Network state health monitoring method and system

Abstract

The invention discloses a health monitoring method and system for a network state. The health monitoring method comprises the following steps: acquiring multi-class historical state data of nodes and links in a network; performing data preprocessing on the multiple types of historical state data; splitting the data into training data and verification data; defining a hybrid neural network monitoring model based on a convolutional neural network and a double-circulation gating unit, and performing health monitoring on a network state; the method comprises the following steps: acquiring multi-class state data in nodes and links in a network within current time and a period of historical time away from the current time, and preprocessing the acquired data; and inputting the preprocessed data into the trained hybrid neural network monitoring model to perform health monitoring on the network state. According to the invention, through the adaptive k-proximity joint fuzzy c-mean reduction algorithm, the hybrid neural network monitoring model based on the convolutional neural network and the double-circulation gating unit is adopted, and multiple types of data are used for monitoring, so that the effectiveness and accuracy of health monitoring are improved.

Description

Technical Field

[0001] The present invention relates to the field of network communications, and in particular to a network health monitoring method and system. Background Art

[0002] In the digital age, more and more IT products are directly or wirelessly connected through distributed networks. Networked devices or products are constantly expanding in scale to respond to the needs of global customers. These devices or products are connected to other systems through the Internet. They cover traditional computing hardware such as laptops, desktops, data centers or storage systems, common mobile devices such as smartphones or tablets, robots and self-driving cars, as well as more and more physical devices and objects. These devices / products are critical to enterprises and organizations, and any failure will lead to business interruption and revenue loss. When the existing network fault detection technology detects a fault, the fault has already occurred in the network, which will cause business interruption and revenue loss in the network. The existing network fault detection method usually collects a certain type of data in the network operation, simply cleans and integrates the data, and detects whether there is a fault in the network through machine learning models such as convolutional neural networks. This approach only cleans and integrates the data before inputting it into the model, which increases the difficulty of model training and detection and consumes a lot of computing resources. Secondly, the existing integration method does not take into account the inherent manifold structure and clustering structure of the data, which may weaken the clustering structure after data integration and affect the accuracy of network model detection. In addition, although the data is integrated in the existing technology, the amount of data after integration is still relatively large, and the use of convolutional neural networks faces the problem of overfitting. In addition, since only one type of data is used in fault detection, the accuracy of the detection results is low. Summary of the invention

[0003] The embodiments of the present application provide a network health monitoring method, system and storage medium, which can prevent the occurrence of failures in advance and effectively improve the accuracy of network health monitoring.

[0004] In a first aspect, the present application provides a network health monitoring method, comprising:

[0005] Step S1, obtaining multiple types of historical status data of nodes and links in the network;

[0006] Step S2, performing data preprocessing on the acquired multiple types of historical status data;

[0007] Data preprocessing includes cleaning and data reduction of acquired data;

[0008] Step S3, splitting the preprocessed data into training data and verification data;

[0009] Step S4, defining a hybrid neural network monitoring model based on a convolutional neural network and a dual-loop gating unit to monitor the health of the network;

[0010] The hybrid neural network monitoring model includes the following layers: input layer; convolution layer: Conv1D layer is used for feature extraction; double recurrent gated unit layer: two sets of double recurrent gated unit layers process the features extracted from Conv1D; connection layer and output layer;

[0011] Step S5, training the hybrid neural network monitoring model by using the training data and the verification data to obtain a trained detection model;

[0012] Step S6, obtaining multiple types of status data of nodes and links in the network at the current time and within a period of historical time from the current time, and preprocessing the obtained data in the manner of step S2;

[0013] Step S7, inputting the data pre-processed in step S6 into the trained hybrid neural network monitoring model to perform health monitoring on the network.

[0014] In a possible implementation, the multiple categories of historical status data or the multiple categories of status data include more than two of node running time, CPU load, memory usage, disk space, bandwidth utilization, packet loss rate, transmission delay, and port throughput.

