Link risk assessment method, device and equipment and computer readable storage medium

Abstract

The invention discloses a link risk assessment method, device and equipment and a computer readable storage medium. The method comprises the following steps: acquiring corresponding multi-dimensional feature values of a link to be predicted N days before a prediction time point, wherein the multi-dimensional feature values comprise feature values of at least two dimensions; determining a new feature value corresponding to each dimension according to a target subinterval corresponding to each feature value on the dimension to which the feature value belongs; inputting the new feature value corresponding to each dimension into a prediction model corresponding to each dimension to obtain a prediction value corresponding to the Mth day after the prediction time point of each dimension; and integrating the predicted values corresponding to all dimensions to perform weighted summation to obtain a link risk assessment value corresponding to the to-be-predicted link. According to the link risk prediction method and device, the future multi-dimensional feature value is predicted based on the known multi-dimensional feature value of the link, and the link risk prediction is performed in combination with the predicted multi-dimensional feature value, so that the manual operation and maintenance workload is reduced, and the link risk hidden danger is identified in advance.

Description

Technical Field

[0001] The present application relates to the field of network operation and maintenance technology, and specifically to a link risk assessment method, apparatus, device, and computer-readable storage medium. Background Art

[0002] In the field of network operation and maintenance technology, link risk analysis currently relies on manual analysis. A common method is to assess link risk based on the link's optical power data. However, this method can only assess the current link status and cannot predict the link's future risk status. Summary of the Invention

[0003] The present application provides a link risk assessment method, apparatus, device and computer-readable storage medium, which can solve the technical problem in the prior art of being unable to predict link risk conditions.

[0004] In a first aspect, an embodiment of the present application provides a link risk assessment method, the link risk assessment method comprising:

[0005] Obtaining a multidimensional feature value corresponding to the link to be predicted N days before the prediction time point, wherein the multidimensional feature value includes feature values of at least two dimensions;

[0006] Determine the new eigenvalue corresponding to each dimension based on the target subinterval corresponding to each eigenvalue in the dimension to which it belongs;

[0007] Input the new feature values corresponding to each dimension into the prediction model corresponding to each dimension, and obtain the predicted value of each dimension M days after the prediction time point;

[0008] The predicted values corresponding to all dimensions are weighted and summed to obtain the link risk assessment value corresponding to the link to be predicted.

[0009] In conjunction with the first aspect, in one embodiment, before obtaining the multidimensional feature value corresponding to the link to be predicted N days before the prediction time point, the method further includes:

[0010] Obtain the multi-dimensional historical feature values corresponding to S sample links N days before the training time point;

[0011] For each dimension, determine the maximum historical eigenvalue, and divide the interval corresponding to zero to the maximum historical eigenvalue into several sub-intervals based on several quantiles, and set the value corresponding to each sub-interval, where different sub-intervals correspond to different values, and the maximum value is 1.

[0012] In conjunction with the first aspect, in one embodiment, determining the new eigenvalue corresponding to each dimension according to the target subinterval corresponding to each eigenvalue in the dimension to which it belongs includes:

[0013] For each eigenvalue, determine its target subinterval among the subintervals corresponding to the dimension to which it belongs;

[0014] The value corresponding to the target subinterval is used as the new eigenvalue corresponding to the dimension to which the eigenvalue belongs.

[0015] In combination with the first aspect, in one embodiment, after setting the value corresponding to each sub-interval, the method further includes:

[0016] Obtain the multi-dimensional true feature values corresponding to S sample links M days after the training time point;

[0017] For each historical feature value, determine its target subinterval among the subintervals corresponding to the dimension to which it belongs, and update the historical feature value with the value corresponding to the target subinterval;

[0018] For each dimension, the new historical eigenvalues and real eigenvalues corresponding to S sample links are used as training samples to train the preset neural network model and obtain a prediction model;

[0019] And so on, we get the prediction model corresponding to each dimension.

