Network quality prediction method and device for wide area network, equipment and medium

Abstract

The invention provides a network quality prediction method and device for a wide area network, equipment and a medium, and the method comprises the steps: determining a plurality of target sub-links in a communication link based on historical network data from a target user side to a target server side, the network quality of a target sub-link under a plurality of preset dimensions is detected through a detection client to obtain historical network quality detection data, an IP address set of a global wide area network and the historical network quality detection data are fused to obtain a network data set, and the network data set is classified based on a classification rule. Selecting a target feature extraction model based on the data type of the classified network data set, extracting multi-dimensional feature data based on the target feature extraction model, fusing the multi-dimensional feature data to obtain target feature data, and inputting the target feature data into a feature processing model to obtain a feature extraction model; and outputting a network prediction result according to the network attributes of different dimensions, and determining and avoiding a network fault.

Description

Technical Field

[0001] The present application relates to the technical field of network quality prediction, and in particular to a method, apparatus, device and medium for predicting network quality of a wide area network. Background Art

[0002] Wide area networks (WANs) are comprised of multiple complex components and infrastructure, including cell egress, regional aggregation switches, ISPs, AS number broadcasts, submarine and terrestrial cables, cross-border lines, and IXP exchange centers. These components work together to transmit a data packet from one end to the other. However, WANs suffer from high jitter and poor stability, coupled with secondary issues caused by external factors like natural disasters and construction, making network quality difficult to guarantee.

[0003] Currently, the main way to circumvent these inherent secondary problems is to schedule exits and paths after they occur.

[0004] However, existing methods only respond after problems occur, which has certain limitations and makes it difficult to fundamentally prevent the occurrence of problems. Even if appropriate avoidance is achieved through scheduling in long WAN links, there will still be a period of time when network quality degrades, affecting the user experience. Summary of the Invention

[0005] In view of this, the purpose of this application is to provide a network quality prediction method, device, equipment and medium for a wide area network. The method determines multiple target sub-links in a communication link based on historical network conditions, and performs network quality detection on the target sub-links under multiple preset dimensions. The historical network quality detection data and the collected wide area network IP address set are vectorized through a target feature extraction model and input into a feature processing model for network quality prediction. This method can predict future network failures in the wide area network and avoid them in advance, fundamentally preventing the occurrence of secondary problems that affect network quality, avoiding the limitations of post-avoidance, and improving network quality and user experience. At the same time, through early prediction and avoidance, higher efficiency and reliability can be achieved in wide area network management.

[0006] In a first aspect, an embodiment of the present application provides a method for predicting network quality of a wide area network, for detecting the network quality of a communication link between a target user terminal and a target server terminal, wherein the target user terminal and the target server terminal include multiple data exchange components, and correspondingly, the communication link includes multiple sub-links, each sub-link including any two nodes between the target user terminal and the target server terminal; the method includes: Based on pre-stored historical network data from a target user end to a target server end, multiple target sub-links in the communication link are determined, and the network quality of the target sub-links under multiple preset dimensions is detected by a detection client to obtain historical network quality detection data of the target sub-links under the multiple preset dimensions, and the historical network quality detection data is normalized based on a preset normalization rule; wherein the multiple preset dimensions include at least source address, destination address, delay, packet loss rate, and intermediate path nodes; Collecting a set of IP addresses of a global wide area network, matching and fusing the set of IP addresses with the historical network quality detection data to obtain a network data set, and classifying the network data set based on a preset classification rule to obtain a classified network data set; wherein the set of IP addresses includes the IP address information of the target sub-link in the historical network quality detection data, and the network data set includes a plurality of IP address subsets obtained by classifying the set of IP addresses and the classified historical network quality detection data; Determine and select a target feature extraction model for the network data set based on the data type of the classified network data set, and extract multi-dimensional feature data corresponding to the network data set based on the target feature extraction model; After the multi-dimensional feature data are fused to obtain target feature data, the target feature data is input into a preset feature processing model, and corresponding network prediction results are output for network attributes of different dimensions. Based on the network prediction results, possible network failures in the future are determined, and the network failures are avoided; wherein the prediction results represent the prediction results under network attributes of different dimensions.

[0007] In a possible implementation, the regularizing of the historical network quality detection data includes: Determining a standard data type and a standard data format for each preset dimension in the historical network quality detection data, and comparing the data types and data formats with the data for each preset dimension in the historical network quality detection data, to filter out useless data and blank data in the historical network quality detection data; The useless data is cleaned, and the blank data is filled based on the standard data type and standard data format under the preset dimension corresponding to the blank data.

[0008] In a possible implementation, the target feature extraction model includes a target time series model and a target general large model, and extracting multi-dimensional feature data corresponding to the network data set based on the target feature extraction model includes: Determining a first target feature dimension for extracting features from the network data set by the target time series model, performing feature extraction on the network data set based on the first target feature dimension, and obtaining time series feature data of the network data set; A second target feature dimension for extracting features from the network data set by the target universal macromodel is determined, and features are extracted from the network data set based on the second target feature dimension to obtain universal feature data of the network data set.

