Network fault prediction method and device based on extended belief rule base

Abstract

The invention discloses a network fault prediction method based on an extended belief rule base, and the method comprises the steps: constructing an extended belief rule based on the state of a network at a past moment, a corresponding reference level and reference value, and a preset attenuation factor, and converting the state of the network at the past moment into a first matching degree; the activation weight of each confidence rule is calculated, and for each activated confidence rule, the basic probability quality allocated to different fault levels is calculated; performing iterative fusion on the basic probability quality of the multiple confidence rules, and obtaining a prediction result of the network fault according to the basic probability quality after iterative fusion; and the rules with the weights lower than the statistical utility are cleared away, the number of the rules is reduced, conditions needing to be met in the parameter optimization process are preset and solved, and collaborative optimization of the structural parameters is achieved. The invention further discloses a network fault prediction device based on the extended belief rule base. The method can effectively predict the network fault, and improves the stability and reliability of the network.

Description

Technical Field

[0001] The present application relates to the technical field of network fault prediction, and more specifically, to a network fault prediction method and device based on an extended confidence rule base. Background Art

[0002] With the rapid development of information technology, network technology has become the cornerstone of modern society. However, as the scale of the network continues to expand, the frequency of network failures has also increased significantly. Studies have shown that fault prediction can detect potential network problems, take measures to avoid risks, reduce service interruptions or delays caused by network failures, and improve user experience and satisfaction. Common fault prediction methods can be divided into data-driven methods, knowledge-driven methods, and semi-quantitative information-driven methods. In contrast, semi-quantitative information-driven methods can make comprehensive use of quantitative data and qualitative knowledge, and therefore have greater application potential in the field of fault prediction. As a reasoning method based on semi-quantitative information, the Belief Rule Base (BRB) can achieve accurate modeling of complex systems through highly readable belief rules and a transparent fusion process.

[0003] However, there are still some problems with the current fault prediction methods based on BRB. In engineering practice, past information is usually used to predict network status. As time goes by, the credibility of information will gradually decay. However, the fusion process of confidence rules in BRB is static, and the decay effect of information credibility is not considered. This method makes time information have no effect on the output results, which reduces the prediction accuracy. In addition, the existing BRB optimization methods mainly focus on attribute weights, rule weights and input reference values. Among them, attribute weights and rule weights represent the relative importance of different attributes and rules, and their values ​​are generally given by experts. In order to eliminate the limitations of expert knowledge, model optimization is required to achieve accurate weight assignment. However, the above research focuses on the optimization of model parameters and ignores the structural complexity of the model, causing the prediction results to deviate from expectations. Summary of the invention

[0004] In view of at least one defect or improvement need in the prior art, the present invention provides a network fault prediction method and device based on an extended confidence rule base, which can solve at least one of the technical problems existing in the background technology.

[0005] To achieve the above object, according to a first aspect of the present invention, a network fault prediction method based on an extended confidence rule base is provided, the method comprising:

[0006] Based on the state of the network at the past moment, the corresponding reference level and the reference value, a decay factor is preset, and an extended confidence rule is constructed to convert the state of the network at the past moment into its first matching degree relative to the corresponding reference level;

[0007] Calculate the activation weight of each confidence rule according to the first matching degree, the reference level and the reference value of the confidence rule, and for each activated confidence rule, calculate the basic probability mass assigned to different fault levels;

[0008] Through the orthogonal and fusion mechanism, the basic probability masses of multiple confidence rules are iteratively integrated, and the prediction results of network faults are obtained according to the basic probability masses after iterative integration;

[0009] According to the size of the rule weight, the rules with weights lower than the preset statistical utility are cleared. Based on the conditions that need to be met in the preset parameter optimization process, the intelligent optimization algorithm is used to solve the problem and realize the collaborative optimization of the structural parameters.

[0010] Furthermore, the network fault prediction method based on the extended confidence rule base, based on the network status at past moments, the corresponding reference level and reference value, preset attenuation factors, and constructing extended confidence rules, specifically includes:

[0011] The predicted network state is y(t), and the network states at M past moments are y(t-τ), τ=1,…,M, M past moments are equivalent to M input attributes, and the preset attenuation factor is α l , then the extended confidence rule is constructed as

[0012] R l :If y(t-τ)isF(H n ),Then y(t)is(H n ,β n,l )

[0013] withattribute weightδ τ , rule weightθ l anddecayfactorα l

[0014] Among them, H n , n=1,2,…,N represents the reference level of y(t-τ), F(H n ) represents the corresponding reference value of the reference level, β n,l Indicates the predicted network state relative to H n Confidence level;

[0015] The attenuation factor is α l for

[0016]

[0017] Among them, λ l represents the decay rate, Δt l Indicates the decay time.

