Method, apparatus and electronic device for determining quality model of transmission link

Abstract

This invention discloses a method, apparatus, and electronic device for determining the quality model of a transmission link. The method includes: acquiring training index data corresponding to the transmission link of a power optical fiber, wherein the training index data is used to characterize the quality status of the transmission link; obtaining a classification result corresponding to the training index data based on a clustering algorithm; determining a quality level label corresponding to the training index data based on the classification result; and training a predetermined initial quality model using a deep residual network algorithm based on the training index data and the corresponding quality level label to obtain a target quality model for the transmission link. This invention solves the technical problems of low model accuracy, poor robustness, and low convergence caused by data noise interference in the quality model of the transmission link.

Description

Technical Field

[0001] This invention relates to the field of electronic information technology, and more specifically, to a method, apparatus, and electronic device for determining the quality model of a transmission link. Background Technology

[0002] Currently, some power fiber optic transmission networks contain aging equipment and transmission links with long service lives. This places higher demands on operation and maintenance (O&M) support, requiring proactive real-time evaluation and monitoring of transmission link quality to ensure the safe and reliable operation of the power fiber optic transmission network. Link quality assessment commonly used in related technologies primarily employs subjective and objective evaluation methods to obtain quality models. Subjective evaluation methods rely heavily on expert experience, are highly subjective, and are prone to issues such as information duplication and loss. Objective evaluation methods, on the other hand, are susceptible to data noise interference, resulting in limited application and low convergence of the quality model.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This invention provides a method, apparatus, and electronic device for determining the quality model of a transmission link, to at least solve the technical problems of low model accuracy, poor robustness, and low convergence caused by data noise interference in the quality model of the transmission link.

[0005] According to one aspect of the present invention, a method for determining the quality model of a transmission link is provided, comprising: acquiring training index data corresponding to a power optical fiber transmission link, wherein the training index data is used to characterize the quality status of the transmission link; obtaining a classification result corresponding to the training index data based on a clustering algorithm; determining a quality level label corresponding to the training index data based on the classification result; and training a predetermined initial quality model using a deep residual network algorithm based on the training index data and the quality level label corresponding to the training index data to obtain a target quality model corresponding to the transmission link.

[0006] Optionally, obtaining the training index data corresponding to the transmission link of the power optical fiber includes: obtaining the initial data of the transmission link of the power optical fiber; and performing data cleaning processing on the initial data based on the variance value corresponding to the initial data to obtain the training index data corresponding to the transmission link.

[0007] Optionally, when the training indicator data consists of multiple sets, and these sets correspond to multiple training indicators, obtaining the classification results based on the clustering algorithm includes: classifying the multiple sets of training indicator data to determine positive and negative indicator data within the sets, wherein the positive indicator data are those whose values ​​are directly proportional to the quality status, and the negative indicator data are those whose values ​​are inversely proportional to the quality status. The positive indicator data is subjected to a first normalization process to obtain the processed positive indicator data; the negative indicator data is subjected to a second normalization process to obtain the processed negative indicator data; based on the processed positive indicator data and the processed negative indicator data, the processed multiple sets of training indicator data are obtained; based on the clustering algorithm, using the processed multiple sets of training indicator data and a preset number of clusters, the classification results corresponding to the processed multiple sets of training indicator data are obtained, wherein the clustering algorithm is a Gaussian mixture clustering algorithm.

[0008] Optionally, when there are multiple transmission links, the step of obtaining the classification results corresponding to the multiple sets of training indicator data based on the clustering algorithm and using the processed multiple sets of training indicator data and a preset number of clusters includes: based on the clustering algorithm, using the processed multiple sets of training indicator data and a preset number of clusters, processing them in the following way to obtain the cluster probabilities corresponding to the processed multiple sets of training indicator data, wherein the cluster probabilities are probabilities based on preset cluster heads:

[0009]

[0010] Where p(x) is the cluster probability of the x-th group of training index data in the processed multi-group training index data, μ i Let ∑ be the mean vector of the i-th transmission link among multiple transmission links. i Let a be the covariance matrix of the i-th transmission link among the plurality of transmission links. i Let p(x|μ) be the mixing coefficient of the i-th transmission link among the plurality of transmission links. i ,∑ iLet be the multivariate Gaussian distribution probability density function of the x-th group of training index data in the i-th transmission link of the plurality of transmission links, where i is the identifier of the plurality of transmission links, k is the number of clusters, and x is the identifier of the plurality of training index data after processing; according to the cluster probability and the number of clusters, the plurality of training index data after processing are divided to obtain the classification results corresponding to the plurality of training index data after processing.