[0009] In a possible implementation, avoiding the network failure includes: Determining a target sub-link and network attributes corresponding to the network failure, and generating a corresponding avoidance strategy based on the network failure; After degrading the target sub-link and network attributes of the network failure, the network failure is avoided based on the avoidance strategy.

[0010] In a possible implementation, the feature processing model includes a fully connected layer, and the method further includes: Determining the task complexity of a preset network quality prediction task and the dimension of the hidden layer of the fully connected layer; The dimension of the hidden layer of the fully connected layer is adjusted based on the task complexity.

[0011] In a possible implementation, before inputting the target feature data into a preset feature processing model, fusing the multi-dimensional feature data includes: Entering the feature data into a preset long-term storage vector database; Based on the first target feature dimension of the time series feature data and the second target feature dimension of the universal feature data, feature fusion is performed on the time series feature data and the universal feature data.

[0012] In one possible implementation, the method further includes: Displaying the feature data and the network prediction results based on a preset visualization platform; Respective alarm thresholds are set for the delay and the packet loss rate, and an alarm is triggered when the network prediction result meets the alarm threshold.

[0013] In a second aspect, an embodiment of the present application further provides a network quality prediction device for a wide area network, for detecting the network quality of a communication link from a target user end to a target server end, wherein the target user end and the target server end include multiple data exchange components, and correspondingly, the communication link includes multiple sub-links, each sub-link includes any two nodes between the target user end and the target server end; the device includes: A detection module is configured to determine multiple target sub-links in a communication link based on pre-stored historical network data from a target user end to a target server end, detect the network quality of the target sub-links under multiple preset dimensions through a detection client, obtain historical network quality detection data of the target sub-links under the multiple preset dimensions, and normalize the historical network quality detection data based on preset normalization rules; wherein the multiple preset dimensions include at least source address, destination address, delay, packet loss rate, and intermediate path nodes; a classification module for collecting a set of IP addresses of a global wide area network, matching and fusing the set of IP addresses with the historical network quality detection data to obtain a network data set, and classifying the network data set based on a preset classification rule to obtain a classified network data set; wherein the set of IP addresses includes the IP address information of the target sub-link in the historical network quality detection data, and the network data set includes a plurality of IP address subsets obtained by classifying the set of IP addresses and the classified historical network quality detection data; an extraction module, configured to determine and select a target feature extraction model for the network data set based on the data type of the classified network data set, and extract multi-dimensional feature data corresponding to the network data set based on the target feature extraction model; The prediction module is used to input the target feature data into a preset feature processing model after fusing the multi-dimensional feature data to obtain the target feature data, output corresponding network prediction results for network attributes of different dimensions, determine possible network failures in the future based on the network prediction results, and avoid the network failures; wherein the prediction results represent the prediction results under network attributes of different dimensions.

[0014] In a possible implementation, the detection module is specifically configured to: Determining a standard data type and a standard data format for each preset dimension in the historical network quality detection data, and comparing the data types and data formats with the data for each preset dimension in the historical network quality detection data, to filter out useless data and blank data in the historical network quality detection data; The useless data is cleaned, and the blank data is filled based on the standard data type and standard data format under the preset dimension corresponding to the blank data.

[0015] In a possible implementation, the target feature extraction model includes a target time series model and a target general large model, and the extraction module is specifically configured to: Determining a first target feature dimension for extracting features from the network data set by the target time series model, performing feature extraction on the network data set based on the first target feature dimension, and obtaining time series feature data of the network data set; A second target feature dimension for extracting features from the network data set by the target universal macromodel is determined, and features are extracted from the network data set based on the second target feature dimension to obtain universal feature data of the network data set.

[0016] In a possible implementation, the prediction module is specifically configured to: Determining a target sub-link and network attributes corresponding to the network failure, and generating a corresponding avoidance strategy based on the network failure; After degrading the target sub-link and network attributes of the network failure, the network failure is avoided based on the avoidance strategy.

[0017] In a possible implementation, the feature processing model includes a fully connected layer, and the device further includes: A determination module, configured to determine the task complexity of a preset network quality prediction task and the dimension of a hidden layer of the fully connected layer; An adjustment module is used to adjust the dimension of the hidden layer of the fully connected layer based on the task complexity.

[0018] In a possible implementation, the prediction module is further specifically configured to: Before inputting the target feature data into a preset feature processing model, entering the feature data into a preset long-term storage vector database; Based on the first target feature dimension of the time series feature data and the second target feature dimension of the universal feature data, feature fusion is performed on the time series feature data and the universal feature data.

[0019] In a possible implementation, the device further includes: A display module, configured to display the feature data and the network prediction results based on a preset visualization platform; An alarm module is used to set respective alarm thresholds for the delay and the packet loss rate, and to issue an alarm when the network prediction result meets the alarm threshold.

[0020] In a third aspect, an embodiment of the present application provides an electronic device comprising: a processor, a storage medium and a bus, wherein the storage medium stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of the wide area network network quality prediction method as described in any one of the first aspects.