[0020] In conjunction with the first aspect, in one embodiment, after obtaining the prediction model corresponding to each dimension, the method further includes:

[0021] Obtain the link status of S sample links M days after the training time point;

[0022] According to the link status of each sample link, determine the actual link risk value of each sample link;

[0023] Input the new historical feature value of each sample link in each dimension into the prediction model corresponding to each dimension to obtain the sample prediction value of each sample link in each dimension;

[0024] The actual link risk values of the S sample links and the sample predicted values of the S sample links in each dimension are combined to obtain the weight corresponding to each dimension.

[0025] In combination with the first aspect, in one embodiment, the S sample links include S1 core layer links, S2 aggregation layer links, and S3 access layer links, where S1, S2, and S3 are all preset values.

[0026] In conjunction with the first aspect, in one embodiment, after obtaining the link risk assessment value corresponding to the link to be predicted, the method further includes:

[0027] Outputting an operation and maintenance prompt based on the link risk assessment value.

[0028] In a second aspect, an embodiment of the present application provides a link risk assessment device, the link risk assessment device comprising:

[0029] An acquisition module, configured to acquire multidimensional feature values corresponding to the link to be predicted N days before the prediction time point, wherein the multidimensional feature values include feature values of at least two dimensions;

[0030] A normalization processing module is used to determine a new eigenvalue corresponding to each dimension based on the target subinterval corresponding to each eigenvalue in the dimension to which it belongs;

[0031] The prediction module is used to input the new feature values corresponding to each dimension into the prediction model corresponding to each dimension, and obtain the predicted value corresponding to each dimension M days after the prediction time point;

[0032] The evaluation module is used to perform weighted summation on the prediction values corresponding to all dimensions to obtain a link risk evaluation value corresponding to the link to be predicted.

[0033] In a third aspect, an embodiment of the present application provides a link risk assessment device, which includes a processor, a memory, and a link risk assessment program stored on the memory and executable by the processor, wherein when the link risk assessment program is executed by the processor, the steps of the link risk assessment method described in the first aspect are implemented.

[0034] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which a link risk assessment program is stored, wherein when the link risk assessment program is executed by a processor, the steps of the link risk assessment method described in the first aspect are implemented.

[0035] The beneficial effects of the technical solutions provided in the embodiments of the present application include:

[0036] In an embodiment of the present application, the multidimensional characteristic values corresponding to the link to be predicted N days before the prediction time point are obtained, and the multidimensional characteristic values include characteristic values of at least two dimensions; according to the target sub-interval corresponding to each characteristic value in the dimension to which it belongs, the new characteristic value corresponding to each dimension is determined; the new characteristic value corresponding to each dimension is input into the prediction model corresponding to each dimension respectively, and the prediction value corresponding to each dimension M days after the prediction time point is obtained; the prediction values corresponding to all dimensions are weighted and summed to obtain the link risk assessment value corresponding to the link to be predicted. Through the embodiment of the present application, the future multidimensional characteristic values are predicted based on the known multidimensional characteristic values of the link, and the link risk prediction is performed in combination with the predicted multidimensional characteristic values, which not only reduces the workload of manual operation and maintenance, but also realizes the early identification of link risk hazards. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1This is a flow chart of an embodiment of a link risk assessment method of the present application;

[0038] Figure 2 This is a schematic diagram of the functional modules of an embodiment of a link risk assessment device of the present application;

[0039] Figure 3 This is a schematic diagram of the hardware structure of the link risk assessment device involved in the embodiment of the present application. DETAILED DESCRIPTION

[0040] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0041] In order to make the objectives, technical solutions and advantages of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0042] In a first aspect, an embodiment of the present application provides a link risk assessment method.

[0043] In one embodiment, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the link risk assessment method of this application. Figure 1 As shown, the link risk assessment method includes:

[0044] Step S10: obtaining a multi-dimensional feature value corresponding to the link to be predicted N days before the prediction time point, wherein the multi-dimensional feature value includes feature values of at least two dimensions;

[0045] In this embodiment, the multi-dimensional fault cumulative duration, fault occurrence number, bandwidth utilization, and input optical power are used as examples, and N days is set according to actual needs, such as 180 days. For example, at time t1 (i.e., the prediction time point), the risk assessment of the link to be predicted is started, and the cumulative fault duration A of the link to be predicted 180 days before time t1 is obtained. history 、Number of failures B history , bandwidth utilization C history And the input optical power D history Among them, the cumulative fault duration and the number of fault occurrences can be obtained based on the collected alarm data; the bandwidth utilization can be obtained based on the traffic tool; and the optical power data can be obtained based on the collected input optical power performance data.

