Ethernet connector network monitoring method and system

Abstract

The present invention relates to the field of digital processing technology, and more specifically, to a method and system for monitoring an Ethernet connector network, the method comprising: collecting multiple network parameters of a connector at several moments before the current moment, and screening the target moment based on preset conditions, obtaining a sample set consisting of all network parameters collected at all target moments; training a pre-constructed SVDD model based on a preset objective function and a sample set, and in the training process, using the generalization ability index of each sample as a weight, weighting the penalty factor in the objective function, and training the SVDD model with the goal of minimizing the objective function with the weighted penalty factor, so as to monitor the anomaly of each network parameter in real time based on the trained SVDD model. The present invention can improve the accuracy of the constructed model, so that the anomaly in the network parameters can be accurately monitored, and the accurate monitoring of the status of the Ethernet connector can be achieved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and more specifically, to an Ethernet connector network monitoring method and system. Background Art

[0002] Ethernet connectors are key components for connecting network devices and realizing data transmission. They are usually used between network terminals such as computers, servers, and switches. They can ensure stable and high-speed network connections between different devices, allowing data to flow smoothly in the network. They are the basic components of modern network architecture. Since frequent network failures can seriously affect business operations, real-time monitoring of the network parameters of the connector is crucial to promptly discover anomalies, quickly locate and solve problems.

[0003] In the related technology, for example, a patent application document with publication number CN110348005A discloses a distribution network equipment status data processing method, device, computer equipment and medium, the method comprising: obtaining historical status data of the distribution network equipment; performing word segmentation processing on the historical status data to obtain multiple word segments; performing text clustering processing, keyword calculation processing and association analysis processing on the above multiple word segments; creating a neural network model based on the output parameters of the association analysis processing, and performing data training on the neural network model; updating the weights of the neural network model based on the output parameters of the data training; monitoring the status information of the distribution network equipment based on the updated neural network model, and repairing the distribution network equipment when it is detected that the status of the distribution network equipment is in an abnormal state based on the status information.

[0004] However, when using neural network models to monitor the status information of distribution network equipment, related technologies directly use sample sets to train the models without considering the generalization capabilities of different samples in the training set. As a result, the trained neural network model may overfit samples with low generalization capabilities, thereby affecting the accuracy of the neural network model and making it impossible to accurately monitor the distribution network equipment. Summary of the invention

[0005] In order to solve the problem of inaccurate monitoring results of distribution network equipment due to low model accuracy, the present invention provides an Ethernet connector network monitoring method and system.

[0006] According to a first aspect of the present invention, there is provided an Ethernet connector network monitoring method, comprising:

[0007] Collect multiple network parameters of the connector at several moments before the current moment, and select the target moment based on the preset conditions to obtain a sample set consisting of all network parameters collected at all target moments;

[0008] The pre-built SVDD model is trained based on the preset objective function and sample set. During the training process, the generalization ability index of each sample is used as the weight to weight the penalty factor in the objective function. The SVDD model is trained with the goal of minimizing the objective function with the weighted penalty factor. Based on the trained SVDD model, the abnormality of each network parameter is monitored in real time.

[0009] A method for obtaining the generalization ability index of any sample includes: calculating the time consistency of any sample: ; For the The temporal consistency of the samples; is the sampling time range of all neighboring samples of the sample; For this sample and The distance between the nearest neighbor samples; is the sampling time of the neighbor sample; is the average sampling time of all neighboring samples of this sample; is the number of neighboring samples of the sample; is the normalization function;

[0010] Get the central equidistant sample of the sample, calculate the average of the difference in the number of neighboring samples between the sample and all central equidistant samples, and normalize them to get the criticality of the sample, and calculate the generalization ability index of the sample. The generalization ability index is negatively correlated with time consistency and criticality.

[0011] The present invention can accurately measure the generalization ability of each sample and set different penalty factors according to the level of generalization ability. A lower penalty factor is set for samples with lower generalization ability, and a higher penalty factor is set for samples with better generalization ability, so that during the training process of the SVDD model, the model's fitting ability for samples with high generalization ability can be improved, while overfitting samples with low generalization ability can be avoided. The accuracy of the constructed SVDD model is guaranteed, so that network parameter anomalies can be accurately monitored, and accurate monitoring of the Ethernet connector status can be achieved.