[0021] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the wide area network network quality prediction method described in any one of the first aspects are executed.

[0022] The embodiments of the present application provide a network quality prediction method, apparatus, device, and medium for a wide area network. Based on pre-stored historical network data from a target user end to a target server end, multiple target sub-links in a communication link are determined, and the network quality of the target sub-links in multiple preset dimensions is detected by a detection client to obtain historical network quality detection data of the target sub-links in multiple preset dimensions. The historical network quality detection data is regularized based on a preset regularization rule, a set of IP addresses of a global wide area network is collected, the IP address set and the historical network quality detection data are matched and fused to obtain a network data set, and the network data set is classified based on a preset classification rule to obtain a classified network data set. A target feature extraction model for the network data set is determined and selected based on the data type of the classified network data set. Multi-dimensional feature data corresponding to the network data set is extracted based on the target feature extraction model. After the multi-dimensional feature data is fused to obtain target feature data, the target feature data is input into a preset feature processing model. Corresponding network prediction results are output for network attributes of different dimensions. Based on the network prediction results, possible network failures in the future are determined, and network failures are avoided. This application determines multiple target sub-links in a communication link based on historical network conditions, and performs network quality detection on the target sub-links under multiple preset dimensions. The historical network quality detection data and the collected wide area network IP address set are vectorized through a target feature extraction model, and input into a feature processing model for network quality prediction. This can predict future network failures in the wide area network and avoid them in advance, fundamentally preventing the occurrence of secondary problems that affect network quality, avoiding the limitations of post-avoidance avoidance, and improving network quality and user experience. At the same time, through early prediction and avoidance, higher efficiency and reliability can be achieved in wide area network management.

[0023] In order to make the above-mentioned objects, features and advantages of the present application 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

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0025] Figure 1 is a flow chart of a method for predicting network quality of a wide area network according to an embodiment of the present application; Figure 2 It is a schematic diagram of the network quality prediction process; Figure 3 is a schematic structural diagram of a network quality prediction device for a wide area network according to an embodiment of the present application; Figure 4 It is a structural diagram of a computer device according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the drawings in the present application only serve the purpose of illustration and description and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowcharts can be implemented out of sequence, and steps without logical context can be reversed or implemented simultaneously. In addition, those skilled in the art, under the guidance of the contents of this application, can add one or more other operations to the flowchart, or remove one or more operations from the flowchart.

[0027] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present application.

[0028] It should be noted that the term "comprising" will be used in the embodiments of the present application to indicate the existence of the features declared thereafter, but does not exclude the addition of other features.

[0029] Considering that the WAN is composed of multiple complex components and infrastructure, including cell egress, regional aggregation switches, ISPs, AS number broadcasts, submarine and terrestrial cables, cross-border lines, and IXP switching centers, these components work together to transmit a data packet from one end to the other. However, the WAN suffers from high jitter and poor stability, coupled with secondary issues caused by external factors such as natural disasters and construction, making network quality difficult to guarantee.

[0030] Currently, these inherent secondary issues are primarily addressed by scheduling egress and path scheduling after they occur. However, existing methods, which only address issues after they occur, have limitations and are difficult to prevent fundamentally. Furthermore, even with appropriate scheduling over long WAN links, network quality can still degrade for a period of time, impacting user experience.

[0031] To address this issue, the present application provides a wide area network (WAN) network quality prediction method, apparatus, device, and medium. This method determines multiple target sub-links in a communication link based on historical network conditions, and performs network quality detection on the target sub-links under multiple preset dimensions. The historical network quality detection data and the collected WAN IP address set are vectorized using a target feature extraction model, and input into a feature processing model for network quality prediction. This method can predict future WAN network failures and avoid them in advance, fundamentally preventing the occurrence of secondary problems that affect network quality, avoiding the limitations of post-event avoidance, and improving network quality and user experience. At the same time, through advance prediction and avoidance, higher efficiency and reliability can be achieved in WAN management.

[0032] Figure 1 FIG. 1 is a flow chart of a method for predicting network quality of a wide area network according to an embodiment of the present application. Figure 1 As shown, the network quality prediction method of the wide area network in the embodiment of the present application may specifically include: S101. Based on pre-stored historical network data from a target user end to a target server end, multiple target sub-links in a communication link are determined, and the network quality of the target sub-links under multiple preset dimensions is detected by a detection client to obtain historical network quality detection data of the target sub-links under multiple preset dimensions, and the historical network quality detection data is regularized based on preset regularization rules.

[0033] S102: Collect IP address sets of the global wide area network, match and fuse the IP address sets with historical network quality detection data to obtain a network data set, and classify the network data set based on preset classification rules to obtain a classified network data set.

[0034] S103 , determining and selecting a target feature extraction model for the network data set based on the data type of the classified network data set, and extracting multi-dimensional feature data corresponding to the network data set based on the target feature extraction model.