[0046] Step S20, determining a new eigenvalue corresponding to each dimension according to the target subinterval corresponding to each eigenvalue in the dimension to which it belongs;

[0047] In this embodiment, the prediction model is used to obtain the predicted value of each dimension M days after the prediction time point. Based on this, in order to improve the accuracy of machine learning, considering that the characteristic values of multiple dimensions that affect the link risk value (including but not limited to the cumulative fault duration, the number of fault occurrences, bandwidth utilization, optical power, etc.) have different value ranges, for example, the cumulative fault duration has a value range of [0, 100+), and the bandwidth utilization has a value range of [0, 1). In order to make the data converge, before the characteristic values are input into the prediction model, the characteristic values of multiple dimensions of the link to be predicted are normalized so that they have the same measurement scale. This can avoid the instability caused by outliers to the prediction model.

[0048] The actual distribution of eigenvalues in different dimensions is difficult to integrate using conventional normalization algorithms. Based on this, the present application proposes a new normalization processing scheme, specifically:

[0049] For each dimension, multiple subintervals are divided and values are assigned to each subinterval. Based on this, when normalizing the eigenvalues of each dimension, the target subinterval of the eigenvalue is first determined, and then the value assigned to the target subinterval is used as the new eigenvalue corresponding to the dimension.

[0050] Step S30: Input the new feature values corresponding to each dimension into the prediction model corresponding to each dimension to obtain the predicted value of each dimension M days after the prediction time point;

[0051] In this embodiment, the prediction model is obtained by training a preset model, and the preset model can be selected based on actual needs, such as an LSTM model, a Prophet model, a Transformer model, etc.

[0052] Taking the LSTM model as an example, the LSTM (Long Short-Term Memory) is a time-recurrent neural network, a special type of recursive neural network, suitable for the prediction and analysis of multi-dimensional time series. The LSTM network structure contains forget gates, update gates, and output gate layers. This allows both more distant and more recent historical data in the time series to have a certain impact on the present and future, forming long-term and short-term memories. This results in the prediction value being determined by the combined effects of recent and distant history.

[0053] In this implementation, the feature values of D dimensions (including but not limited to the cumulative fault duration, number of fault occurrences, bandwidth utilization, optical power, etc.) of the link to be predicted over the past N days are used to predict the feature values of these D dimensions M days in the future through a trained LSTM model (i.e., a prediction model).

[0054] The LSTM model function is as follows:

[0055] Y ij =LSTM i (X ij )

[0056] Among them, input X ij is the new feature value of the j-th link for the i-th dimension, LSTM i is the LSTM model corresponding to the i-th dimension, Y ij Represents the predicted value of the i-th dimension of the j-th link in the next M days.

[0057] Assuming that the number of training sample links is P, the time step is N, and the feature dimension is D, then the value range of i in the above formula is [0, D-1], and the value range of j is [0, P-1], which is used to train LSTM i The data is a P×M array, that is, the eigenvalues of the i-th dimension of the P links over N days of history.

[0058] Output Y i : The predicted value of the i-th dimension of P sample links after M days in the future. It is a P×1 array.

[0059] After training a multi-dimensional LSTM model based on existing sample data, it is applied to the actual live network to predict future multi-dimensional feature values based on historical multi-dimensional feature values. For example, based on N days (assuming 3 months) of historical bandwidth utilization data on a link, bandwidth utilization data for the next M days (assuming 7 days) can be predicted.

[0060] Step S40 , performing weighted summation on the prediction values corresponding to all dimensions to obtain a link risk assessment value corresponding to the link to be predicted.

[0061] In this embodiment, combined with the above description, A history 、B history 、C history and D history After inputting the corresponding prediction models respectively, the prediction value corresponding to each dimension can be obtained, which is recorded as A predict 、B predict 、C predict and D predict .

