A neural network-based ipv6 network renovation problem identification model generation method

Abstract

The application relates to an artificial intelligence technology and discloses a neural network-based IPv6 network reconstruction problem identification model generation method, which comprises the following steps: collecting IPv6 support data of each key node in a network and in the network; performing IPv6 reconstruction problem analysis on the collected data based on a neural network; and outputting the collected data for manual analysis of the IPv6 reconstruction problem; comparing the model analysis result with the manual analysis result to detect whether there is a result deviation; if yes, iteratively training the neural network based on the result deviation; and if no, solidifying the adjustment parameters and the training result of the neural network to generate an IPv6 network reconstruction problem identification model. The application also discloses a control device, a computer device and a computer readable storage medium. The application aims to generate a special model capable of accurately identifying IPv6 network reconstruction problems, so as to improve the accuracy and efficiency of IPv6 network reconstruction problem identification.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method for generating a model for identifying IPv6 network transformation problems based on neural networks, a control device, a computer device, and a computer-readable storage medium. Background Technology

[0002] With the rapid development of the internet, IPv4 address resources are becoming increasingly scarce. IPv6, as the next-generation internet protocol, boasts numerous advantages such as a vast address space, high security, and excellent mobility, making it an inevitable trend for future internet development. During the IPv6 network transformation process, accurately identifying and resolving various problems is crucial to ensuring the smooth progress of the transformation and guaranteeing stable network operation.

[0003] Currently, the identification of most IPv6 network transformation issues relies primarily on the human experience and expertise of network administrators. However, with the continuous expansion of network scale and the increasing complexity of network structures, the efficiency and accuracy of manual analysis face significant challenges. Manual analysis is not only time-consuming and labor-intensive, but also prone to omissions and misjudgments, failing to promptly identify and resolve potential problems.

[0004] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main objective of this application is to provide a method, control device, computer equipment, and computer-readable storage medium for generating a model for identifying IPv6 network transformation problems based on neural networks. The aim is to generate a dedicated model that can accurately identify IPv6 network transformation problems, thereby improving the accuracy and efficiency of IPv6 network transformation problem identification.

[0006] To achieve the above objectives, this application provides a method for generating an IPv6 network transformation problem identification model based on neural networks, comprising the following steps:

[0007] Deploy IPv6 monitoring points to collect IPv6 support data from key network nodes and other network components, and then clean, organize, and store the data.

[0008] Based on a neural network pre-trained for identifying IPv6 network transformation problems, the collected data is analyzed for IPv6 transformation problems to generate model analysis results; and the collected data is output for human analysis of IPv6 transformation problems to generate human analysis results.

[0009] The model analysis results are compared with the human analysis results to detect any discrepancies.

[0010] If so, based on the result deviation, the neural network is iteratively trained, and the process is returned to execute the step of analyzing the collected data for IPv6 transformation problems based on the neural network that has been pre-trained for IPv6 network transformation problem identification, and generating model analysis results.

[0011] If not, the adjustment parameters and training results of the neural network are solidified to generate an IPv6 network transformation problem identification model.

[0012] To achieve the above objectives, this application also provides a control device, comprising:

[0013] The monitoring module is used to deploy IPv6 monitoring points, collect IPv6 support data from key nodes in the network, and clean and organize the data before storing it in the database.

[0014] The analysis module is used to analyze the collected data for IPv6 transformation issues based on a neural network that has been pre-trained for IPv6 network transformation issue identification, and generate model analysis results.

[0015] The output module is used to output the collected data for manual analysis of IPv6 transformation issues and generate human analysis results.

[0016] The detection module is used to compare the model analysis results with the human analysis results to detect whether there is any result deviation.

[0017] The iteration module is used to iteratively train the neural network based on the result deviation if the result is true, and return to the step of performing IPv6 transformation problem analysis on the collected data based on the neural network that has been pre-trained for IPv6 network transformation problem identification, and generating model analysis results.

[0018] The generation module is used to solidify the adjustment parameters and training results of the neural network if otherwise, and generate an IPv6 network transformation problem identification model.

[0019] To achieve the above objectives, this application also provides a computer device, the computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, it implements the steps of the above-described method for generating an IPv6 network transformation problem identification model based on a neural network.

[0020] To achieve the above objectives, this application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for generating an IPv6 network transformation problem identification model based on a neural network.

[0021] The method, control device, computer equipment, and computer-readable storage medium for generating IPv6 network transformation problem identification models based on neural networks provided in this application employ a pre-trained neural network to analyze collected data, enabling rapid and efficient generation of model analysis results. Simultaneously, it outputs data for human analysis, combining the efficiency of machine analysis with the flexibility of human analysis, resulting in more comprehensive problem identification. Furthermore, comparing the model analysis results with human analysis results allows for timely detection of discrepancies. If a discrepancy is found, the neural network is iteratively trained to continuously optimize model performance and improve the accuracy of problem identification. When the results are unbiased, the adjusted parameters of the neural network and the training results are solidified, generating a stable and reliable IPv6 network transformation problem identification model. This model can be effectively applied to the identification of practical IPv6 network transformation problems, thereby improving the accuracy and efficiency of IPv6 network transformation problem identification. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the steps in a method for generating an IPv6 network transformation problem identification model based on a neural network in one embodiment of this application;

[0023] Figure 2 This is a schematic diagram of the control device in one embodiment of this application;

[0024] Figure 3 This is a schematic diagram of the internal architecture of a computer device according to an embodiment of this application.