[0015] In a possible implementation, data reduction uses an adaptive k-neighbor joint fuzzy c-mean reduction algorithm to reduce each type of state data obtained. The adaptive k-neighbor joint fuzzy c-mean reduction algorithm is as follows:

[0016]

[0017] Where X is the cleaned data set, which can be divided into c non-empty disjoint subsets, b k is the center of the kth subset, n is the number of data in the cleaned data set X, J is the reduced data set, λ≥0 is the trade-off parameter, α>1 is the fuzzy index, P is the projection matrix, and P T P=I,x i 、x j are the i-th and j-th cleaned data respectively, X is divided into c non-empty disjoint subsets G1∪G2∪…∪G c , where G is a set of c non-empty disjoint subsets; the centroid of each subset is O = [o1, o2, ..., o k ],k=1,2,…,c;U=[u ki ] is the membership matrix, u ki Represents point x i With subset G kThe membership value between is calculated as:

[0018]

[0019] where d ki =||x i -O k ||2 represents x i With O k The distance between.

[0020] In a possible implementation, the convolutional neural network in the hybrid neural network monitoring model is used as a feature extractor, the dual-loop gating unit performs feature processing, the convolutional layer of the convolutional neural network outputs to the first layer of the dual-loop gating unit, the first layer of the dual-loop gating unit outputs to the second layer of the dual-loop gating unit, the second layer of the dual-loop gating unit is connected to the fully connected layer, and the fully connected layer is connected to the output layer.

[0021] In a possible implementation, the method further includes: presetting a health threshold, comparing the output result with the health threshold, if the output result is greater than the health threshold, it proves that the network is operating normally, if it is less than the threshold, there is a hidden danger of failure in the network, and an alarm message is initiated.

[0022] In a second aspect, the present application provides a network health monitoring system, comprising:

[0023] Data acquisition module, data preprocessing module, data splitting module, model training module, health monitoring module;

[0024] A data acquisition module is used to obtain various types of historical status data of nodes and links in the network;

[0025] A data preprocessing module, used for preprocessing the acquired data;

[0026] Data preprocessing includes cleaning and data reduction of the collected data;

[0027] A data splitting module is used to split the collected device-related data into training data and verification data;

[0028] The health monitoring module is used to define a hybrid neural network monitoring model based on convolutional neural network and dual-loop gating unit to monitor the health of the network;

[0029] A model training module, which performs machine learning on the hybrid neural network monitoring model defined by training data and verification data to obtain a trained hybrid neural network monitoring model;

[0030] The data acquisition module is also used to acquire multiple types of status data of nodes and links in the network at the current time and within a period of historical time from the current time, and preprocess the acquired multiple types of status data using the data preprocessing module;

[0031] The health monitoring module is also used to input the multiple types of status data preprocessed by the data preprocessing module into the trained hybrid neural network monitoring model to perform health monitoring on the network.

[0032] In a possible implementation, the multiple categories of historical status data or the multiple categories of status data include more than two of node running time, CPU load, memory usage, disk space, bandwidth utilization, packet loss rate, transmission delay, and port throughput.

[0033] In a possible implementation, data reduction uses an adaptive k-neighbor joint fuzzy c-mean reduction algorithm to reduce each type of state data obtained. The adaptive k-neighbor joint fuzzy c-mean reduction algorithm is as follows:

[0034]

[0035] Where X is the cleaned data set, which can be divided into c non-empty disjoint subsets, b k is the center of the kth subset, n is the number of data in the cleaned data set X, J is the reduced data set, λ≥0 is the trade-off parameter, α>1 is the fuzzy index, P is the projection matrix, and P T P=I,x i 、x j are the i-th and j-th cleaned data respectively, X is divided into c non-empty disjoint subsets G1∪G2∪…∪G c , where G is a set of c non-empty disjoint subsets; the centroid of each subset is O = [o1, o2, ..., o k ],k=1,2,…,c;U=[u ki ] is the membership matrix, u ki Represents point x i With subset G k The membership value between is calculated as:

[0036]

[0037] where d ki =||x i -O k ||2 represents x i With O k The distance between.

