A network performance evaluation method and device based on dynamic evidence reasoning rules

Abstract

The application discloses a network performance evaluation method based on dynamic evidence reasoning rules, which comprises the following steps: dynamically adjusting the reference value of a test index based on the change of a network state; using the adjusted reference value to convert monitoring data into a confidence distribution form, converting the test index into a corresponding evidence form according to the mapping relationship of IF-Then; fusing the evidences corresponding to multiple indexes based on the evidence reasoning rules to obtain an evaluation result; and optimizing the model parameters by using an optimization algorithm through setting the constraint conditions and optimization targets of parameters. The application further discloses a network performance evaluation device based on dynamic evidence reasoning rules. The network performance evaluation method based on dynamic evidence reasoning rules can more accurately reflect the actual performance of a network in different scenarios according to the change of the network state, and meets the evaluation requirements of different network environments and business types.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of network intelligent operation and maintenance, and more particularly to a network performance evaluation method and device based on dynamic evidential reasoning rules. BACKGROUND

[0002] With the rapid development of technologies such as 5G, edge computing and artificial intelligence, network architecture is evolving towards cloud network convergence and algorithm network convergence, which puts higher requirements on network operation and maintenance. Network performance evaluation, as an important part of intelligent operation and maintenance, directly affects the judgment of network status and the formulation of optimization decisions. However, the existing network performance evaluation methods have some limitations. Traditional evaluation methods are mainly divided into methods based on quantitative data, methods based on qualitative knowledge and methods based on mixed information. The method based on quantitative data is prone to overfitting, and the evaluation results are difficult to explain. The method based on qualitative knowledge relies on expert experience and knowledge, and the evaluation accuracy is relatively low. The method based on mixed information combines quantitative data and qualitative knowledge, and evaluates through the fusion of multi-source information, which has great application potential.

[0003] As a mixed information evaluation method, the evidential reasoning (ER) rule can convert various test indicators into evidence form and fuse through orthogonal and operation, providing a new idea for comprehensively mastering network performance. However, the evidence reference value in the ER rule is usually static and cannot adapt to the dynamic changes of network status. Therefore, the static reference value cannot accurately reflect the network performance in the face of complex and variable network environment, resulting in insufficient evaluation accuracy. This limits the effect of ER rule in practical application, and an evaluation method that can dynamically adjust the reference value is needed to improve the accuracy and reliability of evaluation. SUMMARY

[0004] In view of at least one defect or improvement demand of the prior art, the present application provides a network performance evaluation method and device based on dynamic evidential reasoning rules, which can solve at least one of the problems in the background art.

[0005] To achieve the above-mentioned purpose, according to the first aspect of the present application, a network performance evaluation method based on dynamic evidential reasoning rules is provided, which comprises:

[0006] Based on the changes of network status, the reference value of test indicators is dynamically adjusted;

[0007] The adjusted reference value is used to convert the monitoring data into a confidence distribution form, and the test indicators are converted into corresponding evidence form according to the mapping relationship of IF-Then;

[0008] The evidences corresponding to the plurality of indexes are fused based on the evidence reasoning rules to obtain an evaluation result.

[0009] By setting constraint conditions and optimization objectives of parameters, an optimization algorithm is used to optimize the model parameters.

[0010] Further, the network performance evaluation method based on the dynamic evidence reasoning rules, the network state includes a network in a stable state, a network in a slow change state and a network in a step change state.

[0011] Further, the network performance evaluation method based on the dynamic evidence reasoning rules, the network state includes a network in a stable state, a network in a slow change state and a network in a step change state.

[0012] When the network is in a stable state, there are T1 observation times, and the monitoring data corresponding to each observation time is represented as The mean value is represented as

[0013]

[0014] The distance between each monitoring data and the mean value is calculated There are

[0015]

[0016] Based on the calculated maximum distance And the minimum distance For the N reference levels G of the index n , n = 1, 2, …, N, the step amount Δd i is represented as

[0017]

[0018] Arrange the reference values in ascending order, then when N is odd, the index reference value is

[0019]

[0020] When N is even, the index reference value is

[0021]

[0022] Further, the network performance evaluation method based on the dynamic evidence reasoning rules, the network state includes a network in a stable state, a network in a slow change state and a network in a step change state.

[0023] When the network is in a slow change state, there are T2 observation times, and the monitoring data corresponding to each observation time is represented as The change amount of adjacent time ​ is represented as

[0024]

[0025] is calculated as a new observation sequence, the same calculation method as when the network is in a stable state is used to calculate the index x i The index reference value under the new observation sequence is restored to the original sequence

[0026] Further, the network performance evaluation method based on the dynamic evidence reasoning rule, the reference value of the test index is dynamically adjusted based on the change of the network state, specifically includes:

[0027] When the network is in a step change state, the step change state is divided into a stable state and a slow change state according to the change of the state, and the dynamic reference value of the index is calculated according to the stable state and the slow change state respectively.