[0025] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0026] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0027] Furthermore, descriptions using terms such as "first" and "second" in this application are for descriptive purposes only (e.g., to distinguish identical or similar features) and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" and "second" may explicitly or implicitly include at least one of those features. Additionally, technical solutions from different embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If a combination of technical solutions is contradictory or impossible to implement, such a combination should be considered nonexistent and not within the scope of protection claimed in this application.

[0028] Reference Figure 1 In one embodiment, the method for generating an IPv6 network transformation problem identification model based on neural networks includes:

[0029] Step S10: Deploy IPv6 monitoring points, collect IPv6 support data from key network nodes and other network nodes, and clean and organize the data before storing it in the database.

[0030] Step S20: Based on the neural network that has been pre-trained for IPv6 network transformation problem identification, analyze the collected data for IPv6 transformation problems and generate model analysis results; and Step S30: Output the collected data for manual analysis of IPv6 transformation problems and generate human analysis results.

[0031] Step S40: Compare the model analysis results with the human analysis results to detect any discrepancies.

[0032] Step S50: If so, then based on the result deviation, iteratively train the neural network and return to the step of performing the IPv6 transformation problem analysis on the collected data based on the neural network that has been pre-trained for IPv6 network transformation problem identification, and generating model analysis results.

[0033] Step S60: If not, then solidify the adjustment parameters and training results of the neural network to generate an IPv6 network transformation problem identification model.

[0034] In this embodiment, the execution terminal can be a computer device or other device or apparatus (such as a control device) that controls the computer device.

[0035] As described in step S10, taking into account factors such as network topology and service distribution, IPv6 monitoring points are rationally deployed at key locations in the network to ensure comprehensive and accurate collection of various types of data. Specific deployment locations include, but are not limited to, core network switches, routers, access layer switches and wireless access points, perimeter firewalls, and intrusion detection systems.

[0036] Through the deployed IPv6 monitoring points, comprehensive data on various indicators such as network traffic, active connections, network quality, terminals, and applications are collected. The collected data covers the following aspects:

[0037] (1) Total traffic: Collect the total IPv4 and IPv6 network traffic of key nodes such as Internet exit, data center exit, core switch, office network exit router or switch to understand the overall data transmission scale of the network.

[0038] (2) Specific application traffic: Obtain the specific application traffic of IPv4 and IPv6 at key nodes, such as the traffic of web browsing, video streaming, file transfer, etc., to facilitate the analysis of the network resource usage of different applications.

[0039] (3) Active link data: Statistical analysis of the number of active IPv4 and IPv6 links of key nodes and applications to understand the number of connections communicating in the network and to assess the network’s activity and busyness.

[0040] (4) Network quality data: Monitor the quality of IPv4 and IPv6 networks between key nodes and from key nodes to designated locations, including indicators such as latency, packet loss rate, and bandwidth utilization, in order to evaluate the performance and stability of the network.

[0041] (5) Terminal device data: Perform IPv4 and IPv6 address scanning on terminal devices such as PCs, cameras, and printers to obtain the support of these terminal devices for IPv4 and IPv6 and to determine the network compatibility of the terminal devices.

[0042] (6) Application support data: Collect information on the support of applications such as databases, operating systems, DHCP, and business systems for IPv4 and IPv6, and understand the operation status of applications under different network protocols.

[0043] The collected data may contain noise, duplicates, errors, etc. To ensure data quality, cleaning operations are required.

[0044] Optionally, filtering, smoothing, and other methods can be used to remove random noise data caused by factors such as network environment interference and sensor errors.

[0045] Optionally, duplicate data records can be identified and deleted by comparing key information in the data, such as timestamps and device identifiers, to avoid data redundancy.

[0046] Optionally, check for logical errors, formatting errors, and other issues in the data, and correct or delete erroneous data based on the data context and business rules.

[0047] The cleaned data needs to be standardized to have a uniform format and structure, which will facilitate subsequent storage and analysis.

[0048] Optionally, data from different sources and in different formats can be converted into a unified standard format, such as a unified date and time format, and standardized units and precision of numerical data.

[0049] Optionally, based on the type and purpose of the data, it can be divided into different categories such as traffic data, active link data, and network quality data for easier management and querying.

[0050] Optionally, non-numerical data can be encoded into a numerical form that is easy for computers to process, such as encoding device status information into numeric codes.

[0051] Choose a suitable database management system, such as a relational database (MySQL, Oracle, etc.) or a non-relational database (MongoDB, Redis, etc.). Design a reasonable database table structure based on the data classification and organization results, determining the table fields, data types, indexes, and other information. Finally, import the cleaned and organized data into the database in batches, and create corresponding indexes and views for subsequent analysis and use.

[0052] As described in step S20, the neural network is connected to the IPv6 monitoring system, which is linked to a relational database. The relational database stores the collected data that has been cleaned and organized in step S10. This collected data is obtained by collecting and processing data from key network nodes and IPv6 support data within the network. It forms a specific data structure in the relational database, reflecting various aspects of the network structure, IPv4 / IPv6 status, and more. This information covers network topology, node connection status, traffic distribution of different protocols, and device support for IPv6, providing a comprehensive and detailed data foundation for subsequent analysis.

[0053] The neural network was pre-trained for IPv6 network transformation problem identification and performs analysis based on a "dedicated model for IPv6 network transformation problem identification." This dedicated model was trained on a large amount of historical data and labeled problem samples, and it can learn the relationship between various problems and related data features during the IPv6 network transformation process.