[0038] In a possible implementation, the convolutional neural network in the hybrid neural network monitoring model is used as a feature extractor, the dual-loop gating unit performs feature processing, the convolutional layer of the convolutional neural network outputs to the first layer of the dual-loop gating unit, the first layer of the dual-loop gating unit outputs to the second layer of the dual-loop gating unit, the second layer of the dual-loop gating unit is connected to the fully connected layer, and the fully connected layer is connected to the output layer.

[0039] In a possible implementation, the system also includes: an alarm module, which pre-sets a health threshold and compares the output result with the health threshold. If the output result is greater than the health threshold, it proves that the network is operating normally. If it is less than the threshold, there is a hidden danger of failure in the network and an alarm message is initiated.

[0040] In a third aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions enable the computer to execute the method provided in the first aspect.

[0041] The embodiment of the present invention provides a network health monitoring method and system, which monitors the health status of the network, discovers potential fault risks, and effectively avoids business interruption and revenue loss in the network due to faults; monitors the health status of the network system operation by using multiple types of data, thereby improving the accuracy of network health monitoring; reduces the data before inputting into the neural network, and adopts an adaptive k-nearest joint fuzzy c-mean reduction algorithm during data reduction, which simultaneously learns the projection matrix and the membership matrix, and considers the inherent manifold structure of the data and the clustering of the data, so that the data has a better clustering structure; at the same time, the health monitoring model adopts a hybrid neural network monitoring model based on a convolutional neural network and a dual-loop gating unit, and integrates the convolutional neural network and the dual-loop gating unit into a deep learning framework. This unique combination of parallel and sequential processing enables the neural network to simultaneously utilize the advantages of convolutional layers and loop layers, thereby improving the effectiveness and accuracy of health monitoring.

[0042] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description and the drawings.

[0043] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A flow chart of a method for network status health monitoring provided in an embodiment of the present application;

[0045] Figure 2 A schematic diagram of the structure of a network health monitoring system provided by the present application is provided for the embodiment of the present application;

[0046] Figure 3 A structural diagram of a hybrid neural network monitoring model based on a convolutional neural network and a dual-loop gating unit provided in an embodiment of the present application. DETAILED DESCRIPTION

[0047] In order to better understand the technical solution of the present application, the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0048] It should be clear that the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0049] The terms used in the embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to limit the present application. The singular forms "a", "said" and "the" used in the embodiments of the present application and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings.

[0050] It should be understood that the term "and / or" used in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0051] Figure 1 A method for monitoring network health provided by the present application includes:

[0052] Step S1, obtaining multiple types of historical status data of nodes and links in the network.

[0053] Specifically, when the network is working, data can be collected through the Internet of Things (such as smart sensors, RFID), application programming interfaces (APIs), software development kits, web crawlers, and other methods, and the collected data can be stored in the server. This application obtains the status data of each node and link in the network in real time by setting corresponding sensors, including node running time, CPU load, memory usage, disk space, bandwidth utilization, packet loss rate, transmission delay, port throughput, etc., and uploads the acquired data to the server for storage, and can be obtained from the server when needed.

[0054] Step S2, preprocessing the acquired data.

[0055] Data preprocessing includes cleaning and data reduction of the collected data.

[0056] Since the data contains data impurities such as omissions and duplications, the data must be cleaned. Data cleaning can remove duplicate records, correct erroneous data, handle missing data, and eliminate noise.

[0057] Since the amount of data obtained is relatively large, and multiple different types of data are obtained, which are multi-dimensional data, if the cleaned data is directly input into the model for training, it will increase the difficulty of model training and increase the computing resources required for model training and monitoring. Therefore, it is necessary to reduce the cleaned data.