[0035] S104. After the multi-dimensional feature data are fused to obtain target feature data, the target feature data is input into a preset feature processing model, and corresponding network prediction results are output for network attributes of different dimensions. Based on the network prediction results, possible network failures in the future are determined and avoided.

[0036] In the aforementioned WAN network quality prediction method, multiple target sub-links within a communication link are identified based on historical network conditions. These target sub-links are then subjected to network quality detection under multiple preset dimensions. A target feature extraction model vectorizes the historical network quality detection data and a collection of WAN IP addresses, which are then fed into a feature processing model for network quality prediction. This method can predict future WAN network failures and proactively mitigate them, fundamentally preventing the occurrence of secondary issues that could impact network quality. This avoids the limitations of post-event mitigation and improves network quality and user experience. Furthermore, proactive prediction and mitigation can achieve greater efficiency and reliability in WAN management.

[0037] It should be noted that the network quality prediction method of the wide area network of the present application is used to detect the network quality of the communication link from the target user end to the target server end. The target user end and the target server end include multiple data exchange components. Correspondingly, the communication link includes multiple sub-links, and each sub-link includes any two nodes between the target user end and the target server end.

[0038] The above exemplary steps of the embodiment of the present application are described below with reference to specific examples: S101, based on the pre-stored historical network data from the target user end to the target server end, determine multiple target sub-links in the communication link, and detect the network quality of the target sub-links under multiple preset dimensions through the detection client, obtain the historical network quality detection data of the target sub-links under multiple preset dimensions, and regularize the historical network quality detection data based on preset regularization rules.

[0039] It should be noted that a communication link is also a communication path, and a sub-link is also a sub-path between any two nodes included in the communication path.

[0040] Specifically, the historical network data includes at least historical delays and historical packet loss rates. The historical network data is pre-stored historical network data, for example, stored in a relevant database; the target user end (User end) is the user end that initiates the network quality detection, and the target server end (Server end) is the server end of the network quality detection clock. For example, if a network quality detection is initiated from Beijing to New York, Beijing represents the target user end, and New York represents the target server end; the target sub-link is the target sub-link selected from multiple sub-links of the communication link from the target user end to the target server end, and the detection client is the client used for network quality detection, such as an agent. Multiple preset dimensions include at least source address, destination address, delay, packet loss rate, and intermediate path nodes; the source address corresponds to the user end, and the destination address corresponds to the server end; the regularization rules are pre-set rules for organizing historical network quality detection data.

[0041] In the embodiment of the present application, multiple target sub-links in the communication link are determined based on the historical network data from the target user end to the target server end, and the network quality detection of the target sub-link under multiple preset dimensions is performed on the target sub-link by the detection client to obtain the historical network quality detection data of the target sub-link under multiple preset dimensions, and the historical network quality detection data is regularized according to the regularization rules for subsequent processing. For example, Figure 2 As shown, it can be divided into data layer, model layer and application layer, among which, in the data layer, network quality detection data is obtained.

[0042] Optionally, when regularizing historical network quality detection data, determine the standard data type and standard data format of each preset dimension in the historical network quality detection data, and compare them with the data type and data format of the data of each preset dimension in the historical network quality detection data to filter out useless data and blank data in the historical network quality detection data; clean the useless data, and fill the blank data based on the standard data type and standard data format under the preset dimension corresponding to the blank data.

[0043] Among them, useless data represents data in historical network quality detection data whose data format is inconsistent with the standard data format, and / or data type is inconsistent with the standard data type. For example, if the standard data type of a data is a positive number, then the negative number is useless data; if the standard data format of a data is one decimal place, then the data with two decimal places is useless data; blank data represents data in a preset dimension in historical network quality detection data that is empty.

[0044] Specifically, the network quality detection data set is normalized according to normalization rules, including the source address, destination address, latency, packet loss rate, and intermediate path nodes of the detection data. Useless data is then removed and blank data is filled in. For example, when filling blank data, numbers such as 0 or 1 can be used to ensure normal vectorization of subsequent data.

[0045] It should be noted that the data is preprocessed, for example, missing blanks, inconsistent formats, inconsistent values, etc. are processed, so that all contents of the entire data are preprocessed into a state that can be analyzed.

[0046] S102, collecting IP address sets of the global wide area network, matching and fusing the IP address sets with historical network quality detection data to obtain a network data set, and classifying the network data set based on preset classification rules to obtain a classified network data set.

[0047] Optionally, the IP address set of the global wide area network is collected based on a preset IP address database. The IP address database can be a public database or a paid professional database. The IP address set of the global wide area network is collected through these databases.

[0048] In the embodiment of the present application, the IP address set includes at least a public IP address set and an Ixp Peer interconnected address set; the classification rules include at least specific rules of the IP address, physical space attributes, ISP attributes, and path attributes; the physical space attributes include at least geographical location, distance, and regional coverage; the ISP attributes include at least operator type, service quality, and bandwidth; the IP address set includes the IP address information of the target sub-link in the historical network quality detection data, that is, the specific IP address information in the historical network quality detection data can be determined from the IP address set, and the network data set includes multiple IP address subsets after the IP address set is classified and the classified historical network quality detection data; collect and organize the global wide area network public IP address set and the Ixp Peer interconnected address set, and match the IP address set with the historical network quality detection data obtained in step S101 to obtain a network data set, and classify the network data set based on the classification rules to obtain the classified network data set, that is, the multiple IP address subsets after the IP address set is classified and the classified historical network quality detection data for subsequent processing. For example, Figure 2 As shown, the original detection data is the network data set.