[0011] Optionally, when the training indicator data consists of multiple sets, and the multiple sets of training indicator data correspond to multiple training indicators, determining the quality level label corresponding to the training indicator data based on the classification result includes: processing the multiple sets of training indicator data using the coefficient of variation method to obtain the weight values ​​corresponding to the multiple training indicators respectively; and determining the quality level label corresponding to the multiple sets of training indicator data based on the weight values ​​and the classification result.

[0012] Optionally, the step of training a predetermined initial quality model using a deep residual network algorithm based on the training metric data and the quality level labels corresponding to the training metric data to obtain a target quality model corresponding to the transmission link includes: training the predetermined initial quality model using the deep residual network algorithm based on the training metric data and the quality level labels corresponding to the training metric data to obtain the trained predetermined initial quality model; and optimizing the trained predetermined initial quality model based on the momentum gradient descent optimization method to obtain the target quality model, wherein the momentum gradient descent method is a gradient descent method using an exponentially weighted moving average method.

[0013] Optionally, the method further includes: acquiring test index data of the transmission link under test; inputting the test index data into the target quality model for testing, and obtaining the quality test result corresponding to the test index data of the transmission link under test.

[0014] According to another aspect of the present invention, a transmission link quality model determination apparatus is also provided, comprising: a first acquisition module, configured to acquire training index data corresponding to a power optical fiber transmission link, wherein the training index data is used to characterize the quality status of the transmission link; a classification module, configured to obtain a classification result corresponding to the training index data based on a clustering algorithm; a determination module, configured to determine a quality level label corresponding to the training index data based on the classification result; and a second acquisition module, configured to train a predetermined initial quality model using a deep residual network algorithm based on the training index data and the quality level label corresponding to the training index data, to obtain a target quality model corresponding to the transmission link.

[0015] According to another aspect of the present invention, a non-volatile storage medium is also provided, the non-volatile storage medium storing a plurality of instructions adapted for loading by a processor and executing any one of the transmission link quality model determination methods.

[0016] According to another aspect of the present invention, an electronic device is also provided, comprising: one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the transmission link quality model determination method as described in any one of the present invention.

[0017] In this embodiment of the invention, training index data corresponding to the transmission link of a power optical fiber is acquired, wherein the training index data is used to characterize the quality status of the transmission link; a classification result corresponding to the training index data is obtained based on a clustering algorithm; a quality level label corresponding to the training index data is determined based on the classification result; and a target quality model corresponding to the transmission link is obtained by training a predetermined initial quality model using a deep residual network algorithm based on the training index data and the quality level label corresponding to the training index data. This achieves the goal of classifying the index data of the transmission link, reducing data noise, and improving the processing efficiency of the quality model, thereby achieving the technical effect of high processing efficiency and high convergence of the quality model for transmission link quality. This solves the technical problem of low model accuracy, poor robustness, and low convergence caused by data noise interference in the quality model of the transmission link. Attached Figure Description

[0018] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0019] Figure 1 This is a flowchart of an optional method for determining the quality model of a transmission link according to an embodiment of the present invention;

[0020] Figure 2 This is a clustering diagram illustrating an optional method for determining the quality model of a transmission link according to an embodiment of the present invention.

[0021] Figure 3 This is a residual block diagram of an optional transmission link quality model determination method according to an embodiment of the present invention;

[0022] Figure 4 This is a schematic diagram of an optional transmission link quality model determination device according to an embodiment of the present invention. Detailed Implementation

[0023] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0024] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0025] With the further development and application of power fiber optic transmission networks, the number of services carried by the network is increasing, the network coverage is expanding, and the complexity and management difficulty of the network are also constantly increasing. At the same time, some aging equipment and links in power fiber optic transmission networks still exist, thus placing higher demands on operation and maintenance support. It is necessary to proactively evaluate and monitor link quality in real time to support the safe and reliable operation of the power fiber optic transmission network. Quality assessment of power fiber optic transmission links requires a systematic and comprehensive analysis of link information, environmental information, and network composition.

[0026] In related technologies, quality assessment models for transmission links are mainly implemented through subjective and objective evaluation methods. Subjective evaluation methods require constructing a corresponding evaluation index system, determining the weight of each index based on expert scores, and then using methods such as the analytic hierarchy process (AHP) to comprehensively evaluate link quality. The drawback of this approach is its over-reliance on expert experience, high subjectivity, and susceptibility to issues like information duplication and loss. Objective evaluation methods, on the other hand, typically involve calculating index weights and conducting a comprehensive quality assessment. Objective methods are more objective than subjective methods, avoiding biases caused by human factors. However, simply assigning weights cannot mitigate interference from data noise, which consists of duplicate or corrupted data that is difficult for automated programs to identify, leading to inaccurate results, a narrower application scope, and susceptibility to noise in the data, thus exhibiting significant limitations.