[0054] The neural network performs in-depth analysis of the network's IPv6 transformation based on collected data. It extracts various features from the data, such as the proportion of IPv6 traffic in different time periods, device protocol support status, network latency, and packet loss rate, and inputs these features into a dedicated model for calculation and judgment. Through the model's operations, the neural network can identify deficiencies in the network's IPv6 transformation. For example, if it finds that the proportion of IPv6 traffic is low, the neural network will further analyze possible reasons. These could be that the current terminal does not support the IPv6 protocol, or although the terminal supports it, it has not enabled the IPv6 protocol. It could also be due to poor IPv6 network quality, such as high latency and packet loss, affecting users' use of the IPv6 network and thus leading to a low proportion of IPv6 traffic. Finally, the neural network outputs these identified deficiencies as the model analysis results.

[0055] After the model analysis is completed, the results will be output. The results represent the deficiencies in IPv6 network upgrades identified by the neural network, presented in a structured format for easy processing and comparison later.

[0056] As described in step S30, in addition to using neural networks for model analysis, it is also necessary to output collected data and analyze it manually. Human analysts will work in conjunction with a knowledge base containing various professional knowledge, common problems, and solutions in the field of IPv6 network transformation.

[0057] Optionally, human analysts will conduct a comprehensive and detailed review and study of the collected data. Based on their professional experience and knowledge base, they will uncover potential IPv6 transformation issues from the data. For example, they might examine device configuration files to determine if configuration errors are causing IPv6 malfunctions; or analyze time-series network traffic data to identify the causes of abnormal traffic fluctuations. Human analysis is more flexible and in-depth, taking into account factors that models might overlook, thus complementing the results of model analysis.

[0058] After manual analysis, the results will be re-entered into the system. The results of the manual analysis are the IPv6 transformation issues identified by the analysts, and may be presented in a report detailing the problem's manifestation, possible causes, and suggested solutions.

[0059] As described in step S40, the system reads the model analysis results and human analysis results from the specified storage location. The model analysis results are usually stored in a database table or a file in a specific format in the form of structured data, containing field information such as network node identifier, problem type, and problem severity; the human analysis results may initially exist in the form of a text report, which the system will parse and convert into structured data suitable for comparison, for example, extracting the problem information described in the text and mapping it to the same field structure as the model analysis results.

[0060] To ensure data consistency and comparability, the system standardizes key data in both results. For problem types, problems with different descriptions but the same underlying principle are uniformly categorized into a standard problem classification system. For example, "Low IPv6 traffic - terminal not enabled" and "Low traffic due to terminal IPv6 function not enabled" are both categorized under "terminal IPv6 protocol not enabled." For problem severity, a unified quantification is performed according to preset level standards, such as assigning "high," "medium," and "low" to specific numerical ranges.

[0061] The system compares the problem types in the model analysis results and the human analysis results one by one according to the data records. For each record, it checks whether the problem types identified in the two results are consistent. If they are consistent, the field is marked as successfully matched; if they are inconsistent, it is marked as having a deviation, and the specific difference is recorded, such as the model identifying it as "Type A problem" while the human analysis identifies it as "Type B problem".

[0062] Optionally, the location of the network node where the problem occurred can be compared. Using the unique identifier of the network node (such as IP address, device number, etc.), check whether the problem node indicated in the two results is the same. If the location information is inconsistent, mark it as a discrepancy and record the specific difference node information.

[0063] Optionally, the standardized severity values ​​of the problems can be compared. An allowable error range is set. If the severity values ​​of the two results are within this error range, the comparison is considered successful; if they exceed the error range, a deviation is determined, and the degree of deviation is recorded.

[0064] For each data record, the system comprehensively judges whether there is a deviation based on the comparison results of each field. If any field fails the comparison, the record is determined to have a deviation and is marked with a deviation flag.

[0065] Furthermore, after traversing all data records, the system counts the number of records with deviations and calculates the proportion of deviation records to the total number of records. Based on a preset deviation threshold, it determines whether there is a significant deviation in the overall result. For example, if the proportion of deviation records exceeds 20%, the overall result is considered to have a large deviation, requiring iterative training of the neural network; if the deviation proportion is within a reasonable range, the result is considered to be basically consistent.

[0066] Optionally, the system will compile the comparison results into a detailed deviation report. The report includes details of the records with deviations, such as differences in problem type, location, and severity, as well as overall deviation statistics, such as the number of deviation records and the deviation percentage.

[0067] After comparing the model analysis results with the human analysis results field by field and judging the overall deviation, the computing device will determine whether there is a result deviation based on the preset rules.

[0068] As described in step S50, if there is a result deviation, the neural network is iteratively trained based on the result deviation.

[0069] Optionally, key information related to the deviation can be extracted from the deviation report, including the collected data corresponding to the records with deviations, model analysis results, and human analysis results. This information will serve as an important basis for subsequent iterative training.

[0070] Optionally, the adjusted bias data can be merged with the original training dataset to expand the scale of the training data. This allows the neural network to learn on more diverse data, improving its generalization ability.

[0071] Optionally, the learning rate of the neural network can be adjusted appropriately based on the severity of the bias and the model's convergence. If the bias is large, the learning rate can be increased to accelerate the model's learning speed; if the model is close to convergence, the learning rate can be decreased to prevent the model from skipping the optimal solution.

[0072] Optionally, the merged training dataset can be used to train the neural network anew. During training, the neural network continuously adjusts its internal weights and bias parameters to minimize the error between the prediction results and the labeled information. Specifically, the backpropagation algorithm is used to calculate the gradient of the error, and the network parameters are updated based on the gradient.