[0049] Therefore, the IP address set serves as the basis for subsequent classification and detection. By obtaining comprehensive and accurate IP address information, the accuracy and reliability of subsequent classification and detection can be improved.

[0050] It should be noted that when classifying network data sets, historical network quality detection data can be subdivided according to specific rules (for example), physical space attributes (such as geographic location, distance, regional coverage, etc.) and ISP attributes (such as operator type, service quality, bandwidth, etc.). In addition, IP address sets can be classified according to categories such as country, continent, ISP, etc. to obtain multiple subsets, thereby forming a more refined data set to provide data support and samples for subsequent network quality predictions.

[0051] S103 , determining and selecting a target feature extraction model for the network data set based on the data type of the classified network data set, and extracting multi-dimensional feature data corresponding to the network data set based on the target feature extraction model.

[0052] In the embodiment of the present application, the data type of the network data set, for example, country, delay, packet loss rate, etc., the target feature extraction model is a model for extracting features from the network data set, and the target feature extraction model includes a target time series model and a target general large model, wherein the target time series model is such as Transformer, RNN, CNN-LSTM, etc., and the target general large model is an existing general large model, such as GPT, Deepseek, etc. Based on the data type of the network data set, the target feature extraction model for the network data set is selected, and the multi-dimensional feature data corresponding to the network data set is extracted according to the target feature extraction model. For example, Figure 2 As shown in Figure 1, feature engineering, i.e. feature extraction, is performed on the network data set.

[0053] It should be noted that a target feature extraction model for a network data set can be selected based on the data type and a preset prediction direction. For example, a prediction direction could be latency, i.e., latency prediction. Alternatively, historical network quality detection data can be used to determine historical network failures, and corresponding prediction directions can be determined based on these historical network failures. For example, if historical network quality detection data determines that a range of network delays exist across various dimensions, latency prediction can be performed subsequently.

[0054] Optionally, when extracting multi-dimensional feature data corresponding to the network data set based on the target feature extraction model, a first target feature dimension is determined for feature extraction of the network data set by the target time series model, and feature extraction is performed on the network data set based on the first target feature dimension to obtain time series feature data of the network data set; and a second target feature dimension is determined for feature extraction of the network data set by the target universal large model, and feature extraction is performed on the network data set based on the second target feature dimension to obtain universal feature data of the network data set. The determination of the feature dimension depends on the requirements for feature data and prediction accuracy and can be determined based on actual circumstances.

[0055] Specifically, feature extraction is performed on the above network data set through the target time series model, and the extraction dimension is the first target feature dimension, such as 1024. Feature extraction is performed on the above network data set through the target general large model, and the extraction dimension is the second target feature dimension, such as 2048. In this way, the feature extraction of the network data set is completed and the data is vectorized.

[0056] S104, after fusing the multi-dimensional feature data to obtain the target feature data, the target feature data is input into a preset feature processing model, and corresponding network prediction results are output for network attributes of different dimensions. Based on the network prediction results, possible network failures in the future are determined and avoided.

[0057] In the embodiment of the present application, the feature processing model is a pre-set model for processing feature data, such as the above-mentioned time series model, general large model or other models. The network attributes of different dimensions include at least country, region, ISP, AS number, submarine cable system, etc. The prediction results represent the prediction results under network attributes of different dimensions, such as the results under attributes such as country, region, ISP, AS number, submarine cable system, etc. After the multi-dimensional feature data obtained in step S103 is subjected to feature fusion to obtain target feature data, the target feature data is input into the feature processing model. The feature processing model outputs the corresponding network prediction results for network attributes of different dimensions, determines the possible network failures in the future based on the network prediction results, and avoids the network failures. For example, Figure 2 As shown in the figure, the feature data extracted from the time series model and the general large model are fused and the prediction results are output, such as ISP prediction, AS number prediction, and region prediction.

[0058] Optionally, when fusing multi-dimensional feature data, the feature data is entered into a preset long-term storage vector database; based on the first target feature dimension of the time series feature data and the second target feature dimension of the general feature data, the time series feature data and the general feature data are feature fused.

[0059] Specifically, after extracting the feature data, the extracted feature data is entered into a preset long-term storage vector database for vectorized long-term storage, which is used for analyzing historical data and making predictions about the future, as well as fusing the features extracted from the time series model and the general large model in terms of vector dimensions. For example, the time series feature vector dimension of 384 is fused with the feature dimension of 1024 extracted from the general large model.