[0028] Further, the network performance evaluation method based on the dynamic evidence reasoning rule, the adjusted reference value is used to convert the monitoring data into a confidence distribution form, and the test index is converted into a corresponding evidence form according to the mapping relationship of IF-Then, specifically includes:

[0029] The index and the corresponding dynamic reference value are described as a confidence distribution form as follows

[0030]

[0031] Where, δ n,i (t), n = 1, 2, …, N represents the index x i The confidence degree of the reference level G n at time t;

[0032] The index x i corresponding to the evidence e n,i is represented as

[0033] e n,i : If x i (t) is h i (G n ), then y(t) is {(H1, β 1,j,i ), (H2, β 2,j,i ) …, (H K , β K,j,i )}

[0034] Where, y(t) represents the system output, H k ​k = 1, 2, …, K represents different output states, β k,n,i k = 1, 2, …, K represents the confidence degree assigned to H k ;

[0035] The weight w n,i of the evidence e n,i is represented as

[0036]

[0037] wherein, represents the self weight of the index x i .

[0038] Further, the network performance evaluation method based on the dynamic evidence reasoning rule, the evidence reasoning rule is used for fusing the evidences corresponding to the plurality of indexes, and an evaluation result is obtained, and specifically, the method comprises the following steps.

[0039] All the evidences corresponding to the indexes are arranged in sequence and recorded as e = {e l ; l = 1, 2, …, L}, the evidence weights, reliabilities and confidence degrees are respectively represented as w = {w l ; l = 1, 2, …, L}, r = {r l ; l = 1, 2, …, L} and β = {β k,l ; l = 1, 2, …, L}, the total number of the evidences is L, the analytical expression form of the evidence reasoning rule is used for analysis, and the evaluation result is represented as

[0040]

[0041] wherein, η(t) is an intermediate variable and is represented as

[0042]

[0043] wherein, p k,e(L) (t) represents the confidence degree of the fusion result at the moment t relative to the level k, the utility of p k,e(L) (t) is u k,e(L) , and the numerical form of the evaluation result is represented as

[0044]

[0045] Further, the network performance evaluation method based on the dynamic evidence reasoning rule, the constraint condition and the optimization target of the parameter are specifically as follows.

[0046] The constraint condition of the parameter is as follows.

[0047] For the evidence reliability r l , there is

[0048] 0 < rl <1

[0049] For the index weight w i , there are

[0050]

[0051] wherein, and respectively represent the upper and lower bounds of the preset w i ;

[0052] For the confidence level β k,l , there are

[0053]

[0054] The optimization target of the parameters is:

[0055] For the real output of the network performance The mean square error MSE thereof is represented as

[0056]

[0057] The optimization target is established

[0058]

[0059] Further, the network performance evaluation method based on the dynamic evidence reasoning rule, the optimization algorithm is used to optimize the model parameters, including the constraint condition and the optimization target based on the parameters, the projection covariance matrix adaptive evolution algorithm is used as the optimization tool to optimize the model parameters.

[0060] According to the second aspect of the present application, a network performance evaluation device based on a dynamic evidence reasoning rule is also provided, comprising:

[0061] The adjusting module is used for dynamically adjusting the reference value of the test index based on the change of the network state;

[0062] The conversion module is used for converting the adjusted reference value into the confidence distribution form, converting the test index into the corresponding evidence form according to the mapping relationship of IF-Then, and monitoring the data;

[0063] The fusion module is used for fusing the evidences corresponding to the multiple indexes based on the evidence reasoning rule to obtain the evaluation result;

[0064] The optimization module is used for setting the constraint condition and the optimization target of the parameters, and optimizing the model parameters by using the optimization algorithm.

[0065] Overall, compared with the prior art, the above technical solutions conceived by the present application can achieve the following beneficial effects:

[0066] The network performance evaluation method based on dynamic evidence reasoning rules provided by the present application can accurately reflect the actual performance of the network in different scenarios according to the changes in the network state, ensure that the evaluation results are comprehensive and accurate, have strong adaptability and flexibility, and can be flexibly set and adjusted according to actual needs, thereby meeting the evaluation needs of different network environments and business types. BRIEF DESCRIPTION OF DRAWINGS

[0067] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0068] Figure 1 A flowchart of a network performance evaluation method based on dynamic evidence reasoning rules provided by the present application is shown in the figure.

[0069] Figure 2 A network stable state diagram provided by the present application is shown in the figure.

[0070] Figure 3 A network slow change state diagram provided by the present application is shown in the figure.

[0071] Figure 4 A network step state diagram provided by the present application is shown in the figure.

[0072] Figure 5 Some park network index observation data provided by the present application is shown in the figure.