[0012] Preferably, the difference in the number of neighboring samples between any sample and any sample equidistant from the center of the sample is the comprehensive difference, and the comprehensive difference satisfies the following relationship:

[0013] ;

[0014] In the formula, For the The sample and the The combined difference in the number of neighboring samples of samples with equal center distance; For the The sample The number of neighboring samples of equidistant samples from the center; For the The number of neighboring samples of a sample; For the The sample The distance between the center of each sample and the center of the sample; For the The distance between each sample and the sample center; is the natural exponential function; is the activation function; the value of any dimension of the sample center is the average value of all samples in the corresponding dimension, and the dimension corresponds to the type of network parameter one by one.

[0015] The present invention can accurately measure the difference in the number of neighboring samples between any sample and the central equidistant sample, so as to accurately evaluate the possibility that each sample is at the decision boundary.

[0016] Preferably, the criticality of any sample satisfies the following relationship:

[0017] ;

[0018] In the formula, For the The criticality of samples; For the The sample and the The combined difference in the number of neighboring samples of samples with equal center distance; For the The number of equally spaced samples from the center of each sample; is the normalization function.

[0019] Preferably, a method for obtaining a central equidistant sample of any sample comprises:

[0020] The distance between any sample and the sample center is obtained to obtain a reference distance. If the difference between the distance between another sample and the sample center and the reference distance is less than a preset distance, the other sample is used as the center equidistant sample of any sample.

[0021] Preferably, the generalization ability index of any sample satisfies the following relationship:

[0022] ;

[0023] In the formula, For the The generalization ability index of samples; For the The temporal consistency of the samples; For the The criticality of a sample.

[0024] The present invention integrates various data and can accurately measure the generalization ability of each sample, thereby achieving accurate adaptation of the penalty factor of each sample.

[0025] Preferably, before training the pre-built SVDD model based on the preset objective function and the sample set, the method further comprises:

[0026] Each network parameter of any sample in the sample set is normalized to obtain the corresponding standard sample, and the set of all standard samples is used as the training set to train the pre-built SVDD model.

[0027] The present invention can avoid the influence of magnitude differences of different types of network data on model training results.

[0028] Preferably, a method for obtaining a neighboring sample of any sample includes:

[0029] Get the preset neighborhood distance. If the distance between any two samples is less than the neighborhood distance, either sample will be regarded as the neighboring sample of the other sample.

[0030] Preferably, based on the trained SVDD model, the abnormality of each network parameter is monitored in real time, including:

[0031] All network parameters collected at the current moment are used as test data, and the test data is input into the trained SVDD model, and the anomaly detection result of the test data is output to monitor the anomaly of each network parameter in real time according to the anomaly detection result.

[0032] The present invention can ensure the accuracy of the abnormal detection result of the data to be tested, thereby realizing accurate monitoring of the state of the Ethernet connector.

[0033] According to a second aspect of the present invention, there is provided an Ethernet connector network monitoring system, the system comprising a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the first aspect of the present invention.

[0034] The present invention has the following effects:

[0035] 1. The present invention can improve the fitting ability of the SVDD model for samples with higher generalization ability by adaptively adjusting the penalty factor of each sample, and at the same time avoid the model from excessively fitting samples with lower generalization ability, thereby ensuring the accuracy of the constructed model, so as to accurately identify network parameter anomalies and realize accurate monitoring of the Ethernet connector status.

[0036] 2. The present invention combines the temporal consistency and criticality of samples to accurately identify samples with poor generalization ability, thereby ensuring the accuracy of the determined generalization ability index, thereby achieving precise adaptation of the penalty factor of each sample. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0038] Figure 1 The present invention is a schematic diagram of the steps of a method for monitoring an Ethernet connector network according to an embodiment of the present invention. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0040] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0041] Reference Figure 1 , a method for monitoring an Ethernet connector network, comprising steps S1 to S3, specifically as follows:

[0042] S1: Collect multiple network parameters of the connector at several moments before the current moment, and select the target moment based on preset conditions to obtain a sample set consisting of all network parameters collected at all target moments.

[0043] The network parameters refer to data that can reflect the network-related conditions of the Ethernet connector. The network parameters collected in the present invention include: bandwidth utilization, packet rate, byte rate, number of connections, and number of active sessions. The present invention does not specifically limit the types of network parameters collected.