[0027] According to an embodiment of the present invention, a method embodiment for determining the quality model of a transmission link is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0028] Figure 1 This is a method for determining the quality model of a transmission link according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0029] Step S102: Obtain training index data corresponding to the transmission link of the power optical fiber, wherein the training index data is used to characterize the quality status of the transmission link.

[0030] It is understandable that in order to obtain a quality model of the transmission link, it is necessary to first obtain the training index data of the power fiber transmission link as the training set of the model.

[0031] Optionally, the training metrics data mentioned above can be of various types, such as: link age information, transmission power information, packet acceptance rate, temperature and humidity environmental information, optical signal-to-noise ratio, and other metrics data related to power optical fiber.

[0032] In one optional embodiment, the above-mentioned acquisition of training index data corresponding to the transmission link of the power optical fiber includes: acquiring initial data of the transmission link of the power optical fiber; and performing data cleaning processing on the initial data based on the variance value corresponding to the initial data to obtain the training index data corresponding to the transmission link.

[0033] It is understandable that in order to reduce the impact of data noise on the target quality model and improve the model training efficiency, the initial data of the power fiber optic transmission link is cleaned. The above data cleaning process can reduce redundant data and remove data that is out of the given range or has abnormal format, which is conducive to obtaining training index data with better training effect.

[0034] Alternatively, there can be various methods for cleaning the data, such as cleaning the data with a variance of 0 or close to 0 in the initial data, which helps to remove noisy data.

[0035] Step S104: Based on the clustering algorithm, obtain the classification results corresponding to the above training index data.

[0036] It is understandable that using clustering algorithms to process training index data and obtain classification results is beneficial for objective data classification and helps improve the accuracy of the target quality model.

[0037] In an optional embodiment, when the training indicator data consists of multiple sets, and each set corresponds to multiple training indicators, obtaining the classification results corresponding to the training indicator data based on a clustering algorithm includes: classifying the multiple sets of training indicator data to determine positive and negative indicator data within the multiple sets of training indicator data, wherein the positive indicator data are training indicator data whose numerical value is directly proportional to the quality status, and the negative indicator data are training indicator data whose numerical value is inversely proportional to the quality status. Training indicator data; performing a first normalization process on the positive indicator data to obtain the processed positive indicator data; performing a second normalization process on the negative indicator data to obtain the processed negative indicator data; based on the processed positive and negative indicator data, obtaining the processed multiple sets of training indicator data; based on the clustering algorithm, using the processed multiple sets of training indicator data and a preset number of clusters, obtaining the classification results corresponding to the processed multiple sets of training indicator data, wherein the clustering algorithm is a Gaussian mixture clustering algorithm.

[0038] It is understandable that the training metric data is divided into multiple groups based on various training metrics, with each group potentially containing one or more data points. In practical applications, the transmission path of electric beams involves multiple metric data types, including positive and negative metric data. Simply put, positive metric data represents data where higher values ​​are better, while negative metric data represents data where lower values ​​are better. The positive and negative metric data are then processed to be dimensionless. This involves a first normalization process to obtain the processed positive metric data, and a second normalization process to obtain the processed negative metric data, resulting in multiple groups of processed training metric data. These processed groups of training metric data are then considered dimensionless. A clustering algorithm is then used to process these multiple groups of training metric data, clustering them according to a preset number of clusters to obtain the classification results.

[0039] Alternatively, there can be multiple clustering algorithms, such as the K-means algorithm, where K represents the number of clusters.

[0040] In an optional embodiment, when there are multiple transmission links, the above-mentioned clustering algorithm, using the processed multiple sets of training indicator data and a preset number of clusters, to obtain the classification results corresponding to the processed multiple sets of training indicator data respectively, includes: based on the above-mentioned clustering algorithm, using the processed multiple sets of training indicator data and a preset number of clusters, processing them in the following way to obtain the cluster probabilities corresponding to the processed multiple sets of training indicator data respectively, wherein the cluster probabilities are probabilities based on preset cluster heads:

[0041]

[0042] Where p(x) is the cluster probability of the x-th group of training index data in the above-processed multi-group training index data, μ i Let ∑ be the mean vector of the i-th transmission link among multiple transmission links. i Let p(x|μ) be the covariance matrix of the i-th transmission link among the aforementioned multiple transmission links, ai be the mixing coefficient of the i-th transmission link among the aforementioned multiple transmission links, and p(x|μ) be the covariance matrix of the i-th transmission link among the aforementioned multiple transmission links. i ,∑ iLet be the multivariate Gaussian distribution probability density function of the x-th group of training index data in the i-th transmission link among the processed training index data, where i is the identifier of the multiple transmission links, k is the number of clusters, and x is the identifier of the processed training index data. Based on the cluster probability and the number of clusters, the processed training index data is divided to obtain the classification results corresponding to the processed training index data.

