution is sent to the user terminal to instruct the user terminal to display the target solution. According to the method, a solution prediction model trained to be converged is adopted to predict a target solution, and the method does not depend on operation and maintenance personnel; meanwhile, time can be saved, efficiency of finding a target solution is provided, and user experience is improved.

Description

Prediction method, device, equipment and storage medium for network fault solution

technical field

The present application relates to data processing technology, and in particular to a network failure solution prediction method, device, equipment and storage medium.

Background technique

With the development of society, users' demand for the network is increasing, and the probability of network failures during use is also increased. In order to deal with network failures in a timely manner, operators have also made a lot of efforts.

In the prior art, when a user encounters a network failure in life or work, the user promptly informs the corresponding operation and maintenance personnel of the network failure, and then the operation and maintenance personnel determine the corresponding solution to the network failure through the network failure situation, and then inform How users solve it. Among them, for complex network failure situations, operation and maintenance personnel may need to determine the corresponding solution through offline survey.

It can be seen that in the prior art, when a user encounters a network failure, the user relies on the professionalism of the operation and maintenance personnel to find a solution, and then makes corresponding processing, so that the user has a great dependence on the operation and maintenance personnel; at the same time, Since the operation and maintenance personnel need to spend a lot of time looking for a solution, the efficiency is low, and the user's experience of using the network will also be reduced.

Contents of the invention

This application provides a prediction method, device, equipment, and storage medium for network failure solutions to solve the problem that users have a great dependence on operation and maintenance personnel, and that operation and maintenance personnel need to spend a lot of time looking for solutions. The problem of low efficiency.

In a first aspect, the present application provides a method for predicting a network failure solution, the method comprising:

receiving target network fault data corresponding to the target network sent by the user terminal;

The solution prediction model trained to convergence is used to predict the solution of the target network according to the fault data of the target network to obtain the target solution; the solution prediction model trained to convergence is a plurality of pre-built twins A fusion model of the bidirectional encoder characterization quantity Bert model of the converter;

sending the target solution to the user terminal to instruct the user terminal to display the target solution.

In one way, the solution prediction model that has been trained to convergence includes a plurality of twin Bert models that have been trained to convergence; different twin Bert models have different levels of configuration parameters; Semantic granularity capture feature parameters;

The solution prediction model trained to convergence is used to predict the solution of the target network according to the fault data of the target network to obtain the target solution, including:

Input the target network failure data into each of the twin Bert models that have been trained to converge;

Using each of the twin Bert models that have been trained to converge to perform feature extraction corresponding to the semantic granularity of the target network fault data, so as to obtain a corresponding network fault feature vector;

The solution of the target network is predicted according to the plurality of network fault feature vectors, so as to obtain the target solution.

In one manner, the solution prediction model that has been trained to convergence further includes a linear layer; the linear layer includes a preset prediction label formula;

The predicting the solution of the target network according to the plurality of network fault feature vectors to obtain the target solution includes:

Calculate the network fault fusion feature vector based on the network fault weight formula and a plurality of the network fault feature vectors;

inputting the network fault fusion feature vector into the preset prediction label formula to calculate a target prediction label value;

determining a corresponding target prediction label range based on the target prediction label value;

A corresponding target solution is determined based on the target predicted label range.

In one manner, the determining the corresponding target solution based on the target prediction label range includes:

Obtain the mapping relationship between the pre-stored preset prediction label range and its corresponding solution;

In response to the target predicted label range being consistent with any preset predicted label range in the mapping relationship, a solution corresponding to the consistent preset predicted label range is determined as the target solution.

In one manner, before using the solution prediction model that has been trained to convergence to predict the solution of the target network according to the fault data of the target network, it further includes:

Obtain a training sample set; the network fault data in the training sample set comes from the historical network fault data sent by the user terminal and the network fault data in the expert experience knowledge base; the training sample set includes a plurality of network fault data and each network fault The predicted label value of the solution corresponding to the data;

Using the training sample set to train a plurality of twin preset Bert models;

Determining multiple twin Bert models satisfying preset training convergence conditions as multiple twin Bert models trained to convergence.

In one manner, after the multiple twin Bert models that meet the preset training convergence conditions are determined as multiple twin Bert models that have been trained to convergence, it also includes:

Construct a fusion model of multiple twin Bert models based on multiple twin Bert models that have been trained to convergence and a linear layer;

The fusion model of the plurality of twin Bert models is determined as the solution prediction model that has been trained to convergence.

In one mode, the method also includes:

Obtain a network fault update data set; the network fault update data in the network fault update data set comes from the network fault update data sent by the user terminal and the network fault update data in the expert experience knowledge base; the network fault update data set contains multiple Each network fault update data and the predicted label value of the solution corresponding to each network fault update data;

A plurality of twin Bert models that have been trained to convergence are updated and trained using the network fault update data set;

Determining multiple twin Bert models that meet the preset update training conditions as multiple updated twin Bert models that have been trained to converge;

Construct a fusion model of multiple updated twin Bert models based on multiple updated twin Bert models trained to convergence and a linear layer;

The fusion model of the multiple updated twin Bert models is determined as the latest solution prediction model that has been trained to convergence.

In a second aspect, the present application provides a device for predicting network failure solutions, the device comprising:

A receiving module, configured to receive target network fault data corresponding to the target network sent by the user terminal;

The prediction module is used to use the solution prediction model trained to convergence to predict the solution of the target network according to the fault data of the target network to obtain the target solution; the solution prediction model trained to convergence is pre- The fusion model of the two-way encoder characterization Bert model of multiple twin transformers constructed;

A sending module, configured to send the target solution to the user terminal, so as to instruct the user terminal to display the target solution.