[0073] Figure 6 An evaluation result diagram of a certain park network provided by the present application is shown in the figure. DETAILED DESCRIPTION

[0074] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and should not be used to limit the present application. In addition, the technical features involved in the various embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0075] The terms "first", "second", "third" and the like in the specification and claims of the present application and the above-described drawings are used to distinguish different objects, and are not used to describe a specific order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but can optionally include steps or units not listed or can optionally include other steps or units inherent to such processes, methods, products or devices.

[0076] Figure 1 A flowchart of a network performance evaluation method based on dynamic evidence reasoning rules provided by an embodiment of the present application is shown in FIG. 1. The network performance evaluation method based on dynamic evidence reasoning rules provided by an embodiment of the present application includes the following steps. Figure 1

[0077] Based on the change of the network state, the reference value of the test index is dynamically adjusted;

[0078] The adjusted reference value is used to convert the monitoring data into a confidence distribution form, and the test index is converted into a corresponding evidence form according to the mapping relationship of IF-Then;

[0079] Based on the evidence reasoning rule, the evidences corresponding to the multiple indexes are fused to obtain an evaluation result;

[0080] By setting the constraint conditions and optimization objectives of the parameters, an optimization algorithm is used to optimize the model parameters.

[0081] ​Specifically, the traffic, bandwidth utilization, and other key indicators of the network are monitored in real time to obtain a series of monitoring data. According to the changes in the network state, these data are divided into different time periods for analysis. For example, during working hours, the traffic is large, and the network is in a slow-changing state; while at night, the traffic is small, and the network tends to be stable. For data in the stable state, the mean and variance can be calculated to determine the reference point and fluctuation range of the reference value. For data in the slow-changing state, the trend of change can be analyzed to set the adjustment strategy of the dynamic reference value, such as gradually increasing the reference value according to the traffic growth rate. By dynamically adjusting the reference value, the performance of the network in different states can be more accurately reflected, which helps to discover potential performance problems, such as abnormal decrease in throughput caused by device failure.

[0082] After determining the dynamic reference value, the monitoring data is compared with the reference value to calculate its confidence relative to different reference levels. Taking the time delay (TD) as an example, assuming that its reference value is divided into three levels: good, medium, and poor. At a certain moment, the monitored time delay is calculated according to its proximity to the reference value, and its confidence relative to the good level, the medium level, and the poor level is calculated. The TD indicator and its confidence are represented in the form of evidence, where good, medium, and poor are different states of the system output, and the confidence assigned to each state. The monitoring data is converted into a confidence distribution form and further converted into an evidence form, providing a basis for subsequent evidence fusion. In this way, the confidence degree of the indicator at different performance levels can be quantified, making the evaluation process more scientific and accurate.

[0083] The evidences of all indicators are arranged in order, such as the evidences of TD, packet loss rate (PLR), TP, and other indicators. According to the weights, reliabilities, and confidences of each indicator, the evidences are fused using evidence reasoning rules. By calculating the product of the weights and confidences of each evidence, as well as the conflict processing between different evidences, the fusion result is obtained at a certain moment relative to the confidence of the high, medium, and low performance levels. Through evidence fusion, the influence of multiple indicators on network performance can be considered comprehensively, and a more comprehensive and accurate evaluation result can be obtained. For example, at a certain moment, although the time delay indicator performs well, the packet loss rate is high. Through fusion, the contributions of each indicator can be balanced, avoiding the excessive influence of a single indicator on the evaluation result, and ensuring that the evaluation result can truly reflect the overall performance state of the network.

[0084] The optimization target is set to minimize the mean square error (MSE), that is, the error between the evaluation result and the real state of network performance. The weights of the evidence reliability, the index weight and the confidence are respectively preset. An optimization algorithm such as a genetic algorithm or a particle swarm optimization algorithm is used to adjust the weight, the reliability and the confidence of each index under the premise of meeting the constraint condition, so that the MSE is minimized, the change of network performance can be better reflected, and the evaluation accuracy is improved. Through parameter optimization, the accuracy and stability of the evaluation model can be further improved. The optimized model parameters are more in line with the operation law of the actual network, and the evaluation result is more reliable.

[0085] The network performance evaluation method based on the dynamic evidence reasoning rule provided in the embodiments of the present application dynamically adjusts the reference value of the test index, fuses the evidence corresponding to multiple indexes based on the evidence reasoning rule, comprehensively considers the weight, the reliability and the confidence of each index, converts the monitoring data into a confidence distribution form and further converts it into an evidence form, provides a clear quantitative basis for the evaluation result, and makes the evaluation process and result have strong interpretability. In the parameter optimization process, by setting reasonable constraint conditions, the optimal balance between the evaluation accuracy and the interpretability of the evaluation result is achieved. The method can more accurately reflect the actual performance of the network in different scenarios according to the change of the network state, ensure that the evaluation result is comprehensive and accurate, has strong adaptability and flexibility, and can flexibly set and adjust the evaluation parameters according to actual needs, and meet the evaluation needs of different network environments and business types.