[0018] Furthermore, in the above-mentioned network fault prediction method based on the extended confidence rule base, the converting of the network state at a past time into its first matching degree relative to the corresponding reference level specifically includes:

[0019] Convert the numerical form of y(t-τ) into

[0020]

[0021] Among them, n,τ (t) represents y(t-τ) relative to H n The first match.

[0022] Furthermore, the above-mentioned network fault prediction method based on the extended confidence rule base, the calculation of the activation weight of each confidence rule according to the first matching degree, the reference level and the reference value of the confidence rule, specifically includes:

[0023] Activation weight w l (t) is expressed as

[0024]

[0025] Where L represents the total number of rules, θ l represents the rule weight, δ τ Represents the attribute weight.

[0026] Furthermore, in the above-mentioned network fault prediction method based on the extended confidence rule base, for each activated confidence rule, calculating the basic probability mass assigned to different fault levels specifically includes:

[0027] If a rule is activated, its basic probability mass is calculated as

[0028]

[0029] Among them, m n,l (t) indicates the value assigned to level H n The basic probability mass, m P(Θ),l (t) represents the basic probability mass assigned to the power set P(Θ), which is given by 2 N A complete set of levels.

[0030] Furthermore, the network fault prediction method based on the extended confidence rule base, the basic probability masses of multiple confidence rules are iteratively integrated through the orthogonal and fusion mechanism, specifically including:

[0031] Through the orthogonal and fusion mechanism, the first L1 of the L rules are fused into

[0032]

[0033] in, Indicates that the L1 fused rules are assigned to H n The normalized basic probability mass of represents the normalized basic probability mass assigned to P(Θ) by the L1 fused rules, Indicates that the L1 fused rules are assigned to H n The unnormalized basic probability mass of represents the unnormalized basic probability mass assigned to P(Θ) by the L1 fused rules, and Expressed as

[0034]

[0035] Furthermore, the network fault prediction method based on the extended confidence rule base, obtaining the prediction result of the network fault according to the basic probability mass after iterative fusion, specifically includes:

[0036] Iterative Calculation Until all rules are involved in the fusion, the final fusion result is expressed as

[0037]

[0038] Among them, β n,R(L) (t) indicates the fusion result relative to H n confidence level;

[0039] Level H n The reference value is F(H n ), then the prediction result U R(L) (t) is expressed as follows:

[0040]

[0041] Furthermore, the network fault prediction method based on the extended confidence rule base, according to the size of the rule weight, removes the rules whose weight is lower than the statistical utility, and simplifies the number of rules, specifically including:

[0042] The number of rules is simplified as shown below:

[0043]

[0044] in, is the simplified rule R l The weight, μ l For rule R l The statistical utility of is obtained by the following formula:

[0045]

[0046] Among them, μ T is the threshold of statistical utility;

[0047] After simplification, the calculation method of activation weights is modified as follows:

[0048]

[0049] in, represents the reduced activation weight, Indicates the number of rules after simplification.

[0050] Furthermore, in the above-mentioned network fault prediction method based on the extended confidence rule base, the conditions that need to be met in the preset parameter optimization process are solved by an intelligent optimization algorithm to achieve collaborative optimization of structural parameters, specifically including:

[0051] The actual output is The output of the prediction model U R(L) (t) The mean square error MSE between

[0052]

[0053] Where T represents the number of training data;

[0054] The optimization objective is expressed as follows

[0055] minMSE

[0056] st.

[0057] 0<δ τ <1; 0<θ l <1;

[0058] λ l ≥0;

[0059] 0≤β n,l ≤1;

[0060] in, represents the true result, γ is the adjustment factor, which is used to convert the prediction error and the number of rules to the same order of magnitude;

[0061] The collaborative optimization of structural parameters is achieved through intelligent optimization algorithm.