[0073] After each round of training, the model is evaluated using a validation dataset. Metrics such as accuracy and recall on the validation set are calculated to monitor performance changes. If the model's performance on the validation set improves, it indicates that the iterative training has been effective; if performance declines, further adjustments to training parameters or checks of data quality are needed.

[0074] Once the predetermined number of training epochs has been reached or the model performance meets certain requirements, the process returns to steps S20 and S40 to compare the model analysis results with the human analysis results again to check for any remaining bias. If bias still exists, the iterative training process is repeated until the bias is within an acceptable range.

[0075] As described in step S60, when it is determined that there is no result bias, the adjustment parameters (including weights and biases) of the current neural network are saved. These parameters represent the optimal configuration of the neural network in the current training state, which can accurately identify IPv6 network transformation problems.

[0076] Optionally, the saved parameters can be encapsulated with the neural network structure to form a complete, deployable model. This model can be easily integrated into an IPv6 monitoring system for real-time network problem identification.

[0077] After parameter solidification, a formal IPv6 network transformation problem identification model is generated. This model has high accuracy and reliability, and can accurately identify problems in IPv6 network transformation based on collected network data.

[0078] Optionally, the generated recognition model can be deployed to a real production environment to begin supporting IPv6 network transformation efforts. Simultaneously, a monitoring mechanism should be established to regularly evaluate and update the model's performance to adapt to changes in the network environment.

[0079] In one embodiment, a pre-trained neural network is used to analyze the collected data, which can quickly and efficiently generate model analysis results and output data for manual analysis. This combines the efficiency of machine analysis with the flexibility of manual analysis, making problem identification more comprehensive. At the same time, comparing the model analysis results with the manual analysis results can promptly detect deviations. Once a deviation is found, the neural network is iteratively trained to continuously optimize model performance and improve the accuracy of problem identification. When the results are without deviation, the adjustment parameters of the neural network and the training results are solidified to generate a stable and reliable IPv6 network transformation problem identification model, which can be effectively applied to the identification of actual IPv6 network transformation problems, thereby improving the accuracy and efficiency of IPv6 network transformation problem identification.

[0080] In one embodiment, based on the above embodiments, the neural network is constructed using the RGU variant of RNN deep learning technology, and the training process adopts a supervised learning approach.

[0081] In this embodiment, RNN (Recurrent Neural Network) is a type of neural network specifically designed for processing sequential data. Unlike traditional feedforward neural networks, RNNs have a recurrent structure, which allows them to consider previous input information when processing the current input. This characteristic makes RNNs perform exceptionally well when processing sequential data such as time series data and natural language.

[0082] RGU (Recurrent Gating Unit) is a variant of RNN that improves upon the structure of traditional RNNs. RGU introduces a gating mechanism to control the flow of information through gating units, thereby solving the gradient vanishing or gradient exploding problems that may occur in traditional RNNs when processing long sequences.

[0083] The neural network used for identifying IPv6 network transformation issues is based on a variant of the RGU. The network's input layer receives collected IPv6 support data, which may include sequential information such as network traffic and node states.

[0084] In the hidden layer, RGU units are used to process the input data, and a gating mechanism is used to capture long-term dependencies in the data. The network output layer then outputs the analysis results of the IPv6 transformation problem according to the specific task requirements, such as the problem category and severity.

[0085] Before using supervised learning to train a neural network based on an RGU variant, the collected IPv6-supported data needs to be preprocessed. This includes data cleaning, removing noisy data and missing values; data warping, converting the data into a format suitable for network input; and data partitioning, dividing the dataset into training, validation, and test sets.

[0086] The training set is used to train the model, the validation set is used to evaluate the model's performance during training and to adjust hyperparameters such as learning rate and batch size. The test set is used to perform a final evaluation of the model's generalization ability after training is complete.

[0087] During training, the input data from the training set is fed into a neural network based on an RGU variant, and the model outputs a prediction. The prediction is then compared to the corresponding target label, and a loss function is used to measure the difference between the prediction and the target label.

[0088] Commonly used loss functions include cross-entropy loss and mean squared error loss, with the specific choice depending on the type of problem. Cross-entropy loss is typically used for classification problems, while mean squared error loss is typically used for regression problems.

[0089] Next, optimization algorithms (such as stochastic gradient descent, Adam optimizer, etc.) are used to update the network's weight parameters based on the gradient of the loss function. Through continuous iterative training, the model's predictions will gradually approach the target label until satisfactory performance is achieved.

[0090] During training, the model's performance is evaluated using a validation set. If the model's performance on the validation set is unsatisfactory, the hyperparameters or network structure need to be adjusted, and the model needs to be retrained.

[0091] After training, the model's generalization ability is evaluated using a test set. If the model's performance on the test set meets expectations, it indicates that the model has good generalization ability and can be used for practical IPv6 network transformation problem identification tasks.

[0092] In one embodiment, based on the above embodiments, the method for generating an IPv6 network transformation problem identification model based on neural networks further includes:

[0093] During the training of the RGU variant network, quantum gate operations are used to update the network parameters. The network weights and bias parameters are encoded as quantum states, and the gradient descent update process is simulated through a sequence of quantum gates.

[0094] In this embodiment, a quantum state is a state description of a quantum system, represented by a vector in Hilbert space.

[0095] In the RGU variant neural network, the network weights and bias parameters are real numbers, and these parameters can be encoded into quantum states using amplitude encoding.

[0096] Assume the network has n weights and bias parameters [w1, w2, ..., wn]. n These parameters can be encoded into an n-qubit quantum state.