[0044] It should be noted that bandwidth utilization can reflect the overall load of the network and help identify high-traffic periods and potential bottlenecks; packet rate can measure the level of network activity and help detect abnormal traffic peaks or valleys; byte rate can provide more accurate traffic size information and help evaluate bandwidth utilization efficiency; the number of connections can reflect the number of concurrent connections in the network and help identify potential DDoS attacks or other abnormal connection behaviors; the number of active sessions can help evaluate the activity of users in the network and identify abnormal session growth or decrease.

[0045] Specifically, a network parameter collection system can be deployed on the Ethernet connector first, and then the network parameter collection system can be used to collect Poll network device interface statistics to calculate bandwidth utilization; use or The protocol collects traffic statistics from network devices, analyzes the number of packets and bytes to calculate the packet rate and byte rate; Statistics of current activity Connection number, extract the connection number; through the application layer log (such as server logs) or Data statistics on the number of currently active sessions, extract the number of active sessions; then according to a certain sampling frequency, such as At the same time, the bandwidth utilization, data packet rate, byte rate, number of connections and number of active sessions of the connector are collected to obtain multiple network parameters at each sampling moment. This embodiment does not specifically limit the method of collecting network parameters and the sampling frequency.

[0046] Furthermore, after obtaining multiple network parameters at each sampling moment, the target moment among several sampling moments before the current sampling moment can be screened. The specific process is: determine whether there are abnormal data in all network parameters collected at any sampling moment before the current sampling moment. If not, determine that the sampling moment meets the preset conditions and use the sampling moment as the target data; if so, discard it, thereby obtaining all target moments, and taking all network parameters collected at any target moment as a sample to construct a sample set.

[0047] It should be noted that since there are relatively few abnormal samples in network monitoring, it is impossible to effectively cover all abnormal situations. Therefore, the present invention constructs a sample set based on all network parameters collected at the target time, so that all normal situations can be used to identify abnormal situations.

[0048] S2: The pre-built SVDD model is trained based on the preset objective function and sample set. During the training process, the generalization ability index of each sample is used as the weight to weight the penalty factor in the objective function, and the SVDD model is trained with the goal of minimizing the objective function with the weighted penalty factor.

[0049] It should be noted that due to the different sampling times of each sample in the training set, there are differences in the generalization ability of each sample. However, in the training process of the traditional SVDD (Support Vector Data Description) model, the penalty factors of all samples are the same, and the differences in generalization ability of different samples are not taken into account. However, by adopting a unified penalty factor, the model will fit all samples evenly, which may cause the model to overfit samples with low generalization ability, thereby affecting the accuracy of the model. Therefore, the present invention improves the training process of the traditional SVDD model, and the specific improvement process is: according to the generalization ability index of each sample, the penalty factor of each sample is weighted, so as to train the pre-constructed SVDD model based on the objective function with the weighted penalty factor to obtain a trained SVDD model. Among them, the process of constructing the SVDD model is a prior art, and this embodiment will not be described in detail here; the preset objective function is the objective function used in the training process of the traditional SVDD model, and the expression is: ; In the formula, is the radius of the hypersphere; is the center of the hypersphere; is the penalty factor; For the Slack variables for samples; It should be noted that the hypersphere radius, hypersphere center, penalty factor and slack variable are all professional terms in the description of support vector data, and will not be described in detail in this embodiment.

[0050] In an exemplary embodiment of the present invention, before training the pre-built SVDD model based on the preset objective function and the sample set, it also includes:

[0051] Each network parameter of any sample in the sample set is normalized to obtain the corresponding standard sample, and the set of all standard samples is used as the training set to train the pre-built SVDD model.

[0052] It should be noted that since different types of network parameters may have different magnitudes, each data has different effects on the training results of the SVDD model. Therefore, after obtaining the sample set, each network parameter of each sample in the sample set needs to be normalized.

[0053] Specifically, you can use The function normalizes each network parameter of each sample, and may also use the maximum value of each network parameter in all samples to normalize the value of the corresponding network parameter in each sample. This embodiment does not specifically limit the normalization method.

[0054] Furthermore, after the training set of the pre-built SVDD model is determined, the generalization ability index of each sample in the training set can be determined by the following steps:

[0055] Step 1: Calculate the temporal consistency of any sample;

[0056] It should be noted that if the sampling time of similar samples in the training set is close, then all similar samples are likely to be special data collected continuously over a period of time, and are more likely to be special cases that appear within a specific time range, and cannot well represent the general situation of normal samples, and have low generalization ability during model training. Therefore, the present invention uses this feature to calculate the time consistency of any sample, so as to measure the possibility that each sample is a special case that appears within a specific time range, so as to identify samples with low generalization ability.