[0058] Specifically, the data reduction uses an adaptive k-neighbor joint fuzzy c-mean reduction algorithm to reduce each type of state data obtained. The adaptive k-neighbor joint fuzzy c-mean reduction algorithm is as follows:

[0059]

[0060] Where X is the cleaned data set, which can be divided into c non-empty disjoint subsets, b k is the center of the kth subset, n is the number of data in the cleaned data set X, J is the reduced data set, λ≥0 is the trade-off parameter, α>1 is the fuzzy index, P is the projection matrix, and P T P=I,x i 、x j are the i-th and j-th cleaned data respectively, X is divided into c non-empty disjoint subsets G1∪G2∪…∪G c , where G is a set of c non-empty disjoint subsets; the centroid of each subset is O = [o1, o2, ..., o k ],k=1,2,…,c;U=[u ki ] is the membership matrix, u ki Represents point x i With subset G k The membership value between is calculated as:

[0061]

[0062] where d ki =||x i -O k ||2 represents x i With O k The distance between.

[0063] The adaptive k-nearest joint fuzzy c-means reduction algorithm simultaneously learns the projection matrix and the membership matrix, which makes the data have a better clustering structure. Therefore, the reduction algorithm considers both the inherent manifold structure of the data and the clustering of the data.

[0064] Step S3, split the preprocessed data into training data and verification data.

[0065] Specifically, 80% of the preprocessed data is used as training data and 20% is used as validation data.

[0066] Step S4, defining a hybrid neural network monitoring model based on a convolutional neural network and a dual-loop gating unit to perform health monitoring on the network.

[0067] Specifically, the structure of the hybrid neural network monitoring model based on convolutional neural network and dual recurrent gating unit is as follows: Figure 3 As shown, the structure includes the following layers:

[0068] Input layer;

[0069] Convolutional Layers: Conv1D layers are used for feature extraction, followed by pooling and dropout.

[0070] Double recurrent gating unit layer: Two sets of double recurrent gating unit layers process the features extracted from Conv1D. The first set returns the sequence, and the second set returns only the final output of each sequence. Layer normalization and dropout are applied to the two double recurrent gating unit layers for regularization.

[0071] Fully connected layers and output layer.

[0072] The convolutional neural network is used for feature extraction, and the dual-circular gating unit performs feature processing. The convolutional layer of the convolutional neural network outputs to the first layer of the dual-circular gating unit, the first layer of the dual-circular gating unit outputs to the second layer of the dual-circular gating unit, the second layer of the dual-circular gating unit is connected to the fully connected layer, and the fully connected layer is connected to the output layer.

[0073] The hybrid model integrates two effective neural networks, a convolutional neural network and a dual recurrent gating unit. The Conv1D layer in the convolutional neural network represents feature extraction, and the dual recurrent gating unit layer effectively processes multi-dimensional data. The structure of the hybrid model seamlessly integrates two dual recurrent gating unit layers within the convolutional neural network framework. Placing these dual recurrent gating unit layers before the fully connected layer has advantages, such as effectively training the multi-dimensional features extracted by the convolutional neural network, avoiding overfitting caused by directly using the convolutional neural network to process multi-dimensional data. In addition, this integration facilitates the fusion of spatial features captured by the convolutional neural network with the temporal complexity processed by the dual recurrent gating unit in subsequent layers. It is worth noting that the adaptability of the convolutional neural network and the dual recurrent gating unit enables the model to dynamically adjust its feature extraction strategy to effectively align with the unique characteristics of the training data. This adaptability ensures that the model can capture and learn from various complex situations in the data, thereby improving its overall health monitoring capabilities and improving the accuracy of health monitoring.

[0074] Step S5, training a hybrid neural network monitoring model based on a convolutional neural network and a dual-loop gating unit by using the training data and the verification data to obtain a trained detection model.

[0075] Using training sets and validation sets to train models belongs to the existing technology and will not be described in detail here.