[0062] According to a second aspect of the present invention, there is also provided a network fault prediction device based on an extended confidence rule base, comprising:

[0063] A construction module, for presetting a decay factor based on the state of the network at a past moment, the corresponding reference level and the reference value, constructing an extended confidence rule, and converting the state of the network at a past moment into a first matching degree relative to the corresponding reference level;

[0064] an acquisition module, configured to calculate the activation weight of each confidence rule according to the first matching degree, the reference level and the reference value of the confidence rule, and for each activated confidence rule, calculate the basic probability mass assigned to different fault levels;

[0065] The prediction module is used to iteratively fuse the basic probability masses of multiple confidence rules through orthogonal and fusion mechanisms, and obtain the prediction result of network faults according to the basic probability masses after iterative fusion;

[0066] The optimization module is used to eliminate rules whose weights are lower than the statistical utility according to the size of the rule weights, simplify the number of rules, establish the optimization goals of the structural parameters, preset the conditions that need to be met during the parameter optimization process, and solve them through the intelligent optimization algorithm to achieve collaborative optimization of the structural parameters.

[0067] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art:

[0068] The present invention provides a network fault prediction method based on an extended confidence rule base, which significantly improves the prediction accuracy and real-time performance by introducing a coordinated optimization strategy of attenuation factors and structural parameters. The method can comprehensively utilize quantitative data and qualitative knowledge, effectively simulate the changes of network status over time, reduce the impact of network failures, improve the stability and reliability of the network, and reduce maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] 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 below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0070] Figure 1 A flowchart of a network fault prediction method based on an extended confidence rule base provided in an embodiment of the present application;

[0071] Figure 2 A schematic diagram of network monitoring data provided by an embodiment of the present application;

[0072] Figure 3 A schematic diagram of network data prediction results before and after optimization provided in an embodiment of the present application. DETAILED DESCRIPTION

[0073] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0074] The terms "first", "second", "third", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. 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 optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.

[0075] Figure 1 A flowchart of a network fault prediction method based on an extended confidence rule base provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, a network fault prediction method based on an extended confidence rule base provided in an embodiment of the present application includes the following steps:

[0076] 101 Based on the state of the network at a past moment, the corresponding reference level and the reference value, a decay factor is preset, and an extended confidence rule is constructed to convert the state of the network at a past moment into a first matching degree relative to the corresponding reference level;

[0077] 102, according to the first matching degree, the reference level and the reference value of the confidence rule, calculating the activation weight of each confidence rule, and for each activated confidence rule, calculating the basic probability mass assigned to different fault levels;

[0078] 103 Through the orthogonal and fusion mechanism, the basic probability masses of multiple confidence rules are iteratively integrated, and the prediction result of network fault is obtained according to the basic probability masses after iterative integration;

[0079] 104 According to the size of the rule weight, the rules whose weight is lower than the preset statistical utility are eliminated, and based on the conditions that need to be met in the preset parameter optimization process, the intelligent optimization algorithm is used to solve and realize the collaborative optimization of the structural parameters.

[0080] Specifically, based on the state of the network at past moments, the corresponding reference level and reference value, a decay factor is preset, and an extended confidence rule is constructed to convert the state of the network at past moments into its first degree of match relative to the corresponding reference level. For each past moment, the degree of match with the reference level, i.e., the first degree of match, is calculated based on its actual observed value and the preset reference value. This step is the basis for subsequent predictions and ensures that the model can take into account the influence of time factors. The decay factor is introduced to simulate the decay of the credibility of information over time, which can be determined based on expert knowledge or data optimization to reflect the decay of the credibility of the rule.

[0081] First, the matching degree, the reference level and reference value of the confidence rule are used to calculate the activation weight of each confidence rule. The activation weight reflects the degree of activation of the confidence rule during the fusion process, and its value can be calculated as a function of the rule weight and the attribute weight, while considering the influence of the attenuation factor. For each activated confidence rule, calculate the basic probability mass (Basic Probability Mass, BPM) assigned to different fault levels. BPM is a quantity in the confidence rule base used to represent the confidence of the rule for different results, which reflects the contribution of the rule to the prediction result. This step is the core of the prediction model, ensuring that the model can accurately predict faults based on the current network status and historical data.