[0086] Optionally, the network performance evaluation method based on the dynamic evidence reasoning rule provided in the embodiments of the present application, the network state includes that the network is in a stable state, the network is in a slow change state and the network is in a step change state.

[0087] Specifically, in the actual network environment, the network state will change due to various factors. For example, in a data center network, when the number of users is small and the business traffic is stable, the network is in a stable state; when a large number of users access simultaneously and the business traffic gradually increases, the network is in a slow change state; when a large-scale activity causes the number of users and the business traffic to suddenly increase, the network is in a step change state. By monitoring the network traffic, bandwidth utilization and other indexes in real time, it can be determined which state the network is in, and the corresponding evaluation strategy is adopted according to different states.

[0088] Optionally, the network performance evaluation method based on the dynamic evidence reasoning rule provided in the embodiments of the present application, the reference value of the test index is dynamically adjusted based on the change of the network state, and specifically includes:

[0089] When the network is in a stable state, there are T1 observation times, and the monitoring data corresponding to each observation time is represented as The average value of which is expressed as

[0090]

[0091] The distance between each of the monitoring data and the average value is calculated There are

[0092]

[0093] Based on the calculated maximum distance And the minimum distance For the N reference levels G of the index n , n = 1, 2, …, N, the step amount Δd i Is expressed as

[0094]

[0095] The reference values are arranged in ascending order, and when N is odd, the reference values of the index are

[0096]

[0097] When N is even, the reference values of the index are

[0098]

[0099] Specifically, when the network is in a stable state, the monitoring data x i (t) presents a trend around a certain constant, as shown in Figure 2 At this time, the reference values of the index should be adjusted on the basis of the constant.

[0100] Suppose that there are M test indexes of the network, and the monitoring data of each index at time t is expressed as x i (t), i = 1, 2, …, M, t = 1, 2, …, T, where T represents the number of all observation times.

[0101] In the network stable state, continuous data collection is performed on a certain key indicator, and the data is sorted to obtain a data sequence. The mean of the collected data sequence is calculated as the reference value of the indicator in the stable state. The distance between each observation data and the mean, i.e. the difference between each data point and the mean, is calculated. From all the calculated distances, the maximum distance and the minimum distance are found. The maximum distance represents the maximum deviation of the data point from the mean, and the minimum distance represents the minimum deviation of the data point from the mean. The number of reference levels of the indicator is calculated to determine the step amount, which is based on the maximum distance and the minimum distance and is used to divide the intervals between different reference levels. The reference values are arranged in ascending order, and the reference values of different reference levels are calculated based on the step amount and the mean. When N is odd, the reference values of the indicator are centered on the mean and extended to both sides; when N is even, the reference values are calculated differently, but are still based on the mean and the step amount. According to the above process, h i (G n ) can fully reflect the indicator x i The change trend in different stages.

[0102] By calculating the mean as the reference value, the evaluation result can better represent the overall performance of the network in this state, avoiding the influence of individual data points on the evaluation result. By calculating the distance, determining the maximum and minimum distance, and calculating the step amount, the boundaries between different performance levels can be reasonably divided. The subtle changes in network performance can be more detailed, providing clear quantitative standards for different levels of performance. According to actual needs, different numbers of reference levels are set, and the calculation method of the step amount and the reference value is adjusted accordingly. This flexibility makes the evaluation method adaptable to different evaluation accuracy requirements and application scenarios, improving the applicability and practicality of the evaluation method.

[0103] Optionally, the network performance evaluation method based on dynamic evidence reasoning rules provided by the embodiments of the present application dynamically adjusts the reference value of the test indicator based on the change of the network state, specifically including:

[0104] When the network is in a slow change state, there are T2 observation times, and the monitoring data corresponding to each observation time is represented as The change amount of adjacent time is represented as The change amount of adjacent time

[0105]

[0106] Take as a new observation sequence, and use the same calculation method as when the network is in a stable state to calculate the indicator x i The reference value of the indicator in the new observation sequence The reference value of the index is reduced to the original sequence The reference value of the index is reduced to the original sequence

[0107] Specifically, when the network is in a slow change state, the monitored data x i (t) presents a regular increasing or decreasing trend, as shown in Figure 3 At this time, the data can be preprocessed to observe the change amount of x i (t).

[0108] When it is monitored that the network is in a slow change state, continuous data collection is performed on the key index. Since the data presents a slow change trend, the data needs to be preprocessed to highlight its change characteristics. The change amount of the data at adjacent observation time is calculated, that is, the difference between each data point and its previous data point. The sequence reflects the change rate of the time delay in the slow change state. The change amount sequence is taken as new observation data, and the same calculation method as in the stable state is used to calculate the reference value of the index under the sequence. First, the mean of the change amount sequence is calculated, then the distance between each change amount and the mean is calculated, the maximum and minimum distances are determined, then the step amount is calculated, and the reference values of different reference levels are determined according to the step amount and the mean. The calculated reference value is reduced to the original sequence to obtain the dynamic reference value of the time delay index in the slow change state.