[0076] Step S6, obtaining multiple types of status data of nodes and links in the network at the current time and within a historical period from the current time, and preprocessing the obtained data in the manner of step S2.

[0077] Specifically, multiple types of status data can be obtained for the current time and the last hour, or for the current time and the last day. The specific time can be set as needed. Multiple types of status data may include node running time, CPU load, memory usage, disk space, bandwidth utilization, packet loss rate, transmission delay, port throughput, etc. Preprocess each type of data obtained.

[0078] Step S7, input the data pre-processed by S6 into the trained neural network to perform health monitoring on the network.

[0079] Specifically, after preprocessing each type of data in the multiple types of data, the preprocessed data is input into the trained hybrid neural network monitoring model to obtain an output value, and the output value is used to determine whether the network has a potential fault. Since multiple types of data are used for monitoring, the accuracy of monitoring is improved compared to using a single indicator data.

[0080] Specifically, a health threshold is set in advance, and the output result is compared with the health threshold. If the output result is greater than the health threshold, it proves that the network is operating normally. If it is less than the threshold, there is a hidden danger of network failure. At this time, an alarm message can be initiated to remind relevant personnel to check and eliminate hidden dangers to ensure long-term normal operation of the network.

[0081] In summary, in combination with all the above embodiments, the present application monitors the health status of the network system operation by using multiple types of data and monitoring through multiple indicator data, thereby improving the accuracy of health monitoring. When reducing data, an adaptive k-neighbor joint fuzzy c-mean reduction algorithm is used. The algorithm simultaneously learns the projection matrix and the membership matrix, so that it can simultaneously consider the inherent manifold structure of the data and the clustering of the data, so that the data has a better clustering structure. At the same time, the health monitoring model adopts a hybrid neural network monitoring model based on convolutional neural networks and dual-loop gating units, integrating convolutional neural networks and dual-loop gating units into a deep learning framework. This unique combination of parallel and sequential processing enables the neural network to simultaneously take advantage of the advantages of convolutional layers and recurrent layers, thereby improving effective network status health monitoring.

[0082] Please refer to Figure 2 , Figure 2 A schematic diagram of the structure of a network health monitoring system provided for this application includes:

[0083] Data acquisition module, data preprocessing module, data splitting module, model training module, health monitoring module,

[0084] The data acquisition module is used to obtain various types of historical status data of nodes and links in the network.

[0085] The data preprocessing module preprocesses the acquired data.

[0086] Data preprocessing includes cleaning and data reduction of the collected data.

[0087] The data splitting module splits the collected device-related data into training data and verification data.

[0088] The health monitoring module defines a hybrid neural network monitoring model based on convolutional neural network and dual-loop gating unit to perform health monitoring of the network status.

[0089] The model training module performs machine learning by training the defined deep neural network model using training data and validation data to obtain a trained network model.

[0090] The data acquisition module is also used to acquire multiple types of status data of nodes and links in the network at the current time and within a historical period from the current time, and pre-process the acquired data using the data pre-processing module.

[0091] The health monitoring module inputs the current time and multiple types of status data within a period of historical time preprocessed by the data preprocessing module into the trained neural network to perform health monitoring on the network status.

[0092] In a possible implementation, the multiple categories of historical status data or the multiple categories of status data include more than two of node running time, CPU load, memory usage, disk space, bandwidth utilization, packet loss rate, transmission delay, and port throughput.

[0093] In a possible implementation, data reduction uses an adaptive k-neighbor joint fuzzy c-mean reduction algorithm to reduce each type of state data obtained. The adaptive k-neighbor joint fuzzy c-mean reduction algorithm is as follows:

[0094]

[0095] Where X is the cleaned data set, which can be divided into c non-empty disjoint subsets, b k is the center of the kth subset, n is the number of data in the cleaned data set X, J is the reduced data set, λ≥0 is the trade-off parameter, α>1 is the fuzzy index, P is the projection matrix, and P T P=I,x i 、x j are the i-th and j-th cleaned data respectively, X is divided into c non-empty disjoint subsets G1∪G2∪…∪G c , where G is a set of c non-empty disjoint subsets; the centroid of each subset is O = [o1, o2, ..., o k ],k=1,2,…,c;U=[u ki ] is the membership matrix, u ki Represents point x i With subset G k The membership value between is calculated as:

[0096]

[0097] where d ki =|x i -O k ||2 represents x i With O k The distance between.