[0043] It is understandable that a mathematical expression has been established for multiple transmission links of power optical fiber. Based on the clustering algorithm, the cluster probabilities corresponding to multiple sets of training data are first obtained. Based on the cluster probabilities and the number of clusters, the classification results of the above-mentioned multiple sets of training indicators are obtained.

[0044] It should be noted that the number of clusters represents the number of classifications into which the transmission links are divided. That is, the transmission links are divided into multiple classification results according to the number of clusters. For example, if the number of clusters is set to 3, the Gaussian mixture clustering algorithm is used to cluster the training index data of the transmission links, and the training index data of each transmission link is classified into the corresponding cluster.

[0045] Step S106: Based on the above classification results, determine the quality level label corresponding to the above training indicator data.

[0046] It is understandable that associating training indicator data with quality level labels and introducing the classification ability of clustering algorithms into the establishment of quality models is beneficial to the evaluation effect of the target quality model on transmission links.

[0047] In an optional embodiment, when the training indicator data consists of multiple sets, and the multiple sets of training indicator data correspond to multiple training indicators, the step of determining the quality level label corresponding to the training indicator data based on the classification result includes: processing the multiple sets of training indicator data using the coefficient of variation method to obtain the weight values ​​corresponding to the multiple training indicators respectively; and determining the quality level label corresponding to the multiple sets of training indicator data based on the weight values ​​and the classification result.

[0048] It is understandable that the coefficient of variation method, as an objective weight determination method, combined with the classification algorithm obtained from the above clustering algorithm, helps improve the accuracy of the association between quality grade labels and multiple sets of training indicator data, and reduces the impact of data noise.

[0049] Optionally, the specific implementation of the above-mentioned coefficient of variation method is, for example, calculating the mean of each group of training index data and establishing the data representation as: Among them, A jLet x represent the mean of the j-th training metric data, n represent the number of training metric data samples in the transmission link, and x represent the mean of the j-th training metric data. ij Let represent the value of the j-th training metric data in the multiple sets of training metric data for the i-th transmission link. Then, calculate the mean squared error corresponding to each set of training metric data, and establish its mathematical expression as: Among them, SD j Let represent the mean squared error of the j-th group of training metrics data. Then, the coefficient of variation for each group of training metrics data is obtained, and its mathematical expression is established as: Among them, V j Let represent the coefficient of variation of the j-th group of training metrics data. When calculating the weight values ​​of each group of training metrics data, the mathematical expression is as follows: Where, ω j denoted as the weight value of the j-th group of training index data.

[0050] Alternatively, there are multiple ways to obtain the quality level label. For example, based on the above weight values, the quality score corresponding to each transmission link can be mathematically expressed as follows: Where, q i Let x represent the quality score of the i-th transmission link. ij ω represents the value of the j-th training metric data in the multiple sets of training metric data for the i-th transmission link. j Let be the weight value of the j-th group of training indicator data. Using the quality score obtained above, calculate the average quality score of the training indicator data for each classification result. Sort the training indicator data according to the average quality score of the training indicator data for each classification result, thus dividing the training indicator data into three levels from high to low: "Excellent", "Medium", and "Poor".

[0051] Step S108: Based on the above training index data and the above quality level labels corresponding to the above training index data, the deep residual network algorithm is used to train the predetermined initial quality model to obtain the target quality model corresponding to the above transmission link.

[0052] It is understandable that a deep residual network algorithm is used to train a predetermined initial quality model to obtain a target quality model, which has the ability to classify the test index data in the power fiber optic transmission link.

[0053] In one optional embodiment, the above-mentioned method of training a predetermined initial quality model using a deep residual network algorithm based on the training index data and the quality level labels corresponding to the training index data to obtain a target quality model corresponding to the transmission link includes: training the predetermined initial quality model using the deep residual network algorithm based on the training index data and the quality level labels corresponding to the training index data to obtain the trained predetermined initial quality model; and optimizing the trained predetermined initial quality model based on the momentum gradient descent optimization method to obtain the target quality model, wherein the momentum gradient descent method is a gradient descent method using an exponentially weighted moving average method.

[0054] It is understandable that the training index data and its accompanying quality level labels are used as the training set for a predetermined initial quality model. The model is then trained on this training set, and optimized using the momentum gradient descent optimization method. The momentum gradient descent optimization method, also known as the Momentum optimizer, can solve the oscillation problem during algorithm convergence, thus enhancing the convergence of the target quality model.