[0062] Combine the weight pair A corresponding to each dimensionpredict 、B predict 、C predict and D predict By performing weighted summation, we can obtain the link risk assessment value corresponding to the link to be predicted. Among them, the weight corresponding to the cumulative fault duration is denoted as ω a The weight corresponding to the number of fault occurrences is recorded as ω b The weight corresponding to the bandwidth utilization is recorded as ω c The weight corresponding to the input optical power is denoted as ω d , where the size of each weight can be set according to the correlation between the corresponding dimension and the link risk, that is, the higher the correlation between the dimension and the link risk, the larger the corresponding weight.

[0063] In an embodiment of the present application, the multidimensional characteristic values corresponding to the link to be predicted N days before the prediction time point are obtained, and the multidimensional characteristic values include characteristic values of at least two dimensions; according to the target sub-interval corresponding to each characteristic value in the dimension to which it belongs, the new characteristic value corresponding to each dimension is determined; the new characteristic value corresponding to each dimension is input into the prediction model corresponding to each dimension respectively, and the prediction value corresponding to each dimension M days after the prediction time point is obtained; the prediction values corresponding to all dimensions are weighted and summed to obtain the link risk assessment value corresponding to the link to be predicted. Through the embodiment of the present application, the future multidimensional characteristic values are predicted based on the known multidimensional characteristic values of the link, and the link risk prediction is performed in combination with the predicted multidimensional characteristic values, which not only reduces the workload of manual operation and maintenance, but also realizes the early identification of link risk hazards.

[0064] Furthermore, in one embodiment, before step S10, the method further includes:

[0065] Step S00: Obtain multi-dimensional historical feature values corresponding to S sample links N days before the training time point;

[0066] Step S01: for each dimension, determine the maximum historical eigenvalue, and divide the interval corresponding to zero to the maximum historical eigenvalue into several sub-intervals based on several quantiles, and set the value corresponding to each sub-interval, where different sub-intervals correspond to different values, and the maximum value is 1.

[0067] In this embodiment, it is assumed that 3000 sample links are randomly selected, and the cumulative fault duration, number of fault occurrences, bandwidth utilization, and input optical power of the 3000 sample links N days before the training time point are obtained.

[0068] Taking the cumulative duration of a fault as an example, we can obtain the historical characteristic values of 3000 cumulative fault durations. Assuming that the maximum historical characteristic value is 100, and the quantiles are 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90%, and 100%, the subintervals are:

[0069] Subinterval A1: [0, 100*10%);

[0070] Subinterval A2: [100*10%, 100*20%);

[0071] Subinterval A3: [100*20%, 100*30%);

[0072] Subinterval A4: [100*30%, 100*40%);

[0073] Subinterval A5: [100*40%, 100*50%);

[0074] Subinterval A6: [100*50%, 100*60%);

[0075] Subinterval A7: [100*60%, 100*70%);

[0076] Subinterval A8: [100*70%, 100*80%);

[0077] Subinterval A9: [100*80%, 100*90%);

[0078] Subinterval A10: [100*90%, 100*100%].

[0079] Set the value corresponding to sub-interval A1 to 0.1, the value corresponding to sub-interval A2 to 0.2, the value corresponding to sub-interval A3 to 0.3, the value corresponding to sub-interval A4 to 0.4, the value corresponding to sub-interval A5 to 0.5, the value corresponding to sub-interval A6 to 0.6, the value corresponding to sub-interval A7 to 0.7, the value corresponding to sub-interval A8 to 0.8, the value corresponding to sub-interval A9 to 0.9, and the value corresponding to sub-interval A10 to 0.10.

[0080] Similarly, the above processing is performed for each dimension to obtain the sub-interval division result corresponding to each dimension and the value of each sub-interval.

[0081] It should be noted that the selection of quantiles and the setting of the corresponding values of each sub-interval are not limited to the above description and depend on actual needs.