[0082] Through orthogonal and fusion mechanisms, the basic probability masses of multiple confidence rules are iteratively fused. Based on the basic probability masses after iterative fusion, the prediction results of network failures are obtained. The BPMs of each rule are merged to obtain a comprehensive prediction of the future state of the network. The iterative fusion process allows the model to dynamically adjust the weights of each rule to reflect their contribution to the final prediction results. Based on the basic probability masses after iterative fusion, the prediction results of network failures are obtained. Through the iterative fusion process, the BPMs of multiple confidence rules are merged to obtain a comprehensive prediction of the future state of the network, which improves the accuracy and reliability of the prediction.

[0083] According to the size of the rule weight, the rules with weights lower than the statistical utility are removed to simplify the number of rules. The conditions that need to be met during the preset parameter optimization process are solved by the intelligent optimization algorithm to achieve the collaborative optimization of structural parameters. The complexity of the model is reduced by removing rules that have little effect on the prediction results, thereby improving the efficiency of the model. The conditions that need to be met during the preset parameter optimization process are solved by the intelligent optimization algorithm to achieve the collaborative optimization of structural parameters. The model complexity is reduced by simplifying the number of rules, and the model parameters are adjusted by the optimization algorithm to achieve the best balance between prediction accuracy and model complexity, thereby improving the practicality and efficiency of the model. The constraints in the parameter optimization process include that the physical meaning of attribute weights, rule weights, decay rates, and confidence remain unchanged, and the mean square error (MSE) between the output of the prediction model and the true output is minimized.

[0084] The embodiment of the present application provides a network fault prediction method based on an extended confidence rule base, which significantly improves the prediction accuracy and real-time performance by introducing a coordinated optimization strategy of attenuation factors and structural parameters. The method can comprehensively utilize quantitative data and qualitative knowledge, effectively simulate the changes of network status over time, reduce the impact of network failures, improve the stability and reliability of the network, and reduce maintenance costs.

[0085] Optionally, the network fault prediction method based on the extended confidence rule base provided in the embodiment of the present application, based on the network state at the past time, the corresponding reference level and reference value, preset the attenuation factor, and construct the extended confidence rule, specifically includes:

[0086] The predicted network state is y(t), and the network states at M past moments are y(t-τ), τ=1,…,M, M past moments are equivalent to M input attributes, and the preset attenuation factor is α l , then the extended confidence rule is constructed as

[0087] R l :If y(t-τ)isF(H n ),Then y(t)is(H n ,β n,l )

[0088] withattribute weightδ τ , rule weightθ l anddecayfactorα l

[0089] Among them, H n , n=1,2,…,N represents the reference level of y(t-τ), F(H n ) represents the corresponding reference value of the reference level, βn,l Indicates the predicted network state relative to H n Confidence level;

[0090] The attenuation factor is α l for

[0091]

[0092] Among them, λ l represents the decay rate, Δt l Indicates the decay time.

[0093] Specifically, for fault prediction, the model input is the past moment that lags behind the present, and the model output is the future state of the network. To express the above definition, assume that the future state of the network is y(t), and there are M past moments y(t-τ), τ=1,…,M. In BRB, M past moments are equivalent to M input attributes. At the same time, assume that the attenuation factor is α l , then the extended confidence rule R l can be constructed as

[0094] R l :If y(t-τ)isF(H n ),Then y(t)is(H n ,β n,l )

[0095] withattribute weightδ τ , rule weightθ l anddecayfactorα l

[0096] Among them, H n ,n=1,2,…,N and F(H n ) represent the reference level and corresponding reference value of y(t-τ) respectively. n,l Indicates the prediction result relative to level H n Since the exponential distribution can describe various decay processes, α l It can be further described as follows:

[0097]

[0098] Δt l Indicates the decay time. The decay factor is a time-related variable that reflects the decay of the rule credibility.

[0099] Optionally, the network fault prediction method based on the extended confidence rule base provided in the embodiment of the present application, wherein converting the state of the network at a past moment into its first matching degree relative to the corresponding reference level specifically includes:

[0100] Convert the numerical form of y(t-τ) into

[0101]

[0102] Among them, n,τ (t) represents y(t-τ) relative to H n The first match.

[0103] Specifically, since the input information y(t-τ) is in numerical form, the above information conversion method is usually used. n,τ (t) represents y(t-τ) relative to H n In the embodiment of the present application, only one linear transformation is provided as an example. For other types of input information, a suitable transformation method can be selected according to the specific situation.