[0098] In a possible implementation, the convolutional neural network in the hybrid neural network monitoring model is used as a feature extractor, the dual-loop gating unit performs feature processing, the convolutional layer of the convolutional neural network outputs to the first layer of the dual-loop gating unit, the first layer of the dual-loop gating unit outputs to the second layer of the dual-loop gating unit, the second layer of the dual-loop gating unit is connected to the fully connected layer, and the fully connected layer is connected to the output layer.

[0099] In a possible implementation, the system also includes: an alarm module, which pre-sets a health threshold and compares the output result with the health threshold. If the output result is greater than the health threshold, it proves that the network is operating normally. If it is less than the threshold, there is a hidden danger of failure in the network and an alarm message is initiated.

[0100] Since the embodiments of the system part correspond to the embodiments of the method part, the embodiments of the device part refer to the description of the embodiments of the method part and will not be repeated here.

[0101] An embodiment of the present application also provides a computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are executed on an electronic device, the electronic device executes the aforementioned network status health monitoring method.

[0102] The above-mentioned computer-readable storage medium can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, - but not limited to - an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM) or flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0103] Computer-readable signal media may include a data signal propagated in baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including, but not limited to, electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0104] The program code embodied on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wire, optical cable, radio frequency (RF), etc., or any suitable combination of the foregoing.

[0105] Computer program code for performing the operations of the present specification may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0106] The embodiment of the present application also provides a computer program product. When the computer program product is run on a computer, the computer is enabled to execute part or all of the steps in the above method embodiment.

[0107] Through the description of the above implementation methods, technical personnel in the relevant field can understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0108] The above contents are only specific implementation methods of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for monitoring the health of a network state, comprising: Step S1, obtaining multiple types of historical status data of nodes and links in the network; Step S2, performing data preprocessing on the acquired multiple types of historical status data; Data preprocessing includes cleaning and data reduction of acquired data; Step S3, splitting the preprocessed data into training data and verification data; Step S4, defining a hybrid neural network monitoring model based on a convolutional neural network and a dual-loop gating unit to perform health monitoring on the network status; The hybrid neural network monitoring model includes the following layers: an input layer; Convolutional layer: Conv1D layer is used for feature extraction; Double recurrent gated unit layer: Two sets of double recurrent gated unit layers process the features extracted from Conv1D; Connection layer and output layer; Step S5, training the hybrid neural network monitoring model by using the training data and the verification data to obtain a trained monitoring model; Step S6, obtaining multiple types of status data of nodes and links in the network at the current time and within a period of historical time from the current time, and preprocessing the obtained data in the manner of step S2; Step S7, inputting the data pre-processed in step S6 into the trained hybrid neural network monitoring model to perform health monitoring on the network status.

2. The method according to claim 1, characterized in that: The multiple categories of historical status data or the multiple categories of status data include two or more of node running time, CPU load, memory usage, disk space, bandwidth utilization, packet loss rate, transmission delay, and port throughput.

3. The method according to claim 1, characterized in that: The data reduction adopts an adaptive k-neighbor joint fuzzy c-mean reduction algorithm to reduce each type of state data obtained respectively. The adaptive k-neighbor joint fuzzy c-mean reduction algorithm is as follows: Where X is the cleaned data set, which can be divided into c non-empty disjoint subsets, b k is the center of the kth subset, n is the number of data in the cleaned data set X, J is the reduced data set, λ≥0 is the trade-off parameter, α>1 is the fuzzy index, P is the projection matrix, and P T P=I,x i 、x j are the i-th and j-th cleaned data respectively, X is divided into c non-empty disjoint subsets G1∪G2∪…∪G c , where G is a set of c non-empty disjoint subsets; the centroid of each subset is O = [o1, o2, ..., o k ],k=1,2,…,c;U=[u ki ] is the membership matrix, u ki Represents point x i With subset G k The membership value between is calculated as: where d ki =||x i -O k ||2 represents x i With O k The distance between.