In a third aspect, the present application provides an electronic device, including: a processor, and a memory communicatively connected to the processor;

the memory stores computer-executable instructions;

The processor executes the computer-executed instructions stored in the memory, so as to implement the method as described in the first aspect or any one manner.

In a fourth aspect, the present application provides a computer-readable storage medium, where computer-executable instructions are stored in the computer-readable storage medium, and when the computer-executable instructions are executed by a processor, they are used to implement the first aspect or any one of method described in the method.

The present application provides a network failure solution prediction method, device, equipment, and storage medium, specifically including: receiving the target network failure data corresponding to the target network sent by the user terminal; using the solution prediction model that has been trained to converge according to the target The network fault data predicts the solution of the target network to obtain the target solution; the solution prediction model that has been trained to converge is a fusion model of the bidirectional encoder characterization Bert model of multiple twin transformers built in advance; the target The solution is sent to the user terminal to instruct the user terminal to display the target solution. The prediction device for the network fault solution of the present application (hereinafter referred to as the prediction device) first receives the target network fault data corresponding to the target network sent by the user terminal, and then the prediction device will use the solution prediction model that has been trained to converge to analyze the target network solution For prediction, since the solution prediction model is pre-trained to convergence, it is a better prediction model. At the same time, since the solution prediction model that has been trained to convergence is a fusion model of multiple twin Bert models, the target network failure The data is predicted based on multiple twin Bert models, so the final predicted target solution is more accurate; compared with the existing technology, this application no longer relies on the operation and maintenance personnel to find the solution to the target network fault data, so it reduces the need for Dependency of operation and maintenance personnel; further, operation and maintenance personnel need to spend a lot of time to find the target solution, so the efficiency is low, and this application does not rely on operation and maintenance personnel, with the help of the solution prediction model that has been trained to converge The target solution can be predicted, so it can reduce the time and improve the efficiency of finding the solution, and the user of this application can directly obtain the target solution by himself, further improving the user experience.

Description of drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and together with the description serve to explain the principles of the application.

FIG. 1 is an application scenario diagram of a prediction method for a network failure solution provided by the present application;

FIG. 2 is a schematic flowchart of a method for predicting a network fault solution provided in Embodiment 1 of the present application;

FIG. 3 is a schematic flowchart of a method for predicting a network fault solution provided in Embodiment 2 of the present application;

FIG. 4 is a schematic flowchart of a method for predicting a network failure solution provided in Embodiment 3 of the present application;

FIG. 5 is a schematic flowchart of a method for predicting a network fault solution provided in Embodiment 4 of the present application;

FIG. 6 is a schematic diagram of a solution prediction model that has been trained to convergence provided in Embodiment 4 of the present application;

FIG. 7 is a schematic flowchart of a method for predicting a network fault solution provided in Embodiment 5 of the present application;

FIG. 8 is a schematic diagram of a prediction device for a network fault solution provided in Embodiment 6 of the present application;

FIG. 9 is a schematic structural diagram of an electronic device provided in Embodiment 7 of the present application.

By means of the above drawings, specific embodiments of the present application have been shown, which will be described in more detail hereinafter. These drawings and text descriptions are not intended to limit the scope of the concept of the application in any way, but to illustrate the concept of the application for those skilled in the art by referring to specific embodiments.

Detailed ways

Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numerals in different drawings refer to the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with aspects of the present application as recited in the appended claims.

First, the nouns involved in this application are explained:

Bidirectional Encoder Representation from Transformers (Bert for short): refers to an unsupervised pre-trained language model for natural language processing tasks.

In the prior art, when a user encounters a network failure in life or work, the user promptly informs the corresponding operation and maintenance personnel of the network failure, and then the operation and maintenance personnel determine the corresponding solution to the network failure through the network failure situation, and then inform How users solve it. Among them, for complex network failure situations, operation and maintenance personnel may need to determine the corresponding solution through offline survey.

It can be seen that in the prior art, when a user encounters a network failure, the user relies on the operation and maintenance personnel to find a solution and then make corresponding processing, so the user has a great dependence on the operation and maintenance personnel; at the same time, due to the operation and maintenance personnel It takes a lot of time for personnel to find a solution, so the efficiency is low, so it will also reduce the user's experience of using the network.

In order to solve the defects of the prior art, the inventor of the present solution designs a new solution through creative research. This solution provides a prediction method for network failure solutions. In order to solve the problem that users have a great dependence on operation and maintenance personnel, the prediction device of this solution receives the target network failure data sent by the user terminal, and then passes the training to converge The solution prediction model of the target network predicts the target network fault data to obtain the solution of the target network. Since this solution is obtained by using the solution prediction model that has been trained to converge, it does not rely on the operation and maintenance personnel to find the target solution. , thus reducing the dependence on the operation and maintenance personnel; in order to solve the problem that the operation and maintenance personnel need to spend opening time and low efficiency when looking for the target solution, this solution obtains the target solution through machines instead of manual labor, so it can save time to improve the efficiency of obtaining the target solution. This solution directly sends the obtained target solution to the user terminal, so that the user can make corresponding processing based on the target solution displayed on the user terminal. Since the target solution can be displayed through the user terminal, Therefore, the user can obtain it directly, which in turn can improve the user experience.