[0082] Furthermore, in one embodiment, step S20 includes:

[0083] For each eigenvalue, determine its target subinterval among several subintervals corresponding to the dimension to which it belongs; and use the value corresponding to the target subinterval as the new eigenvalue corresponding to the dimension to which the eigenvalue belongs.

[0084] In this embodiment, referring to the above description, taking the characteristic value corresponding to the cumulative duration of the fault as an example, assuming that the characteristic value is 25, then its target subinterval among the several subintervals corresponding to the cumulative duration of the fault is subinterval A3, and 0.3 is used as the new characteristic value corresponding to the cumulative duration of the fault.

[0085] Furthermore, in one embodiment, the S sample links include S1 core layer links, S2 aggregation layer links, and S3 access layer links, wherein S1, S2, and S3 are all preset values.

[0086] In this embodiment, by diversifying samples, the model can capture more comprehensive data patterns, thereby enhancing the generalization ability of the prediction results.

[0087] Furthermore, in one embodiment, after step S01, the method further includes:

[0088] Step S02, obtaining the multi-dimensional true feature values corresponding to the S sample links M days after the training time point;

[0089] Step S03: for each historical feature value, determine the target subinterval in which it is located among the subintervals corresponding to the dimension to which it belongs, and update the historical feature value with the value corresponding to the target subinterval;

[0090] Step S04: For each dimension, the new historical feature values and the real feature values corresponding to the S sample links are used as training samples to train the preset neural network model to obtain a prediction model;

[0091] And so on, we get the prediction model corresponding to each dimension.

[0092] In this embodiment, taking the LSTM model corresponding to the training bandwidth utilization as an example, the training process is as follows:

[0093] 3000 sample links are selected from the existing network. Based on the topological layering of the PTN / SPN network (core layer / aggregation layer / access layer), 1000 core layer links, 1000 aggregation layer links, and 1000 access layer links are selected. The historical bandwidth utilization feature values of these 3000 sample links N days before the training time point (assuming 180 days) are obtained to form a two-dimensional array A

[3000]

[180] , that is, X = A

[3000]

[180] , where each element in X is a[i][history], specifically the historical bandwidth utilization feature value of the i-th link N days before the training time point.

[0094] Get the true characteristic value y of the bandwidth utilization of these 3000 sample links on the Mth day after the training time point (assuming the 7th day) i , forming an array of 3000 rows and 1 column, that is, Y=[y0 y1 … y2999 ] T As a label.

[0095] For each element a[i][history] in X, the value of the element a[i][history] is updated according to its target subinterval in the subintervals corresponding to the bandwidth utilization, thereby obtaining a new two-dimensional array, namely X'.

[0096] Using X' as the input of the preset neural network model and Y as the label, combined with the neural network training process, we can obtain the LSTM model corresponding to bandwidth utilization.

[0097] Similarly, LSTM models corresponding to other dimensions can be trained separately, such as the LSTM model corresponding to the cumulative fault duration, the LSTM model corresponding to the number of fault occurrences, and the LSTM model corresponding to optical power.

[0098] By putting the trained LSMT model corresponding to each dimension into use, the predicted values corresponding to each dimension can be predicted respectively.

[0099] Furthermore, in one embodiment, after obtaining the prediction model corresponding to each dimension, the method further includes:

[0100] Step S05, obtaining the link status of S sample links M days after the training time point;

[0101] Step S06: determining the actual link risk value of each sample link according to the link status of each sample link;

[0102] Step S07: Input the new historical feature value of each sample link in each dimension into the prediction model corresponding to each dimension to obtain the sample prediction value of each sample link in each dimension;

[0103] Step S08 : The actual link risk values of the S sample links and the sample predicted values of the S sample links in each dimension are integrated to obtain the weight corresponding to each dimension.