[0109] By calculating the change amount of the data at adjacent time, the performance change trend of the network in the slow change state can be accurately captured. Taking the change amount sequence as new observation data for processing makes the evaluation method adapt to the characteristics of the slow change of the network. In the slow change state, the change amplitude of the data is small, but the duration is long, and by analyzing the change amount, the stability of the network performance can be more accurately evaluated. By reducing the reference value to the original sequence, dynamic adjustment of the reference value of the time delay index is realized. This dynamic adjustment mechanism makes the reference value change with the change of the network state, avoiding the evaluation deviation caused by the fixed reference value.

[0110] Optionally, the network performance evaluation method based on the dynamic evidence reasoning rule provided in the embodiments of the present application comprises the following steps:

[0111] When the network is in a step change state, the step change state is divided into a stable state and a slow change state according to the change of the state, and the dynamic reference value of the index is calculated according to the stable state and the slow change state respectively.

[0112] Specifically, when the network is in a step change state, the monitored data x i (t) presents a dramatic change trend, as shown in Figure 4The state can be regarded as a combination of the above two states. For example, x i (t) from the stable state to the slow change state. In this state, the dynamic reference value of the index can be calculated in reference to the stable state and the slow change state.

[0113] In monitoring network performance, the data of the key index is collected in real time, and it is identified that the network is in the step change state. For example, the data of the TP index suddenly changes from the stable state to the slow change state in a certain period of time, and such mutation is the step change. The step change state is divided into two stages of stable state and slow change state. For the data of the stable state part, the reference value calculation method of the stable state is adopted to calculate the reference value in the stable state. For the data of the slow change state part, the change amount of the adjacent time data is calculated to obtain a change amount sequence, and the same method as the stable state is adopted to calculate the reference value in the slow change state. The reference values of the stable state and the slow change state are integrated to obtain the dynamic reference value of the TP index in the step change state, which indicates that the reference value of the TP index gradually transitions from the level of the stable state to the level of the slow change state in the step change process.

[0114] Optionally, the network performance evaluation method based on the dynamic evidence reasoning rule provided in the embodiments of the present application comprises the following steps:

[0115] The index and the corresponding dynamic reference value are described in the form of the confidence distribution as follows

[0116]

[0117] Wherein, δ n,i (t), n = 1, 2, …, N indicates that the index x i The confidence degree of the reference level G n at time t.

[0118] The index x i Corresponding evidence e n,i is expressed as

[0119] e n,i : If x i (t) is h i (G n ), then y(t) is {(H1, β 1,j,i ), (H2, β 2,j,i ) …, (H K , β K,j,i )}

[0120] where y(t) represents the system output, H k ,k=1,2,…,K represents different output states, β k,n,i ,k=1,2,…,K represents the confidence degree assigned to H k .

[0121] The weight w n,i of the evidence e n,i is represented as

[0122]

[0123] where, represents the self weight of the index x i .

[0124] Specifically, based on the above three states, the index x i can be represented completely by the dynamic reference value h i (G n ). According to the rule-based information transformation method, the input information can be described in the form of lower confidence distribution. The evidence corresponding to the key index is calculated.

[0125] By calculating the confidence degree, the comparison result of the monitoring data and the dynamic reference value is quantified into a specific numerical value, so that the performance of the index can be presented in a standardized form. The introduction of the dynamic reference value enables the evaluation method to be flexibly adjusted according to the change of the network state, and to adapt to the dynamic change of the network performance. The monitoring data is transformed into the evidence form, and the confidence degree of the index under different performance levels is comprehensively considered, so that the evaluation result is more comprehensive and accurate.

[0126] Optionally, the network performance evaluation method based on dynamic evidence reasoning rules provided by the embodiment of the application comprises the following steps:

[0127] All the evidences corresponding to the indexes are arranged in order, denoted as e={e l ;l=1,2,…,L}, and the evidence weight, reliability and confidence degree are respectively denoted as w={w l ;l=1,2,…,L}, r={r l ;l=1,2,…,L} and β={β k,l ;l=1,2,…,L}, the total number of evidences is L, the analytical expression form based on the evidence reasoning rule is η(t)=f(e

[0128]

[0129] where η(t) is an intermediate variable, denoted as

[0130]

[0131] wherein p k,e(L) (t) represents the confidence of the fusion result at time t relative to level k, p k,e(L) (t) has a utility u k,e(L) , and the numerical form of the evaluation result is represented as

[0132]

[0133] Specifically, evidence of each key indicator in the network is collected, the evidence is arranged according to the importance or order of the indicators, and the evidence of each indicator is assigned a weight and a reliability, the weight reflecting the importance of each indicator in the overall network performance evaluation, and the reliability representing the credibility of the evidence. According to the evidence reasoning rule, the evidence of multiple indicators is fused. First, the weighted confidence of each evidence is calculated, and then the weighted confidences of all the evidence are fused to obtain a comprehensive evaluation result. According to the fused evaluation result, the final state of the network performance is determined. For example, if the confidence of the “medium” state in the evaluation result is the highest, it is judged that the network performance is in the medium state.