[0060] Optionally, when mitigating a network failure, the target sub-link and network attributes corresponding to the network failure are determined, and a corresponding mitigation strategy is generated based on the network failure. After degrading the target sub-link and network attributes of the network failure, the network failure is avoided based on the mitigation strategy. The network attributes are the attributes of the target sub-link corresponding to the network failure, i.e., the country, AS number, ISP number, etc. where the network failure occurred. The mitigation strategy is a strategy for mitigating future network failures, such as dynamically scheduling traffic paths or adjusting data transmission priorities.

[0061] It should be noted that after a predicted failure occurs, the back-end automated program will be used to dispatch problematic paths, AS numbers, submarine cable systems, etc. to avoid the problem.

[0062] This allows us to predict the possibility of network quality degradation in advance and generate avoidance strategies to prevent network problems from affecting user experience.

[0063] The network quality prediction method for a wide area network provided in an embodiment of the present application determines multiple target sub-links in a communication link based on pre-stored historical network data from a target user end to a target server end, and detects the network quality of the target sub-links in multiple preset dimensions through a detection client to obtain historical network quality detection data of the target sub-links in multiple preset dimensions, and regularizes the historical network quality detection data based on a preset regularization rule, collects a set of IP addresses of a global wide area network, matches and fuses the IP address set and the historical network quality detection data to obtain a network data set, and classifies the network data set based on a preset classification rule to obtain a classified network data set, determines and selects a target feature extraction model for the network data set based on the data type of the classified network data set, and extracts multi-dimensional feature data corresponding to the network data set based on the target feature extraction model; after fusing the multi-dimensional feature data to obtain target feature data, the target feature data is input into a preset feature processing model, and corresponding network prediction results are output for network attributes of different dimensions; based on the network prediction results, possible network failures in the future are determined, and network failures are avoided. The network quality prediction method for a wide area network (WAN) disclosed in this application determines multiple target sub-links in a communication link based on historical network conditions, and performs network quality detection on the target sub-links under multiple preset dimensions. The historical network quality detection data and the collected WAN IP address set are vectorized using a target feature extraction model, and input into a feature processing model for network quality prediction. This method can predict future network failures in the WAN and avoid them in advance, fundamentally preventing the occurrence of secondary problems that affect network quality, avoiding the limitations of post-event avoidance, and improving network quality and user experience. At the same time, through advance prediction and avoidance, higher efficiency and reliability can be achieved in WAN management.

[0064] Furthermore, the feature processing model includes a fully connected layer; determining a task complexity of a preset network quality prediction task and a dimension of a hidden layer of the fully connected layer; and adjusting the dimension of the hidden layer of the fully connected layer based on the task complexity.

[0065] Specifically, the dimension of the hidden layer of the fully connected layer is adjusted according to the task complexity of the network quality prediction task. When the task complexity is high, the dimension of the hidden layer of the fully connected layer is increased. For example, Figure 2 As shown in the figure, the target feature data after feature fusion is input into the fully connected layer of the feature processing model for processing, and the prediction result is output.

[0066] Furthermore, the feature data and network prediction results are displayed on a pre-set visualization platform. Alarm thresholds are set for latency and packet loss rate, and an alarm is triggered when the network prediction results meet the thresholds. For example, if the packet loss rate alarm threshold is set to 0.01, an alarm will be triggered if it exceeds 0.01.

[0067] Specifically, after the above-mentioned data feature extraction, data calculation analysis and prediction, the generated results are displayed through a visualization platform, and alarm thresholds for delay and packet loss rate are given to issue an alarm when the corresponding data exceeds the threshold.

[0068] Therefore, after extracting data features through time series models and general large-scale models, this application performs large-scale model analysis on the data, accurately predicting attributes in different dimensions, and achieving prediction of future WAN network failures. This upgrades the previous post-failure avoidance to pre-failure avoidance, thereby optimizing the problem of poor Internet application quality caused by WAN failures. On the pure WAN side, by avoiding WAN network failures in advance, high-quality, imperceptible, and stable coverage can be achieved for users using the Internet.

[0069] Figure 3 FIG. 1 is a flow chart of a network quality prediction device for a wide area network according to an embodiment of the present application. Figure 3 As shown, the network quality prediction device 300 of the wide area network of the embodiment of the present application is used to detect the network quality of the communication link between the target user end and the target server end. The target user end and the target server end include multiple data exchange components. Correspondingly, the communication link includes multiple sub-links, and each sub-link includes any two nodes between the target user end and the target server end. Specifically, it may include: The detection module 301 is used to determine multiple target sub-links in the communication link based on pre-stored historical network data from the target user end to the target server end, and detect the network quality of the target sub-links under multiple preset dimensions through the detection client, obtain historical network quality detection data of the target sub-links under multiple preset dimensions, and regularize the historical network quality detection data based on preset regularization rules; wherein the multiple preset dimensions include at least source address, destination address, delay, packet loss rate, and intermediate path nodes.