[0104] Optionally, the network fault prediction method based on the extended confidence rule base provided in the embodiment of the present application, wherein the activation weight of each confidence rule is calculated according to the first matching degree, the reference level and the reference value of the confidence rule, specifically includes:

[0105] Activation weight w l (t) is expressed as

[0106]

[0107] Where L represents the total number of rules, θ l represents the rule weight, δ τ Represents the attribute weight.

[0108] Specifically, the activation weight w l (t) reflects the confidence rule R l The activation degree in the fusion process can be expressed by the above formula. Where L represents the total number of rules, θ l and δ τ Denote the rule weight and attribute weight respectively. As shown in the above formula, the decay factor is reflected in the calculation of the activation weight to discount the initial rule.

[0109] Optionally, the network fault prediction method based on the extended confidence rule base provided in the embodiment of the present application, wherein for each activated confidence rule, the basic probability mass assigned to different fault levels is calculated, specifically including:

[0110] If a rule is activated, its basic probability mass is calculated as

[0111]

[0112] Among them, m n,l (t) indicates the value assigned to level H n The basic probability mass, m P(Θ),l (t) represents the basic probability mass assigned to the power set P(Θ), which is given by 2 N A complete set of levels.

[0113] Specifically, Basic Probability Mass (BPM) is a core concept in the confidence rule base used to represent the confidence of rules for different results. In BRB, each rule is associated with one or more output levels, and BPM is the probability value assigned to these levels, which reflects the confidence of the rule for each possible result.

[0114] BPM is the output generated after the rule is activated, which quantifies the support of the rule for a specific output level. Specifically, when a rule is activated, it will generate a BPM value based on the input information and the confidence of the rule itself. This value indicates the trust level of the rule for a specific output level (such as high, medium, or low level of network failure). The BPM values ​​of all activated rules are then aggregated to produce the final prediction result.

[0115] The calculation of BPM involves the reference level and reference value of the confidence rule, as well as the matching degree of the input information with these reference values. In the embodiment provided in the present application, the calculation of BPM also takes into account the decay factor, which simulates the decay of the credibility of the information over time, so that BPM can reflect the decay of the credibility of the rule. In this way, BPM not only includes the confidence of the rule for a specific output level, but also includes the influence of the time factor, thereby improving the accuracy of the prediction model.

[0116] Optionally, the network fault prediction method based on the extended confidence rule base provided in the embodiment of the present application, wherein the basic probability masses of multiple confidence rules are iteratively integrated through the orthogonal and fusion mechanism, specifically includes:

[0117] Through the orthogonal and fusion mechanism, the first L1 of the L rules are fused into

[0118]

[0119] in, Indicates that the L1 fused rules are assigned to H n The normalized basic probability mass of represents the normalized basic probability mass assigned to P(Θ) by the L1 fused rules, Indicates that the L1 fused rules are assigned to H n The unnormalized basic probability mass of represents the unnormalized basic probability mass assigned to P(Θ) by the L1 fused rules, and Expressed as

[0120]

[0121] Optionally, the network fault prediction method based on the extended confidence rule base provided in the embodiment of the present application, wherein the prediction result of the network fault is obtained according to the basic probability mass after iterative fusion, specifically includes:

[0122] Iterative Calculation Until all rules are involved in the fusion, the final fusion result is expressed as

[0123]

[0124] Among them, β n,R(L) (t) indicates the fusion result relative to H n confidence level;

[0125] Level H n The reference value is F(H n ), then the prediction result U R(L) (t) is expressed as follows:

[0126]

[0127] Specifically,

[0128] First, the basic probability mass of each rule is iteratively calculated, including the calculation of and in represents the basic probability mass of the nth rule at the reference level, and Represents the basic probability mass of the nth rule at the reference level under all possible input conditions. The iterative process continues until all rules are included in the fusion calculation.

[0129] The final result of the fusion process is calculated by the following formula

[0130]

[0131] Among them, β n,R(L) (t) indicates the fusion result relative to the nth level F(H) at time t. n ) confidence level. is the estimate of the basic probability mass of the nth rule at time t, and the denominator is the sum of the estimates of the basic probability mass of all rules at time t. This formula ensures that the sum of the confidences of all rules is 1, thus correctly representing the relative likelihood of each fault level.