[0104] In this embodiment, for each sample link, the new historical feature value in each dimension is input into the prediction model corresponding to each dimension, and the sample prediction value in each dimension can be obtained. For example, the sample prediction value of the sample link i in the four dimensions of cumulative fault duration, number of fault occurrences, bandwidth utilization, and optical power is recorded as A i-1_predict 、B i-1_predict 、C i-1_predict and D i-1_predict , for 3000 sample links, we have:

[0105]

[0106] The actual link risk value is determined based on the link status of these 3000 sample links M days after the training time point. The judgment rule is:

[0107] a) If the link is interrupted, the true value of the link risk is 1;

[0108] b) The link is not interrupted, but there are serious alarms of daily concern on the link, and the actual link risk value is 0.75;

[0109] c) The link is not interrupted and there are no serious alarms of daily concern on the link, but the traffic on the link exceeds 75% or the optical power performance has degraded. The actual link risk value is 0.25;

[0110] d) The link is normal, there are no alarms, and the actual link risk value is 0.

[0111] Based on this, we have:

[0112]

[0113] Among them, S i-1_true represents the true link risk value of sample link i on the Mth day after the training time point. Furthermore, for each sample link i, the calculation formula for its true link risk value is as follows:

[0114] S i-1_true =A i-1_predict ×ω a +B o-1_predict ×ω b +C i-1_predict ×ω c +D i-1_predict ×ω d

[0115] The least squares formula can be used to calculate the weight corresponding to each dimension:

[0116] ω=(X T X) -1 X T Y

[0117] in,

[0118] Furthermore, in one embodiment, after step S40, the method further includes:

[0119] Step S50: outputting an operation and maintenance prompt based on the link risk assessment value.

[0120] In this embodiment, different numerical intervals can be set based on actual conditions, and operation and maintenance prompts corresponding to each numerical interval can be set. After obtaining the link risk assessment value, the target numerical interval in which the link risk assessment value is located is determined, and the operation and maintenance prompts corresponding to the target numerical interval are output.

[0121] In a second aspect, an embodiment of the present application also provides a link risk assessment device.

[0122] In one embodiment, referring to Figure 2 , Figure 2 This is a functional module diagram of an embodiment of the link risk assessment device of this application. Figure 2 As shown, the link risk assessment device includes:

[0123] An acquisition module 10 is configured to acquire a multi-dimensional feature value corresponding to the link to be predicted N days before the prediction time point, wherein the multi-dimensional feature value includes feature values of at least two dimensions;

[0124] A normalization processing module 20 is configured to determine a new eigenvalue corresponding to each dimension based on a target subinterval corresponding to each eigenvalue in the dimension to which it belongs;

[0125] Prediction module 30, used to input the new feature value corresponding to each dimension into the prediction model corresponding to each dimension, and obtain the predicted value corresponding to each dimension M days after the prediction time point;

[0126] The evaluation module 40 is configured to perform weighted summation on the prediction values corresponding to all dimensions to obtain a link risk evaluation value corresponding to the link to be predicted.

[0127] Furthermore, in one embodiment, the link risk assessment apparatus further includes a preparation module, configured to:

[0128] Obtain the multi-dimensional historical feature values corresponding to S sample links N days before the training time point;

[0129] For each dimension, determine the maximum historical eigenvalue, and divide the interval corresponding to zero to the maximum historical eigenvalue into several sub-intervals based on several quantiles, and set the value corresponding to each sub-interval, where different sub-intervals correspond to different values, and the maximum value is 1.

[0130] Furthermore, in one embodiment, the normalization processing module is configured to:

[0131] For each eigenvalue, determine its target subinterval among the subintervals corresponding to the dimension to which it belongs;

[0132] The value corresponding to the target subinterval is used as the new eigenvalue corresponding to the dimension to which the eigenvalue belongs.

[0133] Furthermore, in one embodiment, the preparation module is further configured to:

[0134] Obtain the multi-dimensional true feature values corresponding to S sample links M days after the training time point;

[0135] For each historical feature value, determine its target subinterval among the subintervals corresponding to the dimension to which it belongs, and update the historical feature value with the value corresponding to the target subinterval;

[0136] For each dimension, the new historical eigenvalues and real eigenvalues corresponding to S sample links are used as training samples to train the preset neural network model and obtain a prediction model;

[0137] And so on, we get the prediction model corresponding to each dimension.