The application scenarios of a network failure solution prediction method, device, equipment, and storage medium provided by the present application are introduced below.

FIG. 1 is an application scenario diagram of a network failure solution prediction method provided by the present application. As shown in Figure 1, the application scenario diagram includes a user terminal 101 and an electronic device 102, wherein the electronic device 102 includes a prediction device (hereinafter referred to as the prediction device) 103 for network failure solutions, and the prediction device 103 includes The solution prediction model 104 .

Wherein, the user terminal 101 is connected in communication with the electronic device 102, wherein the communication connection may be a wired connection or a wireless connection.

Specifically, the user inputs the target network fault data corresponding to the target network through the user terminal 101 , and then the user terminal 101 sends the target network fault data to the electronic device 102 , and the electronic device 102 transmits the target network fault data to the prediction device 103 .

Further, the prediction device 103 inputs the target network fault data into the solution prediction model 104 that has been trained to convergence, and then uses the prediction model 104 that has been trained to a convergent solution to predict the target fault data to obtain the target solution.

Further, the prediction device 103 sends the target solution to the user terminal 101 to instruct the user terminal 101 to display the target solution.

The method for predicting network fault solutions provided in this application aims to solve the above technical problems in the prior art.

The technical solution of the present application and how the technical solution of the present application solves the above technical problems will be described in detail below with specific embodiments. The following specific embodiments may be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. Embodiments of the present application will be described below in conjunction with the accompanying drawings.

Embodiment one

FIG. 2 is a schematic flowchart of a method for predicting a network fault solution provided in Embodiment 1 of the present application. The execution subject of the method in this embodiment is a predicting device for network failure solutions (hereinafter referred to as the predicting device), as shown in FIG. 2 , and the specific steps are as follows.

S201. Receive target network failure data corresponding to the target network sent by the user terminal.

Wherein, the user terminal is a device used by the user, which may be a mobile phone or a computer, and is not limited here.

Wherein, the target network is a network used by the user in society or work.

Wherein, the target network failure data is a description of the failure of the target network.

Specifically, the user inputs the target network fault data in the user terminal, and then sends the target network fault data to the predicting device, and the predicting device receives the target network fault data.

Wherein, the user terminal may be loaded with a network fault application software, and the user opens the network fault application software to input target network fault data.

S202, using the solution prediction model that has been trained to convergence to predict the solution of the target network according to the fault data of the target network to obtain the target solution; the solution prediction model that has been trained to convergence is a plurality of twin transformers pre-built A Fusion Model of the Bidirectional Encoder Representation Bert Model.

Among them, the solution prediction model that has been trained to convergence is a fusion model of multiple twin Bert models built in advance.

Specifically, the target network fault data is input into the solution prediction model that has been trained to convergence, and then the solution prediction model that has been trained to convergence predicts the target solution through the semantic description of the target network fault data.

S203. Send the target solution to the user terminal to instruct the user terminal to display the target solution.

Specifically, the prediction device sends the target solution to the user terminal in a wired or wireless manner, so that the target solution will be displayed in the user terminal.

In one manner, when the target solution is sent to the user terminal, a reminder message is also generated, thereby reminding the user to check the target solution through the user terminal.

This embodiment provides a method for predicting a network fault solution, which specifically includes: receiving target network fault data corresponding to the target network sent by a user terminal; The solution is predicted to obtain the target solution; the solution prediction model that has been trained to convergence is the fusion model of the bidirectional encoder characterization Bert model of multiple twin transformers pre-built; the target solution is sent to the user terminal, to instruct the user terminal to display the target solution. The prediction device of the network fault solution in this embodiment (hereinafter referred to as the prediction device) first receives the target network fault data corresponding to the target network sent by the user terminal, and then the prediction device will use the solution prediction model that has been trained to converge to solve the problem of the target network. The solution prediction model is pre-trained to convergence, so it is a better prediction model. At the same time, since the solution prediction model that has been trained to convergence is a fusion model of multiple twin Bert models, the target network The fault data is predicted based on multiple twin Bert models, so the final predicted target solution is more accurate; compared with the existing technology, this application no longer relies on the operation and maintenance personnel to find the solution to the target network fault data, so it reduces Dependence on operation and maintenance personnel; further, operation and maintenance personnel need to spend a lot of time to find the target solution, so the efficiency is low, and this application does not rely on operation and maintenance personnel, with the help of the solution prediction model that has been trained to convergence The target solution can be predicted, so it can save time and improve the efficiency of finding the solution, and the user of this application can directly obtain the target solution by himself, further improving the user experience.

Embodiment two

This embodiment is a further refinement of the first embodiment above. The solution prediction model that has been trained to convergence in this embodiment includes multiple twin Bert models that have been trained to convergence; different twin Bert models have different levels of configuration parameters; configuration The parameters are the parameters of the Bert model about the features of semantic granularity.

Among them, the configuration parameters can be Transformers with different layers and the number of hidden layers.

FIG. 3 is a schematic flowchart of a method for predicting a network fault solution provided in Embodiment 2 of the present application. In this embodiment, a solution prediction model that has been trained to convergence is used to predict the solution of the target network according to the fault data of the target network, so as to obtain the target solution, as shown in FIG. 3 , and the specific steps are as follows.

S301. Input target network fault data into respective twin Bert models that have been trained to convergence.

Specifically, the prediction device inputs the target network fault data into the twin Bert model for training to convergence in parallel.