[0057] Specifically, the temporal consistency of any sample satisfies the following relationship:

[0058] ;

[0059] In the formula, For the The temporal consistency of samples, where temporal consistency refers to data that can reflect the temporal uniformity of samples similar to any sample. For example, when the time intervals of samples similar to any sample differ slightly, the temporal consistency of the sample is relatively large. is the sampling time range of all neighboring samples of the sample, where the neighboring samples of any sample refer to the samples whose distance to any sample is within a certain range, which is used to measure the similarity between samples; For this sample and The distance between the nearest neighbor samples; is the sampling time of the neighbor sample; is the average sampling time of all neighboring samples of this sample; is the number of neighboring samples of the sample; is the normalization function.

[0060] in, Reflects the The sample The sampling time of the nearest neighbor samples is the same as that of the The degree of proximity between the average sampling time of all neighboring samples of a sample. The larger the value, the closer the sampling time of the neighboring sample is to the average sampling time of all neighboring samples. Reflects the The sample The nearest neighbor samples and The similarity of samples. The larger the value, the more similar the neighbor sample is to the sample. Larger, and If it is larger, the description is the same as in The neighboring samples with a high degree of similarity to the sample are closer to the average sampling time of all the neighboring samples of the sample, which further indicates that the time consistency of the sample is relatively large.

[0061] ; In the formula, For the The maximum sampling time among all neighboring samples of a sample; For the The minimum sampling time among all neighboring samples of a sample; When it is smaller, the If the sampling time range of all neighboring samples of a sample is shorter, the time consistency of the sample is relatively large.

[0062] In an exemplary embodiment of the present invention, the neighboring samples of any sample may be determined by the following steps:

[0063] Get the preset neighborhood distance. If the distance between any two samples is less than the neighborhood distance, either sample will be regarded as the neighboring sample of the other sample.

[0064] Optionally, the neighborhood distance may be set to 0.1, and when the distance between any two samples is less than 0.1, any sample is taken as a neighboring sample of the other sample. This embodiment does not specifically limit the size of the neighborhood distance.

[0065] Optionally, the nearest neighbor samples of any sample may be determined based on the Euclidean distance, or the Manhattan distance. Of course, a suitable distance may be selected according to specific circumstances. This embodiment does not specifically limit the selected distance type.

[0066] Step 2: Get the central equidistant sample of the sample, calculate the average value of the difference between the number of neighboring samples of the sample and all central equidistant samples, and normalize them to obtain the criticality of the sample;

[0067] It should be noted that, since abnormal samples will gradually appear near the decision boundary, the number of normal samples near the decision boundary will gradually decrease. Therefore, when the samples in the training set are far away from the sample center, the number of similar samples will gradually decrease. However, compared with other samples at the same distance from the sample center, the probability of the existence of a decision boundary near the sample whose number of similar samples decreases too fast is higher. Therefore, the present invention uses this feature to calculate the criticality of each sample to further identify samples with low generalization ability.

[0068] The center equidistant samples refer to samples whose distances from the sample center are the same or whose distances from the sample center are within a certain range.

[0069] In an exemplary embodiment of the present invention, the determination of the center equidistant samples of any sample can be achieved by the following steps:

[0070] The distance between any sample and the sample center is obtained to obtain a reference distance. If the difference between the distance between another sample and the sample center and the reference distance is less than a preset distance, the other sample is used as the center equidistant sample of any sample.

[0071] The value of any dimension of the sample center is the average value of the corresponding dimension of all samples, and the dimension corresponds to the type of network parameter. For example, the value of bandwidth utilization of the sample center is the average value of bandwidth utilization of all samples in the training set.

[0072] Optionally, the preset distance may be set to 0.1, and if the difference between the distance between another sample and the sample center and the reference distance is less than 0.1, the other sample is used as the center equidistant sample of the corresponding sample, so that all center equidistant samples of each sample can be obtained. This embodiment does not specifically limit the size of the preset distance.

[0073] In another embodiment, all samples with the same distance from the sample center may be directly regarded as center equidistant samples.