[0138] Furthermore, in one embodiment, the preparation module is further configured to:

[0139] Obtain the link status of S sample links M days after the training time point;

[0140] According to the link status of each sample link, determine the actual link risk value of each sample link;

[0141] Input the new historical feature value of each sample link in each dimension into the prediction model corresponding to each dimension to obtain the sample prediction value of each sample link in each dimension;

[0142] The actual link risk values of the S sample links and the sample predicted values of the S sample links in each dimension are combined to obtain the weight corresponding to each dimension.

[0143] Furthermore, in one embodiment, the S sample links include S1 core layer links, S2 aggregation layer links, and S3 access layer links, wherein S1, S2, and S3 are all preset values.

[0144] Furthermore, in one embodiment, the link risk assessment device further includes a prompt module, which is configured to:

[0145] Outputting an operation and maintenance prompt based on the link risk assessment value.

[0146] Among them, the functional implementation of each module in the above-mentioned link risk assessment device corresponds to each step in the above-mentioned link risk assessment method embodiment, and its functions and implementation processes are no longer detailed here.

[0147] In a third aspect, an embodiment of the present application provides a link risk assessment device, which may be a personal computer (PC), a laptop computer, a server, or other device with data processing capabilities.

[0148] Reference Figure 3 , Figure 3 FIG2 is a schematic diagram of the hardware structure of the link risk assessment device involved in the embodiment of the present application. In the embodiment of the present application, the link risk assessment device may include a processor, a memory, a communication interface, and a communication bus.

[0149] The communication bus may be of any type and is used to interconnect the processor, memory, and communication interface.

[0150] Communication interfaces include input / output (I / O) interfaces, physical interfaces, and logical interfaces, which are used to interconnect components within the link risk assessment device, as well as interfaces used to interconnect the link risk assessment device with other devices (such as other computing devices or user devices). Physical interfaces can be Ethernet, fiber, or ATM interfaces; user devices can be displays or keyboards.

[0151] The memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical storage, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.

[0152] The processor can be a general-purpose processor that can call a link risk assessment program stored in a memory and execute the link risk assessment method provided in the embodiments of the present application. For example, the general-purpose processor can be a central processing unit (CPU). The method executed when the link risk assessment program is called can be referenced in the various embodiments of the link risk assessment method of the present application and will not be further described here.

[0153] Those skilled in the art will understand that Figure 3 The hardware structure shown in the figure does not constitute a limitation to the present application and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0154] In a fourth aspect, an embodiment of the present application also provides a computer-readable storage medium.

[0155] The computer-readable storage medium of the present application stores a link risk assessment program, wherein when the link risk assessment program is executed by a processor, the steps of the link risk assessment method described above are implemented.

[0156] Among them, the method implemented when the link risk assessment program is executed can refer to the various embodiments of the link risk assessment method of the present application, and will not be repeated here.

[0157] It should be noted that the serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0158] The terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned drawings 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. The terms "first", "second" and "third" are used to distinguish different objects, etc., and do not represent a sequence, nor do they limit the "first", "second" and "third" to different types.

[0159] In the description of the embodiments of this application, the words "exemplary," "for example," or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary," "for example," or "for example" in the embodiments of this application should not be construed as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary," "for example," or "for example" is intended to present the relevant concepts in a concrete manner.

[0160] In the description of the embodiments of the present application, unless otherwise specified, “ / ” means or, for example, A / B can mean A or B; “and / or” in the text is merely a description of the association relationship of associated objects, indicating that three relationships may exist, for example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, in the description of the embodiments of the present application, “multiple” refers to two or more than two.

[0161] In some processes described in the embodiments of the present application, multiple operations or steps are included that appear in a specific order. However, it should be understood that these operations or steps may not be performed in the order in which they appear in the embodiments of the present application or may be performed in parallel. The sequence numbers of the operations are only used to distinguish between different operations, and the sequence numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations or steps may be performed in sequence or in parallel, and these operations or steps may be combined.

[0162] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, of course, it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a number of instructions for enabling a terminal device to execute the methods described in each embodiment of the present application.