S302. Using each of the twin Bert models that have been trained to convergence to perform feature extraction corresponding to the semantic granularity of the target network fault data, so as to obtain corresponding network fault feature vectors.

Among them, the semantic granularity refers to whether the text is divided into words, and the input features of a sentence are represented by words or characters.

Among them, the network fault feature vector is a fixed-length vector.

Specifically, the twin Bert models that have been trained to convergence perform feature extraction corresponding to the semantic granularity of the same network fault data. Since the semantic granularity is inconsistent, the obtained network fault feature vectors are also inconsistent, and then the corresponding network fault feature vectors are respectively obtained. .

S303. Predict the solution of the target network according to the plurality of network fault feature vectors, so as to obtain the target solution.

Specifically, the target solution is predicted according to the characteristics of multiple network fault feature vectors.

This embodiment provides a prediction method for a network fault solution. In this embodiment, the solution prediction model that has been trained to convergence includes multiple twin Bert models that have been trained to convergence; different twin Bert models have different levels of configuration parameters. ; Configuration parameters are the parameters of the Bert model about the features of semantic granularity. When using the solution prediction model that has been trained to convergence to predict the solution of the target network based on the target network fault data to obtain the target solution, it specifically includes: inputting the target network fault data into each of the twins that have been trained to convergence In the Bert model, the twin Bert models that have been trained to convergence are used to extract features corresponding to the semantic granularity of the target network fault data to obtain the corresponding network fault feature vectors; the solution to the target network based on multiple network fault feature vectors Predictions are made to obtain target solutions. This embodiment uses multiple twin Bert models that have been trained to converge to predict network fault data. Since the Bert model has been trained to converge, it is a better model. At the same time, this embodiment uses multiple twin Bert models that have been trained to Convergent twin Bert models. Different twin Bert models have different levels of configuration parameters. Since the configuration parameters are the Bert model’s capture feature parameters for semantic granularity, multiple twin Bert models that have been trained to converge can be used to obtain their corresponding networks. Fault eigenvector, the network fault eigenvector is accurate, and considering multiple configuration parameters, it is more comprehensive, and then based on multiple network fault eigenvectors to predict the target solution, because the network fault eigenvector is accurate and comprehensive, so get The target solution is also accurate.

Embodiment Three

This embodiment is a further refinement of any of the above embodiments. In this embodiment, the solution prediction model that has been trained to convergence also includes a linear layer; the linear layer includes a preset prediction label formula.

Wherein, the preset prediction label formula may be a Softmax function. The preset prediction label formula is a prediction label formula that has been trained to convergence and obtained after training. During the training process, the parameters in the predicted label formula are trained, and after the convergence condition is reached, the final preset predicted label formula is obtained.

FIG. 4 is a schematic flowchart of a method for predicting a network fault solution provided in Embodiment 3 of the present application. This embodiment is a further refinement of any of the above-mentioned embodiments. This embodiment is an optional way to predict the solution of the target network according to multiple network fault feature vectors to obtain the target solution, as shown in FIG. 4 The specific steps are as follows.

S401. Calculate a network fault fusion feature vector based on a network fault weight formula and multiple network fault feature vectors.

Wherein, the network fault weight formula is preset, wherein the weight can be configured according to its own requirements or the importance of each network fault feature vector, which is not limited here.

In one manner, the network failure weight formula can be shown in formula (1):

(1) V＝ω ₁ V ₁ +ω ₂ V ₂ +...+ω _i V _i

Among them, V represents the network fault fusion feature vector, Vi represents the network fault feature vector, i represents the number of twin Bert models, and ω _i represents the corresponding weight parameters.

Specifically, the predicting device inputs a plurality of network fault feature vectors into the network fault weight formula, so as to calculate the network fault fusion feature vector. Among them, the network fault fusion feature vector is a fixed-length vector.

S402. Input the network fault fusion feature vector into a preset prediction label formula to calculate a target prediction label value.

In one manner, the preset predictive label formula can be shown in formula (2):

(2) Y=f(WV+b)

Among them, f(V) represents the activation function, which can be a Softmax function, V represents the network fault fusion feature vector, b represents the preset deviation, W represents the weight matrix of the linear layer, and Y represents the output of the linear layer, which is a fixed-length vector , that is, Y=[Y ₁ ,...Y _m ] ^T , where Y _m is an element of Y, and m is the number of elements.

Specifically, the network fault fusion feature vector V is input into (2) to obtain the output Y of the linear layer, and then a maximum element Y _max is determined from multiple elements of Y, and it is used as the target prediction label value.

S403. Determine a corresponding target predicted label range based on the target predicted label value.

Wherein, the target prediction label range is a range to which the target prediction label value belongs, and the prediction label range may be determined according to the total number of preset solutions.

Specifically, Y is the output of the Softmax function, the value of Y _m is between [0,1], all elements in Y are normalized, and the sum of all elements is equal to 1. Assuming that the total number of preset solutions is n, the interval [0,1] is divided into n interval ranges, each interval range is a predicted label range, and each predicted label range corresponds to the preset solution to which it belongs.

Exemplarily, assuming that the total number of preset solutions is 5, the interval [0,1] is divided into 5 interval ranges, specifically, [0,0.2], (0.2,0.4], (0.4,0.6], ( 0.6,0.8] and (0.8,1]. At the same time, each predicted label range corresponds to a preset solution.