[0097] In traditional neural network training, gradient descent is a commonly used optimization algorithm to update network parameters to minimize the loss function. In quantum computing, quantum gates are the basic units for manipulating quantum states, similar to logic gates in classical computing. To simulate the gradient descent update process of a model, a suitable sequence of quantum gates needs to be designed.

[0098] In quantum systems, a method is needed to compute the gradient of the loss function with respect to the parameters encoded by the quantum state. This can be achieved through quantum measurement and quantum algorithms. For example, a quantum phase estimation algorithm can be used to estimate an approximation of the gradient.

[0099] Once the gradient is estimated, a rotation gate can be designed to update the parameters. A rotation gate can rotate the state of a qubit, thereby changing the amplitude of the quantum state.

[0100] Suppose we want to update the i-th parameter, we can use a single-qubit rotation gate R. i (θ) operates on the qubit encoding this parameter, where θ is related to the estimated gradient and the learning rate. The parameter update process can be simulated through a series of rotating gate operations.

[0101] The implementation process is illustrated below:

[0102] (1) Encode the initial weights and bias parameters of the network into quantum states |ψ0>.

[0103] (2) Input the input data into the neural network based on the RGU variant, obtain the prediction results through quantum measurement, and calculate the loss function L.

[0104] (3) Use quantum algorithms to estimate the gradient of the loss function with respect to the parameters encoded by the quantum state.

[0105] (4) Based on the gradient estimate, select a suitable quantum gate sequence to operate on the quantum state and update the quantum state to |ψ1>.

[0106] (5) The updated quantum state is decoded into a real number by quantum measurement to obtain the updated network weights and bias parameters.

[0107] (6) Repeat the above steps until the loss function converges or the maximum number of iterations is reached.

[0108] Traditional neural network training demands significant computational resources and involves lengthy training times when dealing with large-scale data and complex models. Quantum computing, with its unique quantum mechanical properties such as superposition and entanglement, can significantly accelerate neural network training.

[0109] In one embodiment, based on the above embodiments, the method for generating an IPv6 network transformation problem identification model based on neural networks further includes:

[0110] In the cyclic structure of the RGU variant, entanglement is created between the quantum states corresponding to neurons, so that a change in the state of one neuron can instantly affect other neurons that are entangled with it.

[0111] In this embodiment, in the RGU variant-based neural network, the state of each neuron can be represented by a quantum state. As mentioned earlier, the state information of the neuron can be encoded into the quantum state using methods such as amplitude encoding. For example, the state variable of a neuron can be normalized and represented as the amplitude of a certain qubit in the quantum state. In this way, each neuron in the network corresponds to a specific quantum state.

[0112] To induce entanglement between the quantum states corresponding to neurons, appropriate quantum gate operations need to be designed. In quantum computing, entanglement gates such as the CNOT gate (controlled NOT gate) are available. Through a series of quantum gate operations, originally independent quantum states can be transformed into entangled states.

[0113] Suppose we have two neurons corresponding to quantum states |ψ1> and |ψ2>, which can be manipulated using a CNOT gate. A CNOT gate has a control qubit and a target qubit; when the control qubit is in the |ψ1> state, the target qubit flips. By strategically selecting the control and target qubits and combining them with other single-qubit gate operations, we can entangle these two quantum states.

[0114] For the cyclic structure of the RGU variant, since there are cyclic connections between neurons, a systematic quantum gate operation scheme needs to be designed so that the quantum states corresponding to neurons can gradually become entangled during the cyclic process.

[0115] Optionally, in the cyclic structure of the RGU variant, new input data enters at each time step, and the neuron's state is updated based on the current input and the state of the previous time step. During this process, the entangled quantum states also change with the update of the neuron's state.

[0116] When the state of one neuron changes, due to the properties of quantum entanglement, the states of other neurons entangled with it are instantly affected. This effect propagates continuously within the loop structure, creating a complex network of interconnections between neurons throughout the network.

[0117] Quantum entanglement makes information transmission between neurons more efficient and direct. In traditional neural networks, information propagates gradually through the connection weights between neurons, resulting in a certain delay. However, in the case of quantum entanglement, a change in the state of one neuron can instantly affect other entangled neurons. This allows the network to respond to changes in input data more quickly, improving information processing speed. For IPv6 network transformation problem identification tasks, this means the network can more rapidly detect changes in network state and identify potential problems in a timely manner.

[0118] Furthermore, due to the complex correlations generated by quantum entanglement, the network is able to learn richer features and patterns. When processing IPv6 network data, the network can better capture hidden information and complex dependencies within the data, thereby improving the model's recognition accuracy. For example, in identifying IPv6 address allocation problems, quantum entanglement can help the network discover potential correlations between different nodes, thus more accurately determining whether address allocation anomalies exist.

[0119] In one embodiment, based on the above embodiments, the method for generating an IPv6 network transformation problem identification model based on neural networks further includes:

[0120] Before iteratively training the neural network, the deep learning parameters of the neural network are adjusted. The adjusted parameters include at least one of the following: learning rate adjustment during learning, ordered learning rate adjustment, equal interval adjustment, multi-interval adjustment, and cosine annealing parameters.

[0121] In this embodiment, the learning rate is a key hyperparameter in neural network training, which determines the step size of each parameter update. Dynamically adjusting the learning rate during the learning process involves changing its magnitude according to different training stages.