[0070] The classification module 302 is used to collect IP address sets of the global wide area network, match and fuse the IP address sets with historical network quality detection data to obtain a network data set, and classify the network data set based on preset classification rules to obtain a classified network data set; wherein the IP address set includes the IP address information of the target sub-link in the historical network quality detection data, and the network data set includes multiple IP address subsets after the IP address set is classified and the classified historical network quality detection data.

[0071] The extraction module 303 is used to determine and select a target feature extraction model for the network data set based on the data type of the classified network data set, and extract multi-dimensional feature data corresponding to the network data set based on the target feature extraction model.

[0072] The prediction module 304 is used to input the target feature data into a preset feature processing model after fusing the multi-dimensional feature data to obtain the target feature data, output corresponding network prediction results for network attributes of different dimensions, determine possible network failures in the future based on the network prediction results, and avoid network failures; wherein the prediction results represent the prediction results under network attributes of different dimensions.

[0073] In a possible implementation, the detection module is specifically configured to: Determine the standard data type and standard data format of each preset dimension in the historical network quality detection data, and compare them with the data type and data format of each preset dimension in the historical network quality detection data to filter out useless data and blank data in the historical network quality detection data; Clean useless data and fill blank data based on the standard data type and standard data format under the preset dimension corresponding to the blank data.

[0074] In one possible implementation, the target feature extraction model includes a target time series model and a target general large model, and an extraction module is specifically used to: Determine a first target feature dimension for feature extraction of the target time series model for the network data set, perform feature extraction on the network data set based on the first target feature dimension, and obtain time series feature data of the network data set; A second target feature dimension for feature extraction of the target universal large model for the network data set is determined, and feature extraction is performed on the network data set based on the second target feature dimension to obtain universal feature data of the network data set.

[0075] In a possible implementation, the prediction module is specifically configured to: Determine the target sub-links and network attributes corresponding to the network failure, and generate corresponding avoidance strategies based on the network failure; After degrading the target sub-links and network attributes of the network failure, the network failure is avoided based on the avoidance strategy.

[0076] In one possible implementation, the feature processing model includes a fully connected layer, and the device further includes: A determination module, used to determine the task complexity of a preset network quality prediction task and the dimension of the hidden layer of the fully connected layer; The adjustment module is used to adjust the dimensions of the hidden layers of the fully connected layer based on the task complexity.

[0077] In a possible implementation, the prediction module is further specifically configured to: Before inputting the target feature data into the preset feature processing model, the feature data is entered into a preset long-term storage vector database; Based on the first target feature dimension of the time series feature data and the second target feature dimension of the universal feature data, feature fusion is performed on the time series feature data and the universal feature data.

[0078] In a possible implementation, the device further includes: The display module is used to display feature data and network prediction results based on a preset visualization platform; The alarm module is used to set respective alarm thresholds for delay and packet loss rate, and to issue an alarm when the network prediction result meets the alarm threshold.

[0079] The network quality prediction device for a wide area network provided in an embodiment of the present application determines multiple target sub-links in a communication link based on pre-stored historical network data from a target user end to a target server end, and detects the network quality of the target sub-links in multiple preset dimensions through a detection client to obtain historical network quality detection data of the target sub-links in multiple preset dimensions, and regularizes the historical network quality detection data based on preset regularization rules, collects a set of IP addresses of a global wide area network, matches and fuses the IP address set and the historical network quality detection data to obtain a network data set, and classifies the network data set based on preset classification rules to obtain a classified network data set, determines and selects a target feature extraction model for the network data set based on the data type of the classified network data set, and extracts multi-dimensional feature data corresponding to the network data set based on the target feature extraction model. After fusing the multi-dimensional feature data to obtain target feature data, the target feature data is input into a preset feature processing model, and corresponding network prediction results are output for network attributes of different dimensions. Based on the network prediction results, possible network failures in the future are determined, and network failures are avoided. The network quality prediction device for a wide area network (WAN) of the present application determines multiple target sub-links in a communication link based on historical network conditions, and performs network quality detection on the target sub-links under multiple preset dimensions. The historical network quality detection data and the collected WAN IP address set are vectorized using a target feature extraction model, and input into a feature processing model for network quality prediction. This device can predict future network failures in the WAN and avoid them in advance, fundamentally preventing the occurrence of secondary problems that affect network quality, avoiding the limitations of post-event avoidance, and improving network quality and user experience. At the same time, through advance prediction and avoidance, higher efficiency and reliability can be achieved in WAN management.

[0080] like Figure 4 As shown, an electronic device 400 provided in an embodiment of the present application includes: a processor 401, a memory 402 and a bus, wherein the memory 402 stores machine-readable instructions executable by the processor 401. When the electronic device is running, the processor 401 communicates with the memory 402 through the bus, and the processor 401 executes the machine-readable instructions to perform the steps of the network quality prediction method for the wide area network as described above.

[0081] Specifically, the memory 402 and processor 401 can be general-purpose memories and processors, which are not specifically limited here. When the processor 401 runs the computer program stored in the memory 402, it can execute the wide area network quality prediction method.

[0082] Corresponding to the above-mentioned wide area network network quality prediction method, an embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned wide area network network quality prediction method are executed.