Further, the corresponding target predicted label range is determined according to the target predicted label value. According to the above exemplary example, assuming that the target predicted label value is 0.88, the target predicted label range is determined to be (0.8, 1].

S404. Determine a corresponding target solution based on the target prediction label range.

Exemplarily, in the mapping relationship, the target prediction label range (0.8, 1] corresponds to the preset solution 10, then the preset solution 10 is determined as the target solution corresponding to the target prediction label range.

This embodiment provides a method for predicting network failure solutions. When predicting the solution of the target network according to multiple network failure feature vectors to obtain the target solution, it specifically includes: based on the network failure weight formula and multiple network failures Calculate the network fault fusion feature vector from the fault feature vector; input the network fault fusion feature vector into the preset prediction label formula to calculate the target prediction label value; determine the corresponding target prediction label range based on the target prediction label value; Label ranges determine the corresponding target solutions. The prediction device in this embodiment first calculates the network fault fusion feature vector based on the network fault weight formula and multiple network fault feature vectors, and then inputs the network fault fusion feature vector into the preset prediction label formula to calculate the target prediction label value, because The target prediction label value belongs to a certain target prediction label range, and each target prediction label range corresponds to a solution. Therefore, the prediction device will determine the target prediction label range according to the target prediction label value, so that the target can be accurately determined. solution.

In one way, this way is an optional way to determine the corresponding target solution based on the target prediction label range, and the specific steps are as follows.

Obtain the mapping relationship between pre-stored preset predicted label ranges and their corresponding solutions.

Wherein, the corresponding solutions may be determined by experienced experts.

Wherein, the mapping relationship may be expressed in a table form, or in other ways, which is not limited here.

Specifically, the predicting device obtains a preset mapping relationship between a predicted label range and its corresponding solution from its own storage area, and the mapping relationship may include multiple predicted label ranges and corresponding solutions.

Exemplarily, assuming that the total number of preset solutions is n, the mapping relationship is shown in Table 1 below.

Table 1: The mapping relationship between the preset prediction label range and its corresponding solution

Preset predicted label range corresponding solution [0,1/n] Solution 1 (1/n,2/n] Solution 2 (3/n,4/n] Solution 3 ... solution…… (n-1/n,1] Solution n

In response to that the target predicted label range is consistent with any preset predicted label range in the mapping relationship, a solution corresponding to the consistent preset predicted label range is determined as the target solution.

According to the above illustrative example, the prediction device compares the target predicted label range with the preset predicted label range in Table 1, and finds that any preset predicted label range in Table 1 is consistent with the target predicted label range, for example, The target prediction label range is (3/n,4/n], then the corresponding preset prediction label range in Table 1 is (3/n,4/n], and then the preset prediction label range (3/n ,4/n] corresponds to solution 3, so that solution 3 is determined as the target solution.

When the method determines the corresponding target solution based on the target predicted label range, it specifically includes: obtaining the mapping relationship between the pre-stored preset predicted label range and its corresponding solution; responding to the target predicted label range and the mapping relationship Any preset predicted label range is consistent, and the solution corresponding to the consistent preset predicted label range is determined as the target solution. The prediction device in this embodiment obtains the mapping relationship between the preset prediction label range and its corresponding solution. Since the mapping relationship can accurately reflect the solution corresponding to the preset prediction label range, it responds to the target prediction label range and the mapping relationship. The preset prediction label ranges are consistent, and then the solution corresponding to the consistent preset prediction label range is determined as the target solution, so that the determined target solution is accurate.

Embodiment Four

FIG. 5 is a schematic flowchart of a method for predicting a network failure solution provided in Embodiment 4 of the present application. This embodiment is a further refinement of any of the above-mentioned embodiments. This embodiment is an optional method before using a solution prediction model that has been trained to converge to predict the solution of the target network according to the fault data of the target network, such as As shown in Figure 5, the specific steps are as follows.

S501. Obtain a training sample set; the network fault data in the training sample set comes from the historical network fault data sent by the user terminal and the network fault data in the expert experience knowledge base; the training sample set includes multiple network fault data and the corresponding network fault data The predicted label value for the solution of .

Wherein, the historical network fault data refers to the data sent by the user terminal before. The expert experience knowledge base is the data summed up by the operator's experts.

Wherein, the predicted label value of the solution corresponding to each network fault data included in the training sample set may be pre-processed and manually set. Wherein, the predicted label value of the solution corresponding to the network fault data can reflect the solution of the network fault data.

Specifically, the prediction device acquires network fault data from the user terminal and the expert experience knowledge base, and then the prediction device displays the network fault data to the expert through the display, and the expert inputs the predicted label value of the set solution and sends it to the prediction device, and then The network fault data set includes a plurality of network fault data and a predicted label value of a solution corresponding to each network fault data.

S502. Using the training sample set to train multiple twin preset Bert models.

Specifically, input all the network fault data in the training sample set and the predicted label values of the solutions corresponding to each network fault data to multiple twin preset Bert models for training, so that the preset Bert model can optimize the model parameters based on the training sample set .

S503. Determine multiple twin Bert models satisfying preset training convergence conditions as multiple twin Bert models trained to convergence.

When the preset training convergence conditions are met, multiple twin Bert models that have been trained to convergence are obtained.