[0132] Calculate the prediction result U R(L) (t), which represents the predicted value of the network fault state at time t. This predicted value is obtained by dividing the reference value F(H n ) and its corresponding confidence β n,R(L) (t) and then summing over all levels to get:

[0133]

[0134] Among them, F(H n ) is the reference value of the nth fault level, which is a predefined value used to quantify the severity or impact of the fault level. In this way, the prediction result U R(L) (t) It combines the severity of all fault levels and their likelihood of occurrence to provide a quantitative prediction of the severity of network faults.

[0135] The advantage of this approach is that it can take into account the dynamic changes of network status and, through an iterative fusion process, can more accurately predict the occurrence of network failures. In this way, preventive measures can be taken in advance to reduce the impact of network failures on services.

[0136] Optionally, the network fault prediction method based on the extended confidence rule base provided in the embodiment of the present application, wherein according to the size of the rule weight, rules with weights lower than the statistical utility are removed and the number of rules is simplified, specifically includes:

[0137] The number of rules is simplified as shown below:

[0138]

[0139] in, is the simplified rule R l The weight, μ l For rule R l The statistical utility of is obtained by the following formula:

[0140]

[0141] Among them, μ T is the threshold of statistical utility;

[0142] After simplification, the calculation method of activation weights is modified as follows:

[0143]

[0144] in, represents the reduced activation weight, Indicates the number of rules after simplification.

[0145] Specifically, in the EBRB-based fault prediction model, the rule weight reflects the relative importance of the rules. When the rule weight is too small, its impact on the prediction result can be ignored. In addition, the number of rules directly affects the modeling complexity. Therefore, it is necessary to simplify the number of rules, as shown in the following formula:

[0146]

[0147] in, is the simplified rule R l The weight of μ l For rule R l The statistical utility of can be obtained by the following formula:

[0148]

[0149] μ T is the threshold of statistical utility, and its value can be determined according to the actual situation. After simplification, the number of rules changes, and the calculation method of activation weight needs to be modified accordingly, as shown below:

[0150]

[0151] in, and are the activation weights and number of rules after reduction, respectively.

[0152] Optionally, in the network fault prediction method based on the extended confidence rule base provided in the embodiment of the present application, the conditions that need to be met in the preset parameter optimization process are solved by an intelligent optimization algorithm to achieve collaborative optimization of structural parameters, specifically including:

[0153] The actual output is The output of the prediction model U R(L) (t) The mean square error MSE between

[0154]

[0155] Where T represents the number of training data;

[0156] The optimization objective is expressed as follows

[0157] minMSE

[0158] st.

[0159] 0<δ τ <1; 0<θ l <1;

[0160] λ l ≥0;

[0161] 0≤β n,l ≤1;

[0162] in, represents the true result, γ is the adjustment factor, which is used to convert the prediction error and the number of rules to the same order of magnitude;

[0163] The collaborative optimization of structural parameters is achieved through intelligent optimization algorithm.

[0164] Specifically, after completing the rule reduction, the model parameters need to be updated using monitoring information. To ensure the interpretability of the prediction results, the following conditions need to be met during the parameter optimization process:

[0165] 1) Attribute weight δ τ It reflects the relative importance of input attributes, and its constraints are as follows:

[0166] 0<δ τ <1

[0167] 2) Rule weight θ l It reflects the relative importance of the rules, and its constraints are as follows:

[0168] 0<θ l <1

[0169] 3) Decay rate λ l Should be a non-negative value, and its constraints are as follows:

[0170] λ l ≥0

[0171] 4) Confidence β n,l It reflects the probability of an event, and its constraints are as follows:

[0172]

[0173] Assume that the real output is Then the output U of the prediction model is R(L) (t) The mean square error (MSE) between them can be calculated as:

[0174]

[0175] Where T represents the number of training data. In summary, the optimization objective can be expressed as follows:

[0176] minMSE

[0177] st.

[0178] 0<δ τ <1; 0<θ l <1;

[0179] λ l ≥0;

[0180] 0≤β n,l ≤1;

[0181] in, represents the real result. γ is the adjustment factor, which is used to convert the prediction error and the number of rules to the same order of magnitude. The optimization target can be solved by intelligent optimization algorithm.

[0182] In a specific embodiment, throughput (TP) is selected as a specific indicator reflecting the network fault status. Through the network data acquisition equipment, 100 sets of TP data are obtained, such as Figure 2 As shown. Combined with expert knowledge, the network fault degree is described as low, medium and high, and the corresponding reference values ​​are 3.8, 2.7 and 1.6 respectively. In addition, three moments y(t-1), y(t-2) and y(t-3) are selected as input attributes to construct a fault prediction model based on EBRB. It can be seen that the prediction model has a total of 9 confidence rules, and its initial parameters are shown in Table 1.