4. The method according to claim 1, characterized in that: In the hybrid neural network monitoring model, the convolutional neural network is used as a feature extractor, the dual-loop gating unit performs feature processing, the convolutional layer of the convolutional neural network outputs to the first layer of the dual-loop gating unit, the first layer of the dual-loop gating unit outputs to the second layer of the dual-loop gating unit, the second layer of the dual-loop gating unit is connected to the fully connected layer, and the fully connected layer is connected to the output layer.

5. The method according to claim 1, characterized in that: The method also includes: presetting a health threshold, comparing the output result with the health threshold, if the output result is greater than the health threshold, it proves that the network is running normally, if it is less than the threshold, there is a hidden danger of failure in the network, and an alarm message is issued.

6. A network status health monitoring system, data acquisition module, data preprocessing module, data splitting module, model training module, health monitoring module, A data acquisition module is used to obtain various types of historical status data of nodes and links in the network; A data preprocessing module, used for preprocessing the acquired data; Data preprocessing includes cleaning and data reduction of the collected data; A data splitting module is used to split the collected device-related data into training data and verification data; The health monitoring module is used to define a hybrid neural network monitoring model based on convolutional neural network and dual-loop gating unit to monitor the health of the network; A model training module, which performs machine learning on the hybrid neural network monitoring model defined by training data and verification data to obtain a trained hybrid neural network monitoring model; The data acquisition module is also used to acquire multiple types of status data of nodes and links in the network at the current time and within a period of historical time from the current time, and pre-process the acquired multiple types of status data using the data pre-processing module; The health monitoring module is also used to input the multiple types of status data preprocessed by the data preprocessing module into the trained hybrid neural network monitoring model to perform health monitoring on the network.

7. The system according to claim 6, characterized in that: The multiple categories of historical status data or the multiple categories of status data include more than two types, including node running time, CPU load, memory usage, disk space, bandwidth utilization, packet loss rate, transmission delay, and port throughput.

8. The system according to claim 6, characterized in that: The data reduction adopts an adaptive k-neighbor joint fuzzy c-mean reduction algorithm to reduce each type of state data obtained respectively. The adaptive k-neighbor joint fuzzy c-mean reduction algorithm is as follows: Where X is the cleaned data set, which can be divided into c non-empty disjoint subsets, b k is the center of the kth subset, n is the number of data in the cleaned data set X, J is the compressed data set, λ≥0 is the trade-off parameter, α>1 is the fuzzy index, P is the projection matrix, and P T P=I,x i 、x j are the i-th and j-th cleaned data respectively, X is divided into c non-empty disjoint subsets G1∪G2∪…∪G c , where G is a set of c non-empty disjoint subsets; the centroid of each subset is O = [o1, o2, ..., o k ],k=1,2,…,c;U=[u ki ] is the membership matrix, u ki Represents point x i With subset G k The membership value between is calculated as: where d ki =||x i -O k ||2 represents x i With O k The distance between.

9. The system according to claim 6, characterized in that: In the hybrid neural network monitoring model, the convolutional neural network is used as a feature extractor, the dual-loop gating unit performs feature processing, the convolutional layer of the convolutional neural network outputs to the first layer of the dual-loop gating unit, the first layer of the dual-loop gating unit outputs to the second layer of the dual-loop gating unit, the second layer of the dual-loop gating unit is connected to the fully connected layer, and the fully connected layer is connected to the output layer.

10. A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed, the network health monitoring method according to any one of claims 1 to 5 is implemented.

Images