This embodiment provides a method for predicting a network fault solution. Before using the solution prediction model that has been trained to converge to predict the solution of the target network according to the target network fault data, it specifically includes: obtaining a training sample set; The centralized network fault data comes from the historical network fault data sent by the user terminal and the network fault data in the expert experience knowledge base; the training sample set contains multiple network fault data and the predicted label value of the solution corresponding to each network fault data; The training sample set trains multiple twin preset Bert models; multiple twin Bert models satisfying preset training convergence conditions are determined as multiple twin Bert models that have been trained to convergence. The prediction device in this embodiment first obtains a training sample set, then trains multiple twin preset Bert models based on the training sample set, and then determines the twin Bert models that meet the training convergence conditions as multiple twin Bert models that have been trained to convergence , because the network fault data in the training sample set comes from the historical network fault data sent by the user terminal and the network fault data in the expert experience knowledge base, and the training sample set also contains the predicted label value of the solution corresponding to each network fault data, so the training The source of the sample set is wide and comprehensive, and the data conforms to the actual situation of the user. Therefore, the twin Bert model that has been trained to convergence is better, not only professional, but also in line with the actual situation of the user

In one manner, this manner is an optional manner after determining multiple twin Bert models that meet the preset training convergence conditions as multiple twin Bert models that have been trained to convergence, and the specific content is as follows.

Construct a fusion model of multiple twin Bert models based on multiple twin Bert models that have been trained to convergence and linear layers.

A fusion model of multiple Siamese Bert models is determined as the solution prediction model trained to convergence.

FIG. 6 is a schematic diagram of a solution prediction model trained to convergence provided in Embodiment 4 of the present application. As shown in Figure 6, multiple twin Bert models that have been trained to convergence are Bert model 1, Bert model 2, Bert model 3, and Bert model 4. Multiple twin Bert models that have been trained to convergence and a linear layer construct a A fused model is a solution prediction model that has been trained to convergence. Among them, the configuration parameters of multiple twin Bert models that have been trained to convergence are different.

In this method, after determining multiple twin Bert models that meet the preset training convergence conditions as multiple twin Bert models that have been trained to convergence, it specifically includes: constructing multiple A fusion model of twin Bert models; the fusion model of multiple twin Bert models is determined as a solution prediction model that has been trained to convergence. The prediction device builds a fusion model of multiple twin Bert models that have been trained to convergence and a linear layer. This fusion model is a solution prediction model that has been trained to convergence. Because the solution prediction that has been trained to convergence The model contains multiple twin Bert models and linear layers that have been trained to convergence, so the solution prediction model is a better model.

Embodiment five

FIG. 7 is a schematic flowchart of a method for predicting a network failure solution provided in Embodiment 5 of the present application. This embodiment is a further refinement of any of the above embodiments, as shown in FIG. 7 , and the specific steps are as follows.

S701. Obtain a network fault update data set; the network fault update data in the network fault update data set comes from the network fault update data sent by the user terminal and the network fault update data in the expert experience knowledge base; the network fault update data set includes multiple network faults The fault update data and the predicted label value of the solution corresponding to each network fault update data.

Specifically, the forecasting device obtains the network fault update data from the user terminal and the expert experience knowledge base, and then the forecasting device displays the network fault update data to the expert through the display, and the expert inputs the predicted label value of the set solution and sends it to the predicting device , and further the network fault update data set includes a plurality of network fault update data and the predicted label value of the solution corresponding to each network fault update data.

S702. Perform update training on multiple twin Bert models that have been trained to convergence by using the network fault update data set.

Specifically, input all network fault update data in the network fault update data set and the predicted label value of the solution corresponding to each network fault update data into the twin Bert model that has been trained to convergence for update training, so that the trained to convergent The model parameters in the twin Bert model are further optimized.

S703. Determine multiple twin Bert models satisfying preset update training conditions as multiple updated twin Bert models that have been trained to convergence.

Wherein, the preset update training condition is a pre-set condition specially for the update training.

Specifically, when multiple twin Bert models meet the preset update training conditions, the twin Bert models that have been trained to convergence will be updated, and the model parameters contained in them will be more optimized.

Further, the predicting device determines the multiple twin Bert models satisfying the preset updating training conditions as multiple updated twin Bert models that have been trained to convergence.

S704. Construct a fusion model of multiple updated twin Bert models based on the multiple updated twin Bert models trained to convergence and the linear layer.

Specifically, the prediction device will reselect multiple updated twin Bert models that have been trained to convergence and the linear layer to construct a fusion model of multiple updated twin Bert models. Therefore, the final fusion model of the updated twin Bert model is the latest and more in line with actual needs.

In one manner, the prediction device can also update and train the linear layer, so as to obtain a better preset prediction label formula.

S705. Determine the fusion model of multiple updated twin Bert models as the latest solution prediction model that has been trained to convergence.

This embodiment provides a prediction method for a network fault solution, which specifically further includes: obtaining a network fault update data set; the network fault update data in the network fault update data set is derived from the network fault update data sent by the user terminal and the expert experience knowledge base The network fault update data in ; the network fault update data set contains multiple network fault update data and the predicted label value of the solution corresponding to each network fault update data; the network fault update data set is used to train multiple twin Berts that have been trained to converge The model is updated and trained; multiple twin Bert models that meet the preset update training conditions are determined as multiple updated twin Bert models that have been trained to convergence; based on multiple updated twin Bert models that have been trained to convergence and linear The fusion model of multiple updated twin Bert models is constructed in the layer; the fusion model of multiple updated twin Bert models is determined as the latest solution prediction model that has been trained to convergence. The prediction device in this embodiment can obtain the network fault update data set, so as to perform update training based on the network fault update data contained in the network update data set and the predicted label value of the solution corresponding to each network fault update data, and then can construct the latest existing The solution prediction model trained to convergence, due to the updated training, can obtain the latest solution prediction model that has been trained to convergence, which is more in line with the actual situation; at the same time, due to the network fault update data set, the network fault data comes from the user The network fault update data sent by the terminal and the network fault update data in the expert experience knowledge base, so the latest solution prediction model that has been trained to convergence is also more comprehensive.