[0134] Through evidence fusion, the influence of multiple indicators on network performance can be considered comprehensively, and the limitation of a single indicator on the evaluation result is avoided. The introduction of weight and reliability makes the evaluation process more precise. The weight is assigned according to the importance of the indicators, ensuring that the key indicators occupy a larger proportion in the evaluation; the reliability reflects the credibility of the evidence, which helps to reduce the interference of unreliable evidence on the evaluation result, thereby improving the accuracy of the evaluation.

[0135] Optionally, the network performance evaluation method based on the dynamic evidence reasoning rule provided in the embodiments of the present application sets the constraint condition and the optimization target of the parameter, which specifically includes:

[0136] The constraint condition of the parameter is:

[0137] For the evidence reliability r l , there is

[0138] 0 < r l < 1

[0139] For the indicator weight w i , there is

[0140]

[0141] wherein and respectively represent the upper and lower bounds of the preset w i ;

[0142] For the confidence β k,l , there is

[0143]

[0144] The optimization goal of the parameters is to minimize the mean square error (MSE) between the evaluation result and the true state of the network performance.

[0145] The true output of network performance The mean square error (MSE) is represented as

[0146]

[0147] Establishing the optimization goal

[0148]

[0149] Specifically, in the above DER rule-based network performance evaluation model, the parameters to be optimized include the evidence reliability r l , the index weight w i , and the confidence level β k,l . To make the evaluation result interpretable, the physical meaning of the parameters before and after optimization needs to be kept unchanged, so it is necessary to set reasonable constraint conditions for the parameters in the evaluation model.

[0150] The evidence reliability r l reflects the ability of the information source to provide evidence results. For the evidence reliability, the constraint condition is 0 < reliability < 1, which ensures that the reliability is within a reasonable range and avoids distortion of the evaluation result caused by excessively high or low reliability. The weight w i reflects the relative importance between different indicators. For the index weight, the constraint condition is 0 < weight < 1, and the sum of the weights of all indicators is 1, which ensures that the weight distribution is reasonable and reflects the relative importance of each indicator in the evaluation. In addition, it is necessary to avoid some extreme weight values. The confidence level β k,l is a generalized probability. For the confidence level, the constraint condition is 0 ≤ confidence level ≤ 1, which ensures the non-negativity and normalization of the confidence level.

[0151] The optimization goal is set to minimize the mean square error (MSE), i.e. the error between the evaluation result and the true state of the network performance. The setting of the optimization goal aims to improve the accuracy of the evaluation model and make the evaluation result as close as possible to the true performance of the network.

[0152] By optimizing the model parameters, the error between the evaluation result and the true state of the network performance is minimized, significantly improving the accuracy of the evaluation model. During the optimization process, the constraint conditions ensure that the physical meaning of the parameters before and after optimization remains unchanged, making the optimized evaluation result clearly interpretable. The introduction of the optimization algorithm enhances the robustness of the evaluation model, enabling it to maintain high evaluation performance in different network environments and data distributions, adapt to changes in network state and fluctuations in data, and avoid evaluation errors caused by unreasonable parameters.

[0153] Optionally, the network performance evaluation method based on dynamic evidence reasoning rules provided by the embodiment of the application optimizes the model parameters by using an optimization algorithm, includes constraint conditions and optimization objectives based on parameters, and optimizes the model parameters by using a projection covariance matrix adaptive evolution algorithm as an optimization tool.

[0154] Specifically, the projection covariance matrix adaptive evolution algorithm (PCMA-EA) is used as an optimization tool. The PCMA-EA is a population-based optimization algorithm that continuously iteratively optimizes model parameters by simulating mutation, selection, and recombination operations in the biological evolution process. The specific steps include parameter initialization, Gaussian sampling, parameter projection, population update, and termination operation.

[0155] Under the premise of meeting the constraint conditions, the PCMA-EA is used to optimize the model parameters. First, the population size, optimization step, and optimization times are initialized. Then, the initial population is generated by Gaussian sampling, and the parameters are projected according to the constraint conditions to ensure that the parameters meet the constraints. Next, the population is continuously updated according to the size of the MSE and the parameter constraints, and the inferior individuals are eliminated and the superior individuals are retained. Finally, when the optimization times reach the set threshold or the MSE meets the requirements, the optimization is terminated, and the optimized model parameters are obtained.