[0055] Optionally, the aforementioned predetermined initial quality model can be set to various values. For example, for deep residual network algorithms, it may contain multiple residual blocks. Each residual block uses a fully pre-activated structure with three convolutional layers. The first layer uses a 1×1 convolutional kernel for feature dimensionality reduction; the second layer uses a 3×3 convolutional kernel to operate on the lower-dimensional feature vector; and the third layer uses a 1×1 convolutional kernel for dimensionality reconstruction. A normalization layer and activation function are applied before each convolutional layer, but no activation function is applied outside the residual part. The formula for the residual block is expressed as follows: x 输入 =F+x 输入 , where x 输入 This represents the input to the residual block, x. 输入 represents the output of the residual block, and F represents the residual mapping. For the test index data, a 7×7 convolution and a 3×3 max pooling operation are first performed, followed by four residual blocks. Finally, a global average pooling layer is connected to reduce the dimensionality of the extracted high-dimensional features, and a fully connected layer combined with a normalized exponential function is used to calculate the probability of the test index data belonging to each quality level, thus outputting the quality test result of the test index data.

[0056] Optionally, the specific optimization method of the above momentum gradient descent optimization method can be of various types. For example, the momentum gradient descent optimization method based on exponential weighted moving average can be selected to fine-tune the parameters of the deep residual network algorithm. The weight update formula of the momentum gradient descent optimization method is as follows: ν=βν+(1-β)dw, w=w-αν, where ν is the gradient calculated using exponential weighted average, dw is the original gradient, β is a hyperparameter of the momentum gradient descent optimization method, which can be set to 0.9, α is the learning rate, which can be set to 0.01, and w is the weight decay rate, which can be set to 0.0001.

[0057] In an optional embodiment, the method further includes: acquiring test index data of the transmission link under test; inputting the test index data into the target quality model for testing, and obtaining the quality test result corresponding to the test index data of the transmission link under test.

[0058] It is understandable that, given a well-trained target quality model, inputting the test index data of the transmission link under test will yield the corresponding quality test results.

[0059] Based on the above embodiments and optional embodiments, the present invention proposes an optional implementation method, and specific examples are given for ease of understanding:

[0060] Various performance data of the power fiber optic transmission link were collected as training metrics, including link age, transmission power, packet acceptance rate, temperature and humidity environmental information, and optical signal-to-noise ratio, totaling 20 sets of data. The variance of each set of training metrics was calculated; the specific values ​​below are for illustrative purposes only and do not constitute a limitation. Table 1 shows the training metrics data before data cleaning.

[0061] Table 1

[0062]

[0063]

[0064] Data cleaning is achieved by filtering out training metric data with variance of 0 or close to 0. The number of training metric data sets was reduced from 20 in Table 1 to 12. The multiple sets of training metric data after dimensionality reduction are shown below. Table 2 is a schematic table of the training metric data after data cleaning.

[0065] Table 2

[0066]

[0067]

[0068] The Gaussian mixture clustering algorithm sets the number of clusters to 3 in order to divide the quality level of the transmission link into 3 levels. Figure 2 This is a clustering diagram illustrating an optional transmission link quality model determination method according to an embodiment of the present invention, such as... Figure 2 As shown, the training index data of the transmission links are input into the Gaussian mixture clustering algorithm, which divides the training index data of each transmission link into the corresponding clusters to obtain the classification results.

[0069] The weight value of each group of training indicator data was calculated using the coefficient of variation method. Based on the weight value and the classification results obtained by the above clustering algorithm, the quality level labels corresponding to the multiple groups of training indicator data were determined. Table 3 is a schematic table of the weight values ​​of multiple groups of training indicator data.

[0070] Table 3

[0071]

[0072]

[0073] Based on the calculated training metrics and the aforementioned weight values, the quality score for each transmission link is mathematically expressed as follows: Where, q i Let x represent the quality score of the i-th transmission link. ij ω represents the value of the j-th training metric data in the multiple sets of training metric data for the i-th transmission link. j Let be the weight value of the training indicator data for the j-th group. Using the quality score obtained above, calculate the average quality score of the training indicator data for each classification result. Sort the training indicator data according to the average quality score of the training indicator data for each classification result, thus dividing the training indicator data into three levels from high to low: "Excellent," "Medium," and "Poor." Table 4 is a schematic table of quality level labels.

[0074] Table 4

[0075] Classification results Average quality score Quality grade label 1 88.16 excellent 2 69.72 medium 3 52.54 Difference

[0076] Training metric data with quality level labels is fed into a deep residual network algorithm for training to obtain the target quality model of the power optical fiber transmission link. The test metric data is then input into the target quality model to obtain the corresponding quality test results. Table 5 shows a schematic table of the quality test results for the test metric data.