[0122] In the early stages of training, the model's parameters differ significantly from the optimal values. At this point, a larger learning rate can be used to allow the model's parameters to update quickly, accelerating convergence. As training progresses and the model gradually approaches the optimal solution, continuing to use a large learning rate may cause the model to fluctuate around the optimal solution, failing to converge stably. Therefore, in the later stages of training, the learning rate needs to be gradually reduced to allow the model to fine-tune its parameters and improve its accuracy.

[0123] For example, use a learning rate of 0.1 for the first 10 epochs of training, adjust the learning rate to 0.01 starting from the 11th epoch, and then adjust it to 0.001 again in the 21st epoch.

[0124] Ordered learning rate adjustment involves changing the learning rate sequentially according to a pre-defined order. This method typically involves gradually reducing the learning rate based on the number of training epochs or changes in the loss function, following a certain pattern. For example, after every 5 epochs of training, the current learning rate might be multiplied by a fixed decay factor, such as 0.9. Assuming an initial learning rate of 0.05, after the 5th epoch, the learning rate would become 0.05 * 0.9 = 0.045; after the 10th epoch, it would become 0.045 * 0.9 = 0.0405, and so on.

[0125] Equal-interval adjustment refers to adjusting the learning rate every fixed number of training epochs during the training process. This adjustment method is relatively simple and direct, and can ensure a relatively stable change in the learning rate during model training. For example, the learning rate is halved every 10 epochs. If the initial learning rate is 0.02, then after the 10th epoch, the learning rate becomes 0.01; after the 20th epoch, the learning rate becomes 0.005, and so on.

[0126] Multi-interval adjustment involves setting different intervals to adjust the learning rate based on different training stages. Unlike equal-interval adjustment, the intervals are not fixed but flexibly set according to the model's training progress. For example, in the first 20 epochs of training, the learning rate is adjusted every 5 epochs; from the 21st to the 50th epoch, it is adjusted every 10 epochs; and from the 51st epoch onwards, it is adjusted every 20 epochs. This allows for more flexible adjustment of the learning rate based on the model's convergence speed at different stages.

[0127] Optionally, cosine annealing can be used to adjust the learning rate. The principle behind cosine annealing is based on the periodic variation of the cosine function, causing the learning rate to gradually decrease from its maximum value to its minimum value within a cycle, and then increase again, repeating this cycle. This adjustment method can help the model escape local optima and find a better global optimum.

[0128] Optionally, based on the concept of multi-interval adjustment and the characteristics of the cosine function, cosine annealing can also be used to adjust the learning rate. In cosine annealing, the learning rate gradually decreases in the form of a cosine curve as training progresses, and then increases again after reaching a minimum point. Specifically, within each adjustment interval, the learning rate starts from an initial value, gradually decreases to a minimum value according to the cosine function, and then returns to the initial value after the interval ends, starting the adjustment for the next interval. For example, setting an adjustment period of 30 epochs, with an initial learning rate of 0.03 and a minimum learning rate of 0.001, within these 30 epochs, the learning rate will start from 0.03, gradually decrease to 0.001 according to the cosine function, and then return to 0.03 after the 30th epoch, starting the adjustment for the next 30 epochs.

[0129] By adjusting these deep learning parameters, the neural network can better adapt to different training data and task requirements during iterative training, improving the model's performance and accuracy, thereby generating a more effective IPv6 network transformation problem identification model.

[0130] In one embodiment, based on the above embodiments, the method for generating an IPv6 network transformation problem identification model based on neural networks further includes:

[0131] Before iteratively training the neural network, the relationship pointers of the collected data are adjusted.

[0132] In this embodiment, if a discrepancy is found between the model analysis results and the human analysis results, the relationship orientation of the collected data can be adjusted before iterative training of the neural network is prepared.

[0133] Optionally, a comprehensive review of the collected and stored data can be conducted to clarify the inherent relationships between different data fields. For example, IPv6 address allocation may be related to the connection status of network devices, and traffic data at different time periods may be related to network application usage. By deeply analyzing these relationships, a detailed data relationship graph can be constructed, providing a foundation for subsequent adjustments.

[0134] From the analyzed data relationships, key relationships closely related to IPv6 network transformation issues were identified. For example, during the IPv6 address transition, the mapping relationship between old and new addresses, the relationship between address allocation rules and network topology, etc., have a significant impact on model training and problem identification.

[0135] Check for potential errors in data import, such as incorrect associations between data fields. Correct these errors to ensure that the relationships between data accurately reflect the actual network situation. For example, if you find that the status data of a network device is associated with an incorrect IPv6 address, correct it promptly.

[0136] For identified key relationships, enhance their relevance when data is entered into the database. This can be achieved by adding related fields or adjusting the data storage structure to highlight the importance of these relationships. For example, creating a dedicated relationship table in the database to store key information related to IPv6 address allocation and network device status makes these relationships clearer and more explicit in the data.

[0137] Introducing New Relationships: Based on human analysis and actual network conditions, new relationships that were previously overlooked or undiscovered are introduced. For example, with the development of IPv6 networks, new network applications or services may emerge, which have specific relationships with network performance and security issues. These new relationships are incorporated into the data storage's relational system, enriching the data's relational orientation.

[0138] Update the data already stored in the database based on the adjusted data relationships. This may involve modifying the database table structure and updating data records. Ensure that the data in the database accurately reflects the adjusted relationships to provide accurate data support for subsequent iterative training.

[0139] By adjusting the relationship pointers of the collected data in the database as described above, the association structure of the data can be optimized, providing more accurate and targeted data for subsequent neural network iterative training, thereby improving the model's performance and its ability to identify IPv6 network transformation issues.