[0074] Further, after determining the central equidistant samples of any sample, the difference in the number of neighboring samples between the sample and any central equidistant sample can be calculated. In an exemplary embodiment of the present invention, the difference in the number of neighboring samples between any sample and any central equidistant sample of the sample is a comprehensive difference, and the comprehensive difference satisfies the following relationship:

[0075] ;

[0076] In the formula, For the The sample and the The combined difference in the number of neighboring samples of samples with equal center distance; For the The sample The number of neighboring samples of equidistant samples from the center; For the The number of neighboring samples of a sample; For the The sample The distance between the center of each sample and the center of the sample; For the The distance between each sample and the sample center; is a natural exponential function, where the natural exponential function refers to a function with a natural constant An exponential function with base ; is an activation function, where the activation function is a function used to retain positive values ​​and set negative values ​​to zero.

[0077] in, Reflects the The sample and the The difference in the number of neighboring samples of the center equidistant samples. The larger the value, the The number of neighboring samples of a sample decreases faster than the number of neighboring samples of the sample's center equidistant from the sample, which indicates that the sample is more likely to be on the decision boundary.

[0078] Reflects the The sample The center is equally spaced from the sample The closer the samples are, the larger the value is. The weight of the value is relatively large, so that the The probability that a sample is on the decision boundary.

[0079] Furthermore, after determining the comprehensive difference in the number of neighboring samples between any sample and each of the samples with equal distance from the center of the sample, the criticality of the sample can be calculated. Specifically, the criticality of any sample satisfies the following relationship:

[0080] ;

[0081] In the formula, For the The criticality of samples; For the The sample and the The combined difference in the number of neighboring samples of samples with equal center distance; For the The number of equally spaced samples from the center of each sample; is the normalization function.

[0082] In another embodiment, other normalization methods may be used, such as using a deformed hyperbolic tangent function for normalization. This embodiment does not specifically limit the normalization method.

[0083] Step 3: Calculate the generalization ability index of the sample. The generalization ability index is negatively correlated with time consistency and criticality.

[0084] Specifically, the generalization ability index of any sample satisfies the following relationship:

[0085] ;

[0086] In the formula, For the The generalization ability index of samples; For the The temporal consistency of the samples; For the The criticality of a sample.

[0087] Optionally, when the temporal consistency of any sample is large and the criticality of the sample is large, the sample is more likely to be a sample with low generalization ability. When using the sample to train the SVDD model, a smaller penalty factor is set to reduce the hypersphere boundary around the sample, thereby reducing the model's fitting of the sample, ensuring the accuracy of the fitting results, and further realizing accurate monitoring of the network parameters of the Ethernet connector.

[0088] Furthermore, after determining the generalization ability index of each sample in the training set, the generalization ability index of each sample can be used as a weight to weight the penalty factor of the corresponding sample, thereby obtaining an objective function with a weighted penalty factor, and training the SVDD model with the goal of minimizing the objective function with the weighted penalty factor to obtain a trained SVDD model; wherein, the expression of the objective function with a weighted penalty factor is: ; In the formula, is the radius of the hypersphere; is the center of the hypersphere; is the weighted penalty factor; For the Slack variables for samples; is the total number of samples in the training set; when the training set is used as input to train the SVDD model in this embodiment, the kernel function used is the radial basis kernel function. It should be noted that the training process of the SVDD model is a prior art and will not be described in detail in this embodiment.

[0089] S3: Based on the trained SVDD model, the anomaly of each network parameter is monitored in real time.

[0090] In an exemplary embodiment of the present invention, real-time monitoring of each network parameter can be achieved by the following steps:

[0091] All network parameters collected at the current moment are used as test data, and the test data is input into the trained SVDD model, and the anomaly detection result of the test data is output to monitor the anomaly of each network parameter in real time according to the anomaly detection result.

[0092] It should be noted that after the data to be tested is input into the trained SVDD model, if the distance between the data to be tested and the center of the hypersphere of the trained SVDD model is less than or equal to the radius of the hypersphere, the data to be tested is judged to be normal data; if it is greater than the radius of the hypersphere, the data to be tested is judged to be abnormal data, thereby obtaining the anomaly detection result of the data to be tested and realizing the monitoring of the abnormal situation of each network parameter.

[0093] Furthermore, when it is determined that the data to be tested is abnormal, an alarm is issued, thereby promptly reminding the staff to perform relevant maintenance in a timely manner when an abnormality occurs in the connector network.