[0077] Table 5

[0078]

[0079]

[0080] This invention provides a specific method for setting up an initial quality model, which is obtained using a deep residual network algorithm and based on user-defined parameters. The deep residual network algorithm comprises multiple residual blocks, each using a fully pre-activated structure with three convolutional layers. The first layer uses a 1×1 convolutional kernel for feature dimensionality reduction; the second layer uses a 3×3 convolutional kernel to operate on the lower-dimensional feature vectors; and the third layer uses a 1×1 convolutional kernel for dimensionality reconstruction. Figure 3 This is a residual block diagram of an optional transmission link quality model determination method according to an embodiment of the present invention, as shown below. Figure 3 As shown, a normalization layer and activation function are applied before each convolutional layer, but no activation function is applied outside the residual block. The formula for the residual block is as follows: x 输入 =F+x 输入 , where x 输入 This represents the input to the residual block, x. 输入 The output of the residual block is represented by F, and the residual mapping is represented by F. For the test index data, a 7×7 convolution and a 3×3 max pooling operation are first performed, followed by four residual blocks. Finally, a global average pooling layer is connected to reduce the dimensionality of the extracted high-dimensional features, and a fully connected layer combined with a normalized exponential function is used to calculate the probability of the test index data belonging to each quality level, thus outputting the quality test result of the test index data. The parameters of the deep residual network algorithm for the initial quality model are set, and Table 6 shows the configuration of the initial quality model.

[0081] Table 6

[0082]

[0083]

[0084] The hyperparameters of the deep residual network algorithm are set, and the training index data of power optical fibers with quality grade labels are fed in. The initial quality model is trained iteratively multiple times to obtain the target quality model of the power optical fiber transmission link. The hyperparameters are set as follows: number of clusters is set to 3, training index data size is set to 256, learning rate is set to 0.01, and weight decay rate is set to 0.0001. The training index data of power optical fiber transmission links with quality grade labels are input into the initial quality model for training to obtain the target quality model. The momentum gradient descent optimization method based on exponential weighted moving average is selected to optimize the deep residual network algorithm. The weight update formula of the momentum gradient descent optimization method is as follows: ν=βν+(1-β)dw, w=w-αν, where ν is the gradient calculated using exponential weighted average, dw is the original gradient, β is a hyperparameter of the momentum gradient descent optimization method, which can be set to 0.9, α is the learning rate, which can be set to 0.01, and w is the weight decay rate, which can be set to 0.0001.

[0085] For a well-trained target quality model, inputting the test index data of the transmission link under test can yield the corresponding quality test results.

[0086] The above-mentioned optional implementation methods achieve at least one of the following effects: feature engineering (i.e., data cleaning) is performed on various indicator data of the transmission link to obtain a feature set that is correlated with the quality status of the transmission link, reducing data noise interference. Based on this, clustering and quality level labels are determined. Compared with related methods, this invention comprehensively considers the impact of various attributes on transmission link quality. While using objective methods for weight calculation and quality scoring, it retains a certain degree of subjectivity in the quality assessment process by classifying quality levels, and reduces data noise interference. This invention transforms the transmission link quality assessment problem into a multi-classification problem of determining the transmission link quality level, and uses a deep residual network algorithm to construct the target quality model of the transmission link. Compared with neural network algorithms in related technologies, the deep residual network algorithm has lower complexity and shorter convergence time. Furthermore, the deep residual network algorithm has a deeper network depth, solving the problem of deep network degradation and avoiding gradient vanishing, thus improving classification accuracy. The target quality model trained by the above method is less affected by data noise, has stronger robustness, faster convergence speed, and higher accuracy, and is more suitable for establishing target quality models for power fiber optic transmission links.

[0087] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0088] This embodiment also provides a transmission link quality model determination apparatus, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the terms "module" and "apparatus" can refer to a combination of software and / or hardware that performs a predetermined function. Although the apparatus described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0089] According to embodiments of the present invention, an apparatus embodiment for implementing a method for determining the quality model of a transmission link is also provided. Figure 4 This is a schematic diagram of a transmission link quality model determination device according to an embodiment of the present invention, such as... Figure 4 As shown, the above-mentioned transmission link quality model determination device includes: a first acquisition module 402, a classification module 404, a determination module 406, and a second acquisition module 408. The device will be described below.

[0090] The first acquisition module 402 is used to acquire training index data corresponding to the transmission link of the power optical fiber, wherein the training index data is used to characterize the quality status of the transmission link.

[0091] The classification module 404 is connected to the first acquisition module 402 and is used to obtain the classification results corresponding to the above training index data based on the clustering algorithm.

[0092] The determination module 406, connected to the classification module 404, is used to determine the quality level label corresponding to the above training indicator data based on the above classification results.

[0093] The second acquisition module 408, connected to the determination module 406, is used to train a predetermined initial quality model using a deep residual network algorithm based on the above-mentioned training index data and the above-mentioned quality level labels corresponding to the above-mentioned training index data, so as to obtain the target quality model corresponding to the above-mentioned transmission link.