[0183] Table 1 Initial model parameters

[0184]

[0185] In order to improve the prediction accuracy, it is necessary to optimize the model structure and parameters according to the monitoring data. In the optimization process, all odd-numbered data constitute the training set, and the remaining data are used as the test set, and the threshold μ of the statistical utility is set as T The optimized model parameters and prediction results are shown in Table 2 and Figure 3 As shown in Table 2, two confidence rules are simplified in the optimization process, thereby reducing the complexity of the model. Figure 3 It can be seen that the MSE distribution of the prediction model before and after optimization is 0.0518 and 0.0850, which shows that the optimized model can accurately track the network fault status.

[0186] Table 2 Optimized model parameters

[0187]

[0188]

[0189] The embodiment of the present application also provides a network fault prediction device based on an extended confidence rule base, comprising:

[0190] A construction module, for presetting a decay factor based on the state of the network at a past moment, the corresponding reference level and the reference value, constructing an extended confidence rule, and converting the state of the network at a past moment into a first matching degree relative to the corresponding reference level;

[0191] an acquisition module, configured to calculate the activation weight of each confidence rule according to the first matching degree, the reference level and the reference value of the confidence rule, and for each activated confidence rule, calculate the basic probability mass assigned to different fault levels;

[0192] The prediction module is used to iteratively fuse the basic probability masses of multiple confidence rules through orthogonal and fusion mechanisms, and obtain the prediction result of network faults according to the basic probability masses after iterative fusion;

[0193] The optimization module is used to eliminate rules whose weights are lower than the statistical utility according to the size of the rule weights, simplify the number of rules, establish the optimization goals of the structural parameters, preset the conditions that need to be met during the parameter optimization process, and solve them through the intelligent optimization algorithm to achieve collaborative optimization of the structural parameters.

[0194] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0195] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0196] In the several embodiments provided in the present application, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of the units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some service interfaces, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0197] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0198] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0199] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable memory, which can include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0200] The above is only an exemplary embodiment of the present disclosure, and the scope of the present disclosure cannot be limited thereto. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the specification and practicing the disclosure here, those skilled in the art will easily think of the implementation scheme of the present disclosure. This application is intended to cover any modification, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the technical field not recorded in the present disclosure. The description and examples are regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

[0201] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0202] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A network fault prediction method based on an extended confidence rule base, characterized in that: The following steps are involved: Based on the state of the network at the past moment, the corresponding reference level and the reference value, a decay factor is preset, and an extended confidence rule is constructed to convert the state of the network at the past moment into its first matching degree relative to the corresponding reference level; Calculate the activation weight of each confidence rule according to the first matching degree, the reference level and the reference value of the confidence rule, and for each activated confidence rule, calculate the basic probability mass assigned to different fault levels; Through the orthogonal and fusion mechanism, the basic probability masses of multiple confidence rules are iteratively integrated, and the prediction results of network faults are obtained according to the basic probability masses after iterative integration; According to the size of the rule weight, the rules with weights lower than the preset statistical utility are cleared. Based on the conditions that need to be met in the preset parameter optimization process, the intelligent optimization algorithm is used to solve the problem and realize the collaborative optimization of the structural parameters.

2. The network fault prediction method based on the extended confidence rule base according to claim 1, characterized in that: The method of presetting the attenuation factor based on the network status at the past time, the corresponding reference level and reference value, and constructing the extended confidence rule specifically includes: The predicted network state is y(t), and the network states at M past moments are y(t-τ), τ=1,…,M, M past moments are equivalent to M input attributes, and the preset attenuation factor is α l , then the extended confidence rule is constructed as R l :If y(t-τ)isF(H n ),Then y(t)is(H n ,β n,l ) withattribute weightδ τ ,ruleweightθ l anddecayfactorα l Among them, H n , n=1,2,…,N represents the reference level of y(t-τ), F(H n ) represents the corresponding reference value of the reference level, β n,l Indicates the predicted network state relative to H n Confidence level; The attenuation factor is α l for Among them, λ l represents the decay rate, Δt l Indicates the decay time.