Embodiment six

The following are device embodiments of the present application. FIG. 8 is a schematic diagram of a prediction device for a network failure solution provided in Embodiment 6 of the present application. As shown in FIG. 8 , the device 800 includes the following modules.

The receiving module 801 is configured to receive target network failure data corresponding to the target network sent by the user terminal;

The prediction module 802 is used to predict the solution of the target network according to the fault data of the target network by adopting the solution prediction model trained to convergence to obtain the target solution; the solution prediction model trained to convergence is a pre-built multiple A fusion model of the bidirectional encoder representation Bert model of Siamese transformers;

The sending module 803 is configured to send the target solution to the user terminal, so as to instruct the user terminal to display the target solution.

In one way, the solution prediction model that has been trained to convergence contains multiple twin Bert models that have been trained to convergence; different twin Bert models have different levels of configuration parameters; configuration parameters are the Bert model's grasp of semantic granularity Get the parameters of the feature; the prediction module 802, when using the solution prediction model trained to convergence to predict the solution of the target network according to the fault data of the target network, to obtain the target solution, is specifically used for:

Input the target network fault data into the twin Bert models that have been trained to convergence; respectively use the twin Bert models that have been trained to convergence to extract the features of the corresponding semantic granularity of the target network fault data to obtain the corresponding network fault features vector; the solution of the target network is predicted based on multiple network fault feature vectors to obtain the target solution.

In one manner, the solution prediction model that has been trained to converge also includes a linear layer; the linear layer includes a preset prediction label formula; the prediction module 802 performs the target network solution according to a plurality of network fault feature vectors Prediction, when obtaining the target solution, is specifically used for:

Calculate the network fault fusion feature vector based on the network fault weight formula and multiple network fault feature vectors; input the network fault fusion feature vector into the preset prediction label formula to calculate the target prediction label value;

A corresponding target prediction label range is determined based on the target prediction label value; a corresponding target solution is determined based on the target prediction label range.

In one manner, the prediction module 802, when determining the corresponding target solution based on the target prediction label range, is specifically used to:

Acquiring a pre-stored mapping relationship between a preset predicted label range and its corresponding solution; in response to the target predicted label range being consistent with any preset predicted label range in the mapping relationship, corresponding to the consistent preset predicted label range The solution of is identified as the target solution.

In one manner, before using the solution prediction model that has been trained to converge to predict the solution of the target network according to the target network fault data, this embodiment provides a network fault solution prediction device, which further includes: obtaining module, training module, and determination module.

Among them, the obtaining module is used to obtain the training sample set; the network fault data in the training sample set comes from the historical network fault data sent by the user terminal and the network fault data in the expert experience knowledge base; the training sample set contains multiple network fault data and The predicted label value of the solution corresponding to each network fault data; the training module is used to train the preset Bert model of multiple twins using the training sample set; the determination module is used to combine multiple twins that meet the preset training convergence conditions The Bert model is determined to be multiple twin Bert models that have been trained to convergence.

In one manner, after determining multiple twin Bert models that meet the preset training convergence conditions as multiple twin Bert models that have been trained to converge, this embodiment provides a network fault solution prediction device, which also includes : Building blocks.

Among them, the construction module is used to construct the fusion model of multiple twin Bert models based on multiple twin Bert models trained to convergence and the linear layer; the determination module is also used to determine the fusion model of multiple twin Bert models as already Solution prediction model trained to convergence.

In one manner, this embodiment provides an apparatus for predicting network fault solutions.

Wherein, the obtaining module is also used to obtain the network fault update data set; the network fault update data in the network fault update data set comes from the network fault update data sent by the user terminal and the network fault update data in the expert experience knowledge base; the network fault update The data set contains multiple network fault update data and the predicted label value of the solution corresponding to each network fault update data; the training module is also used to update and train multiple twin Bert models that have been trained to convergence using the network fault update data set The determination module is also used to determine multiple twin Bert models that meet the preset update training conditions as multiple updated twin Bert models that have been trained to converge; the building block is also used to train based on multiple updated Convergent twin Bert models and linear layers construct a fusion model of multiple updated twin Bert models; the determination module is also used to determine the fusion model of multiple updated twin Bert models as the latest one that has been trained to convergence Solution prediction model.

Embodiment seven

FIG. 9 is a schematic structural diagram of an electronic device provided in Embodiment 7 of the present application. As shown in FIG. 9 , the electronic device 900 may include: a processor 901 , and a memory 902 communicatively connected to the processor 901 . Wherein, the memory 902 stores computer-executable instructions; the processor 901 executes the computer-executable instructions stored in the memory 902, so as to implement any one of the method embodiments as in Embodiment 1 to Embodiment 5 above. The specific implementation methods and technical effects are similar and will not be repeated here.

Wherein, in this embodiment, the memory 902 and the processor 901 are connected through a bus. The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into address bus, data bus, control bus and so on. For ease of representation, only one thick line is used in FIG. 9 , but it does not mean that there is only one bus or one type of bus.