[0156] As an efficient optimization algorithm, the PCMA-EA can quickly converge to the optimal solution, shortening the time of model parameter optimization. Compared with traditional manual adjustment or simple optimization methods, the PCMA-EA can significantly improve the optimization efficiency and accelerate the development and application process of the evaluation model.

[0157] In one specific embodiment, in order to verify the effectiveness of the proposed method, a case study is carried out on a certain campus network, including the following steps:

[0158] 1. Establishment of performance evaluation model

[0159] In this embodiment, the time delay (TD), packet loss rate (PLR), and throughput (TP) are selected as the performance evaluation indicators of the campus network. By checking the historical running information of the network, 100 groups of data of the three indicators are obtained, as shown in Table 1. Figure 5

[0160] Based on expert knowledge, three evaluation levels {good, medium, and poor} are set for the campus network and the observed data of the indicators. The evaluation results are shown in Table 2. Figure 5 ​It can be seen that the observation data of TD and PLR belong to a slowly changing state, and the corresponding reference values are static, as shown in Table 1.

[0161] Table 1 Static reference value of index

[0162]

[0163]

[0164] In addition, from the observation data of TP, Figure 5 It can be seen that the observation data of TP is from a slowly changing state to a stable state, so TP needs to set a dynamic reference value. The setting of the dynamic reference value is shown in Table 2.

[0165] Table 2 Dynamic reference value of index

[0166]

[0167] According to Table 2, the reference values of TP at the first to 60th time points are 4, 3 and 2, respectively. Since after the 60th time point, TP transitions to another state, its reference value is dynamically updated to 2.8, 2.25 and 1.7. At the same time, it is assumed that the network performance state can be described by three levels of high (High, H), medium (Medium, M) and low (Low, L). A performance evaluation model based on DER is constructed. Since some model parameters cannot be accurately set, parameter optimization of the evaluation model is needed.

[0168] 2. Optimization of performance evaluation model

[0169] The above performance evaluation model is optimized, wherein 50 groups of data are used as training data, and the rest are used as test data. The optimized model parameters are shown in Table 3. In addition, the optimized index weights are 1, 0.95 and 0.57, respectively. By using the parameters in Table 3, the output results of the model on the test set are shown in Table 4. Figure 6 The corresponding MSE can be calculated as 0.0268. Figure 6 It can be seen that the optimization result basically conforms to the real performance state of the campus network.

[0170] Table 3 Optimized model parameters

[0171]

[0172]

[0173] The embodiment of the application also provides a network performance evaluation device based on a dynamic evidence reasoning rule, comprising:

[0174] The adjusting module is configured to dynamically adjust the reference values of the test indexes based on the changes of the network state.

[0175] a transformation module, configured to use the adjusted reference value to transform the monitoring data into a belief distribution form, and transform the test indexes into corresponding evidence forms according to the IF-Then mapping relationship;

[0176] a fusion module, configured to fuse the evidences corresponding to the plurality of indexes based on the evidence reasoning rules to obtain an evaluation result;

[0177] an optimization module, configured to optimize the model parameters by setting constraint conditions and optimization targets of the parameters, and using an optimization algorithm.

[0178] It should be noted that, for the foregoing method embodiments, in order to simply describe, they are all described as a series of action combinations, but those skilled in the art should know that the present application is not limited to the action sequence described, because according to the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should know that the embodiments described in the specification all belong to preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0179] In the above embodiments, the description of each embodiment has its own emphasis, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0180] In several embodiments provided by the present application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are only schematic. The division of the units is only a logical function division. There can be another division manner in actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between the units can be indirect couplings or communication connections through some interfaces, devices or units, and can be electrical or other forms.

[0181] The units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, that is, they can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0182] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The above integrated unit can be realized in the form of hardware, or in the form of software functional unit.

[0183] The above-described embodiments are merely exemplary embodiments of the present disclosure, and the present disclosure is not limited thereto. That is, any equivalent changes and modifications made in accordance with the teachings of the present disclosure are still included in the scope of the present disclosure. Embodiments of the present disclosure will be readily apparent to those skilled in the art in view of the disclosure herein with the description and practice of the present disclosure. The present application is intended to cover any variations, uses, or adaptations of the present disclosure following the general principles thereof and including such insubstantial variations and modifications thereof as come within the scope of the present disclosure. The scope of the present disclosure is defined by the appended claims rather than the description and embodiments thereof.

[0184] Any of the technical features of the above embodiments can be combined, and for the sake of brevity, not all possible combinations of the various technical features described above are repeated, however, any combination of the technical features should be considered as within the scope of the present disclosure, as long as the combination does not result in a contradiction.