3. The network fault prediction method based on the extended confidence rule base according to claim 2, characterized in that: The converting the state of the network at a past moment into its first matching degree relative to the corresponding reference level specifically includes: Convert the numerical form of y(t-τ) into Among them, n,τ (t) represents y(t-τ) relative to H n The first match.

4. The network fault prediction method based on the extended confidence rule base according to claim 3, characterized in that: The step of calculating the activation weight of each confidence rule according to the first matching degree, the reference level and the reference value of the confidence rule specifically includes: Activation weight w l (t) is expressed as Where L represents the total number of rules, θ l represents the rule weight, δ τ Represents the attribute weight.

5. The network fault prediction method based on the extended confidence rule base according to claim 4, characterized in that: For each activated confidence rule, the basic probability mass assigned to different fault levels is calculated, specifically including: If a rule is activated, its basic probability mass is calculated as Among them, m n,l (t) indicates the value assigned to level H n The basic probability mass, m P(Θ),l (t) represents the basic probability mass assigned to the power set P(Θ), which is given by 2 N A complete set of levels.

6. The network fault prediction method based on the extended confidence rule base according to claim 5, characterized in that: The basic probability masses of multiple confidence rules are iteratively integrated through the orthogonal and fusion mechanisms, specifically including: Through the orthogonal and fusion mechanism, the first L1 of the L rules are fused into in, Indicates that the L1 fused rules are assigned to H n The normalized basic probability mass of represents the normalized basic probability mass assigned to P(Θ) by the L1 fused rules, Indicates that the L1 fused rules are assigned to H n The unnormalized basic probability mass of represents the unnormalized basic probability mass assigned to P(Θ) by the L1 fused rules, and Expressed as 7. The network fault prediction method based on the extended confidence rule base according to claim 6, characterized in that: The method of obtaining the prediction result of the network failure according to the basic probability mass after iterative fusion specifically includes: Iterative Calculation Until all rules are involved in the fusion, the final fusion result is expressed as Among them, β n,R(L) (t) indicates the fusion result relative to H n confidence level; Level H n The reference value is F(H n ), then the prediction result U R(L) (t) is expressed as follows:

8. The network fault prediction method based on the extended confidence rule base according to claim 7, characterized in that: According to the size of the rule weights, the rules with weights lower than the statistical utility are removed, and the number of rules is simplified, specifically including: The number of rules is simplified as shown below: in, is the simplified rule R l The weight, μ l For rule R l The statistical utility of is obtained by the following formula: Among them, μ T is the threshold of statistical utility; After simplification, the calculation method of activation weights is modified as follows: in, represents the reduced activation weight, Indicates the number of rules after simplification.

9. The network fault prediction method based on the extended confidence rule base according to claim 8, characterized in that: The conditions that need to be met during the preset parameter optimization process are solved by an intelligent optimization algorithm to achieve collaborative optimization of structural parameters, specifically including: The actual output is The output of the prediction model U R(L) (t) The mean square error MSE between Where T represents the number of training data; The optimization objective is expressed as follows minMSE st. 0<δ τ <1;0<θ l <1; l l ≥0; 0≤β n,l ≤1； in, represents the true result, γ is the adjustment factor, which is used to convert the prediction error and the number of rules to the same order of magnitude; The collaborative optimization of structural parameters is achieved through intelligent optimization algorithm.

10. A network fault prediction device based on an extended confidence rule base, characterized in that: include: A construction module, for presetting a decay factor based on the state of the network at a past moment, the corresponding reference level and the reference value, and constructing an extended confidence rule to convert the state of the network at a past moment into a first matching degree relative to the corresponding reference level; an acquisition module, configured to calculate the activation weight of each confidence rule according to the first matching degree, the reference level and the reference value of the confidence rule, and for each activated confidence rule, calculate the basic probability mass assigned to different fault levels; The prediction module is used to iteratively fuse the basic probability masses of multiple confidence rules through orthogonal and fusion mechanisms, and obtain the prediction result of network faults according to the basic probability masses after iterative fusion; The optimization module is used to eliminate rules whose weights are lower than the statistical utility according to the size of the rule weights, simplify the number of rules, establish the optimization goals of the structural parameters, preset the conditions that need to be met during the parameter optimization process, and solve them through the intelligent optimization algorithm to achieve collaborative optimization of the structural parameters.

Images