[0140] In one embodiment, based on the above embodiments, the adjustment of the relationship pointing to the collected data storage includes:

[0141] By identifying data originating from a unified node through key node IDs and using the sequential numbering of these IDs, a network structure tree is formed. This structure also links the different data collected by each node, creating interrelationships between them.

[0142] In this embodiment, adjusting the relationship orientation of the collected data before iteratively training the neural network is a crucial step in improving the model training effect. This step uses key node IDs and numbers to construct a network structure tree and associate the data collected from each node.

[0143] First, all critical nodes in the network are accurately identified. These critical nodes include devices such as routers, switches, and servers that perform important functions in the IPv6 network.

[0144] Each key node is assigned a unique ID. This ID serves as the node's identity identifier, possessing uniqueness and stability to ensure accurate identification and differentiation of different nodes throughout the entire data processing process.

[0145] Optionally, IPv6 support data collected from various key nodes can be identified using the corresponding key node ID. For example, if IPv6 traffic data or address allocation data from a router is collected, this data can be bound to the key node ID of that router. In this way, the source node of each piece of data can be clearly identified, providing a foundation for subsequent data management and analysis.

[0146] Optionally, key node IDs can be sequentially numbered based on the network topology and data flow. For example, starting with the core nodes of the network, each node can be assigned a number according to its hierarchy and connection order. This numbering method reflects the location and connection relationships of nodes in the network, allowing the network structure to be represented in the numbering system.

[0147] Optionally, a network structure tree can be constructed based on the IDs of the key nodes after their numbers. The root node of the tree is usually the core node of the network, and the child nodes expand sequentially according to their numbers, forming a hierarchical tree structure.

[0148] A network structure tree visually illustrates the connections and hierarchical structure between nodes in a network, which helps in understanding the flow path and distribution of data within the network.

[0149] Optionally, for each key node, the different types of data it collects can be integrated. For example, a router node may collect IPv6 traffic data, device status data, address allocation data, etc., and these data can be centrally stored in a database record associated with that node ID.

[0150] Optionally, based on the network structure tree, the relationships between data between different nodes can be established. For example, the traffic data of a branch node may be affected by its upstream core node. The network structure tree can clarify this relationship and reflect it in the data.

[0151] It can also link data from different nodes at the same point in time or within the same time period for comprehensive analysis and comparison.

[0152] Optionally, the established data relationships can be stored in a database to form an interconnected data system. This can be achieved by creating related tables, setting foreign keys, and other methods to store related data in the database.

[0153] In this way, these correlated data can be easily acquired and utilized during iterative training of neural networks, improving the accuracy and effectiveness of training. By adjusting the correlation index after the collected data is entered into the database, the neural network can be assisted in understanding the relationships between the data, thereby improving the output accuracy of the neural network analysis results.

[0154] Optionally, the adjustment method mainly adopts the following approach: identifying data from a unified node through key node IDs, forming a network structure tree through the sequential numbering of IDs, and simultaneously associating the total traffic, application traffic, and network quality collected by each node, so that different data such as traffic, applications, network quality, and end-to-end data form interrelationships.

[0155] In addition, refer to Figure 2 This application also provides a control device Z10, comprising:

[0156] The monitoring module Z11 is used to deploy IPv6 monitoring points, collect IPv6 support data from key nodes in the network, and clean and organize the data before storing it in the database.

[0157] Analysis module Z12 is used to analyze the collected data for IPv6 transformation problems based on a neural network that has been pre-trained for IPv6 network transformation problem identification, and generate model analysis results.

[0158] The output module Z13 is used to output the collected data for manual analysis of IPv6 transformation issues and generate human analysis results;

[0159] The detection module Z14 is used to compare the model analysis results with the human analysis results to detect whether there is any result deviation.

[0160] The iteration module Z15 is used to iteratively train the neural network based on the result deviation if the result is true, and return to the step of performing the IPv6 transformation problem analysis on the collected data based on the neural network that has been pre-trained for IPv6 network transformation problem identification, and generating model analysis results.

[0161] The generation module Z16 is used to solidify the adjustment parameters and training results of the neural network if no, and generate an IPv6 network transformation problem identification model.

[0162] Optionally, the control device Z10 can be a virtual control device (such as a virtual machine) or a physical device (such as a physical device other than a computer device that can execute the corresponding method).

[0163] Furthermore, this application also provides a computer device whose internal architecture can be as follows: Figure 3 As shown, the system includes a processor, memory, communication interface, and input interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores data called by the computer programs. The communication interface is used for data communication with external terminals. The input interface is used to receive signals input from external devices. When the computer program is executed by the processor, it implements a neural network-based IPv6 network transformation problem identification model generation method as described in the above embodiment.

[0164] Those skilled in the art will understand that Figure 3 The structures shown are merely block diagrams of a portion of the structures related to the present application and do not constitute a limitation on the computer devices to which the present application is applied. For example, in some alternative embodiments, the computer device may further include an output interface (not shown in the figures), and the output interface is also connected to the system bus and used to output corresponding signals to peripherals.