Embodiment Eight

The present application provides a computer-readable storage medium, in which computer-executable instructions are stored, and when the computer-executable instructions are executed by a processor, they are used to implement any method embodiment as in the above-mentioned embodiment 1 to embodiment 5, and the specific implementation The method and technical effect are similar and will not be repeated here.

Other embodiments of the present application will be readily apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any modification, use or adaptation of the application, these modifications, uses or adaptations follow the general principles of the application and include common knowledge or conventional technical means in the technical field not disclosed in the application . The specification and examples are to be considered exemplary only, with a true scope and spirit of the application indicated by the following claims.

It should be understood that the present application is not limited to the precise constructions which have been described above and shown in the accompanying drawings, and various modifications and changes may be made without departing from the scope thereof. The scope of the application is limited only by the appended claims.

Claims (10)

1. A method of predicting a network failure solution, the method comprising:

receiving target network fault data corresponding to a target network sent by a user terminal;

predicting the solution of the target network according to the target network fault data by adopting a solution prediction model trained to be converged so as to obtain a target solution; the solution prediction model trained to be converged is a fusion model of a bi-directional encoder characterization quantity Bert model of a plurality of twin converters which are constructed in advance;

And sending the target solution to the user terminal to instruct the user terminal to display the target solution.

2. The method of claim 1, wherein the trained-to-converged solution prediction model comprises a plurality of trained-to-converged twin Bert models; different twin Bert models have different hierarchical configuration parameters; the configuration parameters are parameters of the Bert model on grabbing features of semantic granularity;

the predicting the solution of the target network according to the target network fault data by adopting the solution prediction model trained to be converged so as to obtain the target solution comprises the following steps:

respectively inputting the target network fault data into each trained-to-converge twin Bert model;

respectively adopting each trained to-be-converged twin Bert model to perform feature extraction of corresponding semantic granularity on the target network fault data so as to obtain corresponding network fault feature vectors;

and predicting the solution of the target network according to a plurality of network fault feature vectors to obtain the target solution.

3. The method of claim 2, wherein the solution prediction model trained to converge further comprises a linear layer; the linear layer comprises a preset predictive label formula;

Predicting the solution of the target network according to the network fault feature vectors to obtain the target solution, including:

calculating a network fault fusion feature vector based on a network fault weight formula and a plurality of network fault feature vectors;

inputting the network fault fusion feature vector into the preset predictive label formula to calculate a target predictive label value;

determining a corresponding target prediction tag range based on the target prediction tag value;

and determining a corresponding target solution based on the target prediction tag range.

4. The method of claim 3, wherein the determining a corresponding target solution based on the target prediction tag range comprises:

acquiring a pre-stored preset predictive label range and a mapping relation of a corresponding solution;

and responding to the fact that the target prediction tag range is consistent with any preset prediction tag range in the mapping relation, and determining a solution corresponding to the consistent preset prediction tag range as a target solution.

5. The method of any of claims 1-4, wherein prior to predicting a solution to a target network from the target network failure data using the trained solution prediction model, further comprising:

Acquiring a training sample set; the network fault data in the training sample set is derived from historical network fault data sent by the user terminal and network fault data in the expert experience knowledge base; the training sample set comprises a plurality of network fault data and predictive label values of solutions corresponding to the network fault data;

training a plurality of twin preset Bert models by adopting the training sample set;

and determining a plurality of twin Bert models meeting preset training convergence conditions as a plurality of twin Bert models trained to converge.

6. The method according to claim 5, wherein after determining the plurality of twin Bert models satisfying the preset training convergence condition as the plurality of twin Bert models trained to converge, further comprising:

constructing a fusion model of a plurality of twin Bert models based on the twin Bert models trained to be converged and the linear layer;

determining a fusion model of the plurality of twin Bert models as a solution prediction model trained to converge.

7. The method according to claim 1, wherein the method further comprises:

acquiring a network failure update data set; the network fault updating data in the network fault updating data set is derived from the network fault updating data sent by the user terminal and the network fault updating data in the expert experience knowledge base; the network fault updating data set comprises a plurality of network fault updating data and prediction tag values of solutions corresponding to the network fault updating data;

Updating and training a plurality of twin Bert models trained to be converged by adopting the network fault updating data set;

determining a plurality of twin Bert models meeting preset updating training conditions as a plurality of updated twin Bert models trained to be converged;

constructing a fusion model of a plurality of updated twin Bert models based on the updated twin Bert models trained to be converged and the linear layer;

and determining a fusion model of the updated twin Bert models as a latest trained-to-converged solution prediction model.

8. A network failure solution prediction apparatus, the apparatus comprising:

the receiving module is used for receiving target network fault data corresponding to a target network sent by the user terminal;

the prediction module is used for predicting the solution of the target network according to the target network fault data by adopting the solution prediction model trained to be converged so as to obtain the target solution; the solution prediction model trained to be converged is a fusion model of a bi-directional encoder characterization quantity Bert model of a plurality of twin converters which are constructed in advance;

and the sending module is used for sending the target solution to the user terminal so as to instruct the user terminal to display the target solution.

9. An electronic device, comprising: a processor, and a memory communicatively coupled to the processor;

the memory stores computer-executable instructions;

the processor executes computer-executable instructions stored in the memory to implement the method of any one of claims 1-8.

10. A computer readable storage medium having stored therein computer executable instructions which when executed by a processor are adapted to carry out the method of any one of claims 1-8.

Images