[0185] Those skilled in the art easily understand that the above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modifications, equivalent replacements and improvements made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A network performance evaluation method based on dynamic evidence reasoning rules, characterized in that: include: Dynamically adjust the reference value of test indicators based on changes in network status; The adjusted reference values ​​are used to convert monitoring data into confidence distribution form, and the test indicators are converted into corresponding evidence forms according to the IF-Then mapping relationship; Based on the evidence reasoning rules, the evidence corresponding to multiple indicators is integrated to obtain the evaluation results; By setting the constraint conditions and optimization objectives of the parameters, the optimization algorithm is used to optimize the model parameters; The network status includes the network being in a stable state, the network being in a slowly changing state, and the network being in a step-changing state; The dynamically adjusting the reference value of the test indicator based on the change of the network status specifically includes: When the network is in a stable state, there are observation time, and the monitoring data corresponding to each observation time is expressed as , whose mean is expressed as Calculate the distance between each monitoring data and the mean ,have Based on the calculated maximum distance and minimum distance , for this indicator N Reference levels , its step amount Expressed as Arrange the reference values ​​in ascending order. N When it is an odd number, the reference value of the indicator is when N When it is an even number, the reference value of the indicator is 。 2. The network performance evaluation method based on dynamic evidence reasoning rules according to claim 1, characterized in that: The dynamically adjusting the reference value of the test indicator based on the change of the network status specifically includes: When the network is in a slowly changing state, there is a observation time, and the monitoring data corresponding to each observation time is expressed as , adjacent moments The amount of change Expressed as Will As a new observation sequence, the same calculation method as when the network is in a stable state is used to calculate the indicator The reference value of the indicator under the new observation sequence ,Will The indicator reference value restored to the original sequence .

3. The network performance evaluation method based on dynamic evidence reasoning rules according to claim 2 is characterized in that: The dynamically adjusting the reference value of the test indicator based on the change of the network status specifically includes: When the network is in a step change state, the step change state is decomposed into a stable state and a slowly changing state according to the change of the state, and the dynamic reference value of the indicator is calculated according to the stable state and the slowly changing state respectively.

4. The network performance evaluation method based on dynamic evidence reasoning rules according to claim 3, characterized in that: The adjusted reference value is used to convert the monitoring data into a confidence distribution form, and the test indicators are converted into corresponding evidence forms according to the IF-Then mapping relationship, specifically including: The indicators and corresponding dynamic reference values ​​are described as confidence distribution forms as follows in, Indicator At the moment Relative to reference level confidence level; index The corresponding evidence Expressed as :if yes , is ; in, Indicates the system output, Indicates different output states, Indicates assignment to confidence level; evidence Weight Expressed as in, Indicator its own weight.

5. The network performance evaluation method based on dynamic evidence reasoning rules according to claim 3, characterized in that: The evidence corresponding to multiple indicators is integrated based on the evidence reasoning rules to obtain the evaluation results, specifically including: Arrange the evidence corresponding to all indicators in order, and record them as , the weight of evidence, reliability and confidence are expressed as 、 and The total number of evidences is , based on the analytical expression of the evidence reasoning rules, the evaluation results are expressed as in, is an intermediate variable, expressed as in, Indicates the fusion result at time t Relative to level The confidence level, The utility is , the numerical form of the evaluation result is expressed as 。 6. The network performance evaluation method based on dynamic evidence reasoning rules according to claim 5 is characterized in that: The constraints and optimization objectives of the parameter settings specifically include: The constraints on the parameters are: Regarding the reliability of evidence ,have For indicator weights ,have in, and Respectively indicate pre-set The upper and lower bounds of For confidence ,have The optimization objectives of the parameters are: Real output for network performance , its mean square error MSE Expressed as Establish optimization goals 。 7. The network performance evaluation method based on dynamic evidence reasoning rules according to claim 6 is characterized in that: The optimization algorithm is used to optimize the model parameters, including parameter-based constraints and optimization targets, and the projection covariance matrix adaptive evolution algorithm is used as an optimization tool to optimize the model parameters.

8. A network performance evaluation device based on dynamic evidence reasoning rules, characterized in that: include: An adjustment module is used to dynamically adjust the reference value of the test indicator based on changes in network status; The conversion module is used to convert the monitoring data into a confidence distribution form using the adjusted reference value, and convert the test indicators into corresponding evidence forms according to the IF-Then mapping relationship; The fusion module is used to fuse the evidence corresponding to multiple indicators based on the evidence reasoning rules to obtain the evaluation results; The optimization module is used to optimize the model parameters by setting the constraint conditions and optimization objectives of the parameters and using the optimization algorithm; The network status includes the network being in a stable state, the network being in a slowly changing state, and the network being in a step-changing state; The dynamically adjusting the reference value of the test indicator based on the change of the network status specifically includes: When the network is in a stable state, there are observation time, and the monitoring data corresponding to each observation time is expressed as , whose mean is expressed as Calculate the distance between each monitoring data and the mean ,have Based on the calculated maximum distance and minimum distance , for this indicator N Reference levels , its step amount Expressed as Arrange the reference values ​​in ascending order. N When it is an odd number, the reference value of the indicator is when N When it is an even number, the reference value of the indicator is 。

Images