[0163] The above are only preferred embodiments of the present application and do not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A link risk assessment method, characterized in that: The link risk assessment method includes: Obtaining a multidimensional feature value corresponding to the link to be predicted N days before the prediction time point, wherein the multidimensional feature value includes feature values of at least two dimensions; Determine the new eigenvalue corresponding to each dimension based on the target subinterval corresponding to each eigenvalue in the dimension to which it belongs; Input the new feature values corresponding to each dimension into the prediction model corresponding to each dimension, and obtain the predicted value of each dimension M days after the prediction time point; The predicted values corresponding to all dimensions are weighted and summed to obtain the link risk assessment value corresponding to the link to be predicted.

2. The link risk assessment method according to claim 1, wherein: Before obtaining the multidimensional feature value corresponding to the link to be predicted N days before the prediction time point, the method further includes: Obtain the multi-dimensional historical feature values corresponding to S sample links N days before the training time point; For each dimension, determine the maximum historical eigenvalue, and divide the interval corresponding to zero to the maximum historical eigenvalue into several sub-intervals based on several quantiles, and set the value corresponding to each sub-interval, where different sub-intervals correspond to different values, and the maximum value is 1.

3. The link risk assessment method according to claim 2, wherein: Determining a new eigenvalue corresponding to each dimension according to a target subinterval corresponding to each eigenvalue in the dimension to which it belongs includes: For each eigenvalue, determine its target subinterval among the subintervals corresponding to the dimension to which it belongs; The value corresponding to the target subinterval is used as the new eigenvalue corresponding to the dimension to which the eigenvalue belongs.

4. The link risk assessment method according to claim 2, wherein: After setting the value corresponding to each sub-interval, the method further includes: Obtain the multi-dimensional true feature values corresponding to S sample links M days after the training time point; For each historical characteristic value, determine the target subinterval in which it is located among the subintervals corresponding to the dimension to which it belongs; Update the historical feature value with the value corresponding to the target subinterval; For each dimension, the new historical eigenvalues and real eigenvalues corresponding to S sample links are used as training samples to train the preset neural network model and obtain a prediction model; And so on, we get the prediction model corresponding to each dimension.

5. The link risk assessment method according to claim 4, wherein: After obtaining the prediction model corresponding to each dimension, the method further includes: Obtain the link status of S sample links M days after the training time point; According to the link status of each sample link, determine the actual link risk value of each sample link; Input the new historical feature value of each sample link in each dimension into the prediction model corresponding to each dimension to obtain the sample prediction value of each sample link in each dimension; The actual link risk values of the S sample links and the sample predicted values of the S sample links in each dimension are combined to obtain the weight corresponding to each dimension.

6. The link risk assessment method according to claim 2, wherein: The S sample links include S1 core layer links, S2 aggregation layer links and S3 access layer links, where S1, S2 and S3 are all preset values.

7. The link risk assessment method according to any one of claims 1 to 6, characterized in that: After obtaining the link risk assessment value corresponding to the link to be predicted, the method further includes: Outputting an operation and maintenance prompt based on the link risk assessment value.

8. A link risk assessment device, characterized in that: The link risk assessment device includes: An acquisition module, configured to acquire multidimensional feature values corresponding to the link to be predicted N days before the prediction time point, wherein the multidimensional feature values include feature values of at least two dimensions; A normalization processing module is used to determine a new eigenvalue corresponding to each dimension based on the target subinterval corresponding to each eigenvalue in the dimension to which it belongs; The prediction module is used to input the new feature values corresponding to each dimension into the prediction model corresponding to each dimension, and obtain the predicted value corresponding to each dimension M days after the prediction time point; The evaluation module is used to perform weighted summation on the prediction values corresponding to all dimensions to obtain a link risk evaluation value corresponding to the link to be predicted.

9. A link risk assessment device, characterized in that: The link risk assessment device includes a processor, a memory, and a link risk assessment program stored in the memory and executable by the processor, wherein when the link risk assessment program is executed by the processor, the steps of the link risk assessment method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a link risk assessment program, wherein when the link risk assessment program is executed by a processor, the steps of the link risk assessment method according to any one of claims 1 to 7 are implemented.

Images