[0094] In a transmission link quality model determination device provided by this invention, a first acquisition module 402 is used to acquire training index data corresponding to the transmission link of a power optical fiber, wherein the training index data is used to characterize the quality status of the transmission link; a classification module 404, connected to the first acquisition module 402, is used to obtain the classification result corresponding to the training index data based on a clustering algorithm; a determination module 406, connected to the classification module 404, is used to determine the quality level label corresponding to the training index data based on the classification result; a second acquisition module 408, connected to the determination module 406, is used to train a predetermined initial quality model using a deep residual network algorithm based on the training index data and the quality level label corresponding to the training index data, to obtain the target quality model corresponding to the transmission link. This achieves the goal of classifying the transmission link index data, reducing data noise, and improving the processing efficiency of the quality model, thereby achieving the technical effect of high processing efficiency and high convergence of the quality model for transmission link quality, and solving the technical problems of low model accuracy, poor robustness, and low convergence caused by data noise interference in the transmission link quality model.

[0095] It should be noted that the above modules can be implemented by software or hardware. For example, for the latter, it can be implemented in the following ways: the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0096] It should be noted that the first acquisition module 402, classification module 404, determination module 406, and second acquisition module 408 mentioned above correspond to steps S102 to S108 in the embodiments. The instances and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the content disclosed in the above embodiments. It should be noted that the above modules, as part of the device, can run in a computer terminal.

[0097] It should be noted that the optional or preferred implementation methods of this embodiment can be found in the relevant descriptions in the embodiments, and will not be repeated here.

[0098] The aforementioned transmission link quality model determination device may further include a processor and a memory. The first acquisition module 402, the classification module 404, the determination module 406, the second acquisition module 408, etc., are all stored in the memory as program units, and the processor executes the aforementioned program units stored in the memory to realize the corresponding functions.

[0099] The processor contains a core that retrieves the corresponding program unit from memory. One or more cores may be configured. Memory may include non-persistent memory in computer-readable media, random access memory (RAM), and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory includes at least one memory chip.

[0100] This invention provides a non-volatile storage medium storing a program that, when executed by a processor, implements a method for determining the quality model of a transmission link.

[0101] As shown, this embodiment of the invention provides an electronic device, which includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it performs the following steps: acquiring training index data corresponding to a power optical fiber transmission link, wherein the training index data is used to characterize the quality status of the transmission link; obtaining a classification result corresponding to the training index data based on a clustering algorithm; determining a quality level label corresponding to the training index data based on the classification result; and training a predetermined initial quality model using a deep residual network algorithm based on the training index data and the corresponding quality level label to obtain a target quality model corresponding to the transmission link. The device in this document can be a server, PC, etc.

[0102] The present invention also provides a computer program product, which, when executed on a data processing device, is suitable for executing an initialization program having the following method steps: acquiring training index data corresponding to the transmission link of a power optical fiber, wherein the training index data is used to characterize the quality status of the transmission link; obtaining a classification result corresponding to the training index data based on a clustering algorithm; determining a quality level label corresponding to the training index data based on the classification result; and training a predetermined initial quality model using a deep residual network algorithm based on the training index data and the quality level label corresponding to the training index data to obtain a target quality model corresponding to the transmission link.

[0103] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0104] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0105] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0106] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0107] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0108] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0109] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0110] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0111] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0112] The above are merely embodiments of the present invention and are not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A method for determining the quality model of a transmission link, characterized in that, include: Acquire training index data corresponding to the transmission link of the power optical fiber, wherein the training index data is used to characterize the quality status of the transmission link. Based on the clustering algorithm, the classification results corresponding to the training index data are obtained; Based on the classification results, the quality level label corresponding to the training indicator data is determined, including: when the training indicator data consists of multiple sets, and the multiple sets of training indicator data correspond to multiple training indicators, the coefficient of variation method is used to process the multiple sets of training indicator data to obtain the weight values ​​corresponding to the multiple training indicators respectively; based on the weight values, the quality score of each transmission link is obtained in the following manner: ,in, Indicates the first The quality score corresponding to each transmission link Indicates the first The first of multiple sets of training metric data for the transmission link The numerical values ​​of the training metrics data. For the first The weight values ​​of the training indicator data are calculated; based on the quality score of each transmission link, the average quality score of the training indicator data for each classification result is calculated; the training indicator data for each classification result is sorted according to the average quality score, and the multiple sets of training indicator data are divided into 3 levels from high to low according to the average quality score, so as to obtain the quality level label corresponding to the multiple sets of training indicator data respectively. Based on the training metric data and the quality level labels corresponding to the training metric data, a deep residual network algorithm is used to train a predetermined initial quality model to obtain the target quality model corresponding to the transmission link.