[0094] The present invention also provides an Ethernet connector network monitoring system, the system includes a memory and a processor, and a computer program is stored in the memory, the computer program integrates the function of an Ethernet connector network monitoring method, when the computer program is executed, through an Ethernet connector network monitoring method, the present invention can improve the accuracy of the constructed model, so that anomalies in network parameters can be accurately monitored, and accurate monitoring of the Ethernet connector status can be achieved.

[0095] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0096] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A method for monitoring an Ethernet connector network, characterized in that: include: Collect multiple network parameters of the connector at several moments before the current moment, and select the target moment based on the preset conditions to obtain a sample set consisting of all network parameters collected at all target moments; The pre-constructed SVDD model is trained based on the preset objective function and the sample set. During the training process, the generalization ability index of each sample is used as a weight to weight the penalty factor in the objective function, and the SVDD model is trained with the goal of minimizing the objective function with the weighted penalty factor, so as to monitor the abnormality of each network parameter in real time based on the trained SVDD model; The generalization ability index of any sample satisfies the following relationship: ； In the formula, For the The generalization ability index of samples; For the The temporal consistency of the samples; For the The criticality of samples; A method for obtaining the generalization ability index of any sample includes: calculating the time consistency of any sample: ; For the The temporal consistency of the samples; is the sampling time range of all neighboring samples of the sample; For this sample and The distance between the nearest neighbor samples; is the sampling time of the neighbor sample; is the average sampling time of all neighboring samples of this sample; is the number of neighboring samples of the sample; is the normalization function; The central equidistant sample of the sample is obtained, the average value of the difference between the number of neighboring samples of the sample and all central equidistant samples is calculated, and the difference is normalized to obtain the criticality of the sample, and the generalization ability index of the sample is calculated. The generalization ability index is negatively correlated with the time consistency and the criticality.

2. The Ethernet connector network monitoring method according to claim 1, characterized in that: The difference in the number of neighboring samples between any sample and any sample with equal distance from the center of the sample is the comprehensive difference, which satisfies the following relationship: ； In the formula, For the The sample and the The combined difference in the number of neighboring samples of samples with equal center distance; For the The sample The number of neighboring samples of equidistant samples from the center; For the The number of neighboring samples of a sample; For the The sample The distance between the center of each sample and the center of the sample; For the The distance between each sample and the sample center; is the natural exponential function; is an activation function; wherein the value of any dimension of the sample center is the average value of the values ​​of all samples in the corresponding dimension, and the dimension corresponds to the type of network parameter one by one.

3. The Ethernet connector network monitoring method according to claim 2, characterized in that: The criticality of any sample satisfies the following relationship: ； In the formula, For the The criticality of samples; For the The sample and the The combined difference in the number of neighboring samples of samples with equal center distance; For the The number of equally spaced samples from the center of each sample; is the normalization function.

4. The Ethernet connector network monitoring method according to claim 3, characterized in that: A method for obtaining a central equidistant sample of any sample includes: The distance between any sample and the sample center is obtained to obtain a reference distance. If the difference between the distance between another sample and the sample center and the reference distance is less than a preset distance, the other sample is used as the center equidistant sample of the any sample.

5. The Ethernet connector network monitoring method according to claim 1, characterized in that: Before training the pre-built SVDD model based on the preset objective function and the sample set, the method further includes: Each network parameter of any sample in the sample set is normalized to obtain a corresponding standard sample, and the set consisting of all standard samples is used as a training set to train the pre-constructed SVDD model.

6. The Ethernet connector network monitoring method according to claim 1, characterized in that: A method for obtaining a neighboring sample of any sample includes: A preset neighborhood distance is obtained, and if the distance between any two samples is less than the neighborhood distance, any sample is taken as a neighboring sample of another sample.

7. The Ethernet connector network monitoring method according to claim 1, characterized in that: Based on the trained SVDD model, the abnormality of each network parameter is monitored in real time, including: All network parameters collected at the current moment are used as test data, the test data are input into the trained SVDD model, and the anomaly detection result of the test data is output to monitor the anomaly of each network parameter in real time according to the anomaly detection result.

8. An Ethernet connector network monitoring system, characterized in that: The Ethernet connector network monitoring system comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the Ethernet connector network monitoring method as claimed in any one of claims 1 to 7.

Images