[0165] Furthermore, this application also proposes a computer-readable storage medium comprising a computer program that, when executed by a processor, implements the steps of the neural network-based IPv6 network transformation problem identification model generation method described in the above embodiments. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0166] In summary, the method, control device, computer equipment, and computer-readable storage medium for generating IPv6 network transformation problem identification models based on neural networks provided in this application embodiment utilize a pre-trained neural network to analyze collected data, enabling rapid and efficient generation of model analysis results. Simultaneously, it outputs data for manual analysis, combining the efficiency of machine analysis with the flexibility of manual analysis, resulting in more comprehensive problem identification. Furthermore, comparing the model analysis results with manual analysis results allows for timely detection of discrepancies. If a discrepancy is found, the neural network is iteratively trained to continuously optimize model performance and improve the accuracy of problem identification. When the results are without discrepancies, the adjusted parameters of the neural network and the training results are solidified, generating a stable and reliable IPv6 network transformation problem identification model. This model can be effectively applied to the identification of actual IPv6 network transformation problems, thereby improving the accuracy and efficiency of IPv6 network transformation problem identification.

[0167] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media provided in this application and in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0168] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0169] The above description is only a preferred embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural changes made based on the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A neural network-based IPv6 network renovation problem identification model generation method, characterized in that, The method comprises the following steps: deploying IPv6 monitoring points, collecting IPv6 support data of key nodes in the network and in the network, and completing data cleaning and regularizing into a database; based on a neural network pre-trained for IPv6 network reconstruction problem identification, analyzing the collected data for IPv6 reconstruction problems, and generating model analysis results; and outputting the collected data for manual analysis of IPv6 reconstruction problems to generate artificial analysis results; wherein the neural network is constructed based on a variant of RNN with a gating mechanism, and the training process adopts a supervised learning method; in the network training process of the neural network, quantum gate operations are used to update the parameters of the network, wherein the weights and bias parameters of the network are encoded as quantum states, and the gradient descent update process is simulated through a quantum gate sequence; in the recurrent structure of the neural network, there is a recurrent connection between neurons, and based on the controlled non-gate, entanglement is generated between the quantum states corresponding to the neurons, so that the state change of one neuron can affect other neurons entangled with it; comparing the model analysis results with the artificial analysis results to detect whether there is a result deviation; if yes, iteratively training the neural network based on the result deviation, and returning to the step of analyzing the collected data for IPv6 reconstruction problems based on the neural network pre-trained for IPv6 network reconstruction problem identification to generate model analysis results; if no, solidifying the adjustment parameters and training results of the neural network to generate an IPv6 network reconstruction problem identification model; wherein the IPv6 network reconstruction problem identification model learns the relationship between various problems in the IPv6 network reconstruction process and the relevant data features in the IPv6 support data; the IPv6 network reconstruction problem identification model is used to identify problems in the IPv6 network reconstruction according to the collected relevant network data. 2.The neural network-based IPv6 network transformation problem identification model generation method of claim 1, wherein, The method for generating an IPv6 network reconstruction problem identification model based on a neural network further comprises: before iteratively training the neural network, adjusting the deep learning parameters of the neural network, wherein the adjustment parameters include at least one of learning rate adjustment, ordered learning rate adjustment, equal interval adjustment, multiple interval adjustment, and cosine annealing parameters in learning. 3.The neural network-based IPv6 network transformation problem identification model generation method of claim 1, wherein, The method for generating an IPv6 network reconstruction problem identification model based on a neural network further comprises: before iteratively training the neural network, adjusting the relationship direction of the collected data into the database. 4.The method of claim 3, wherein, The relationship direction adjustment of the collected data into the database comprises: identifying and binding the data of the nodes through the key node ID, and forming a network structure tree through the sequential numbering of the key node ID, while associating different data collected by each node to form a mutual association relationship.

5. A control device characterized by comprising: The method comprises the following steps: a monitoring module for deploying IPv6 monitoring points, collecting IPv6 support data of key nodes in the network and in the network, and completing data cleaning and regularizing into a database; The analysis module is configured to analyze the collected data based on the neural network pre-trained for IPv6 network reconstruction problem identification, and generate a model analysis result; the neural network is constructed based on a RNN variant of a gating mechanism, and a supervised learning method is used in the training process; in the network training process of the neural network, quantum gate operations are used to update the parameters of the network, wherein the weights and bias parameters of the network are encoded as quantum states, and the gradient descent update process is simulated through a quantum gate sequence; in the recurrent structure of the neural network, there is a recurrent connection between neurons, and entanglement is generated between the quantum states corresponding to the neurons based on a controlled non-gate, so that the state change of one neuron can affect other neurons entangled with it; The output module is configured to output the collected data for manual analysis of IPv6 reconstruction problems and generate a manual analysis result. The detection module is configured to compare the model analysis result with the manual analysis result, and detect whether there is a result deviation. The iteration module is configured to, if yes, iteratively train the neural network based on the result deviation, and return to the step of analyzing the collected data based on the neural network pre-trained for IPv6 network reconstruction problem identification, and generating a model analysis result. The generation module is configured to, if no, solidify the adjustment parameters and training results of the neural network, and generate an IPv6 network reconstruction problem identification model; the IPv6 network reconstruction problem identification model learns the relationship between various problems in the IPv6 network reconstruction process and the relevant data features in the IPv6 support data; and the IPv6 network reconstruction problem identification model is used to identify problems in the IPv6 network reconstruction based on the collected relevant network data.

6. A computer device, comprising: The computer device includes a memory, a processor, and a computer program stored on the memory and executable on the processor, and the computer program implements the steps of the neural network-based IPv6 network reconstruction problem identification model generation method according to any one of claims 1 to 4 when executed by the processor.

7. A computer readable storage medium characterized in that, The computer readable storage medium stores a computer program, and the computer program implements the steps of the neural network-based IPv6 network reconstruction problem identification model generation method according to any one of claims 1 to 4 when executed by the processor.

Images