2. The method according to claim 1, characterized in that, The acquisition of training metric data corresponding to the transmission link of the power optical fiber includes: Obtain the initial data of the transmission link of the power optical fiber; Based on the variance value corresponding to the initial data, the initial data is cleaned to obtain the training index data corresponding to the transmission link.

3. The method according to claim 1, characterized in that, When the training indicator data consists of multiple sets, and these multiple sets correspond to multiple training indicators, the step of obtaining the classification result corresponding to the training indicator data based on the clustering algorithm includes: The multiple sets of training indicator data are classified to determine the positive and negative indicator data in the multiple sets of training indicator data. The positive indicator data are the training indicator data whose numerical value is directly proportional to the quality status, and the negative indicator data are the training indicator data whose numerical value is inversely proportional to the quality status. The positive indicator data is subjected to a first normalization process to obtain the processed positive indicator data; The negative index data is subjected to a second normalization process to obtain the processed negative index data; Based on the processed positive indicator data and the processed negative indicator data, the processed multiple sets of training indicator data are obtained; Based on the clustering algorithm, the processed training index data and the preset number of clusters are used to obtain the classification results corresponding to the processed training index data, wherein the clustering algorithm is a Gaussian mixture clustering algorithm.

4. The method according to claim 3, characterized in that, When there are multiple transmission links, the step of obtaining the classification results corresponding to the multiple sets of training indicator data based on the clustering algorithm and using the processed multiple sets of training indicator data and a preset number of clusters includes: Based on the clustering algorithm, using the processed training index data and a preset number of clusters, the data is processed in the following way to obtain the cluster probabilities corresponding to the processed training index data, wherein the cluster probabilities are probabilities based on a preset cluster head: ; in, The first of the multiple sets of training index data after processing The cluster probabilities of the training metric data. For multiple transmission links, the first The mean vector of each transmission link. For the first of the plurality of transmission links The covariance matrix of each transmission link. For the first of the plurality of transmission links The mixing coefficient of each transmission link, For the first of the plurality of transmission links The first of the processed training metric data in the transmission link The multivariate Gaussian probability density function of the training index data. For the identification of the multiple transmission links, The number of clusters. The identifier for the processed multiple sets of training indicator data; Based on the cluster probability and the number of clusters, the processed training index data is divided into multiple sets to obtain the classification results corresponding to the processed training index data.

5. The method according to claim 1, characterized in that, The step of training a predetermined initial quality model using a deep residual network algorithm based on the training metric data and the quality level labels corresponding to the training metric data to obtain the target quality model corresponding to the transmission link includes: Based on the training index data and the quality level label corresponding to the training index data, the deep residual network algorithm is used to train the predetermined initial quality model to obtain the trained predetermined initial quality model. The predetermined initial quality model after training is optimized based on the momentum gradient descent optimization method to obtain the target quality model. The momentum gradient descent method is a gradient descent method using the exponentially weighted moving average method.

6. The method according to any one of claims 1 to 5, characterized in that, The method further includes: Obtain the test index data of the transmission link under test; The test index data is input into the target quality model for testing, and the quality test results corresponding to the test index data of the transmission link under test are obtained.

7. A device for determining the quality model of a transmission link, characterized in that, include: The first acquisition module is used to acquire training index data corresponding to the transmission link of the power optical fiber, wherein the training index data is used to characterize the quality status of the transmission link. The classification module is used to obtain the classification results corresponding to the training indicator data based on the clustering algorithm; The determination module is used to determine the quality level label corresponding to the training indicator data based on the classification result, including: when the training indicator data consists of multiple sets, and the multiple sets of training indicator data correspond to multiple training indicators, processing the multiple sets of training indicator data using the coefficient of variation method to obtain the weight values ​​corresponding to the multiple training indicators respectively; and obtaining the quality score of each transmission link based on the weight values ​​in the following manner: ,in, Indicates the first The quality score corresponding to each transmission link Indicates the first The first of multiple sets of training metric data for the transmission link The numerical values ​​of the training metrics data. For the first The weight values ​​of the training indicator data are calculated; based on the quality score of each transmission link, the average quality score of the training indicator data for each classification result is calculated; the training indicator data for each classification result is sorted according to the average quality score, and the multiple sets of training indicator data are divided into 3 levels from high to low according to the average quality score, so as to obtain the quality level label corresponding to the multiple sets of training indicator data respectively. The second acquisition module is used to train a predetermined initial quality model using a deep residual network algorithm based on the training index data and the quality level label corresponding to the training index data, so as to obtain the target quality model corresponding to the transmission link.

8. A non-volatile storage medium, characterized in that, The non-volatile storage medium stores multiple instructions, which are adapted to be loaded by a processor and executed by the transmission link quality model determination method according to any one of claims 1 to 6.

9. An electronic device, characterized in that, include: One or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the transmission link quality model determination method according to any one of claims 1 to 6.

Images