[0083] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system and device described above can refer to the corresponding process in the method embodiment, and will not be repeated in this application. In the several embodiments provided in this application, it should be understood that the disclosed system, device and method can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0084] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed across multiple network elements. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0085] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0086] If the functions are implemented as software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of this application, or the portion that contributes to the prior art, or the portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the deployment method described in each embodiment of this application. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, ROM, RAM, a magnetic disk, or an optical disk.

[0087] The above are only specific embodiments of the present application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for predicting network quality of a wide area network, characterized in that: The method is used to detect the network quality of a communication link between a target user terminal and a target server terminal, wherein the target user terminal and the target server terminal include multiple data exchange components, and correspondingly, the communication link includes multiple sub-links, each sub-link includes any two nodes between the target user terminal and the target server terminal; the method includes: Based on pre-stored historical network data from a target user end to a target server end, multiple target sub-links in the communication link are determined, and the network quality of the target sub-links under multiple preset dimensions is detected by a detection client to obtain historical network quality detection data of the target sub-links under the multiple preset dimensions, and the historical network quality detection data is normalized based on a preset normalization rule; wherein the multiple preset dimensions include at least source address, destination address, delay, packet loss rate, and intermediate path nodes; Collecting a set of IP addresses of a global wide area network, matching and fusing the set of IP addresses with the historical network quality detection data to obtain a network data set, and classifying the network data set based on a preset classification rule to obtain a classified network data set; wherein the set of IP addresses includes the IP address information of the target sub-link in the historical network quality detection data, and the network data set includes a plurality of IP address subsets obtained by classifying the set of IP addresses and the classified historical network quality detection data; Determine and select a target feature extraction model for the network data set based on the data type of the classified network data set, and extract multi-dimensional feature data corresponding to the network data set based on the target feature extraction model; After the multi-dimensional feature data are fused to obtain target feature data, the target feature data is input into a preset feature processing model, and corresponding network prediction results are output for network attributes of different dimensions. Based on the network prediction results, possible network failures in the future are determined, and the network failures are avoided; wherein the prediction results represent the prediction results under network attributes of different dimensions.

2. The method according to claim 1, characterized in that The regularizing of the historical network quality detection data includes: Determining a standard data type and a standard data format for each preset dimension in the historical network quality detection data, and comparing the data types and data formats with the data for each preset dimension in the historical network quality detection data, to filter out useless data and blank data in the historical network quality detection data; The useless data is cleaned, and the blank data is filled based on the standard data type and standard data format under the preset dimension corresponding to the blank data.

3. The method according to claim 2, characterized in that The target feature extraction model includes a target time series model and a target general large model. The multi-dimensional feature data corresponding to the network data set is extracted based on the target feature extraction model, including: Determining a first target feature dimension for extracting features from the network data set by the target time series model, performing feature extraction on the network data set based on the first target feature dimension, and obtaining time series feature data of the network data set; A second target feature dimension for extracting features from the network data set by the target universal macromodel is determined, and features are extracted from the network data set based on the second target feature dimension to obtain universal feature data of the network data set.

4. The method according to claim 3, characterized in that The avoiding the network failure includes: Determining a target sub-link and network attributes corresponding to the network failure, and generating a corresponding avoidance strategy based on the network failure; After degrading the target sub-link and network attributes of the network failure, the network failure is avoided based on the avoidance strategy.

5. The method according to claim 4, characterized in that The feature processing model includes a fully connected layer, and the method further includes: Determining the task complexity of a preset network quality prediction task and the dimension of the hidden layer of the fully connected layer; The dimension of the hidden layer of the fully connected layer is adjusted based on the task complexity.

6. The method according to claim 5, characterized in that Before inputting the target feature data into a preset feature processing model, fusing the multi-dimensional feature data includes: Entering the feature data into a preset long-term storage vector database; Based on the first target feature dimension of the time series feature data and the second target feature dimension of the universal feature data, feature fusion is performed on the time series feature data and the universal feature data.

7. The method according to claim 6, characterized in that The method further comprises: Displaying the feature data and the network prediction results based on a preset visualization platform; Respective alarm thresholds are set for the delay and the packet loss rate, and an alarm is triggered when the network prediction result meets the alarm threshold.

8. A network quality prediction device for a wide area network, characterized in that: The device is used to detect the network quality of a communication link between a target user terminal and a target server terminal, wherein the target user terminal and the target server terminal include multiple data exchange components, and correspondingly, the communication link includes multiple sub-links, each sub-link includes any two nodes between the target user terminal and the target server terminal; the device includes: A detection module is configured to determine multiple target sub-links in a communication link based on pre-stored historical network data from a target user end to a target server end, detect the network quality of the target sub-links under multiple preset dimensions through a detection client, obtain historical network quality detection data of the target sub-links under the multiple preset dimensions, and normalize the historical network quality detection data based on preset normalization rules; wherein the multiple preset dimensions include at least source address, destination address, delay, packet loss rate, and intermediate path nodes; a classification module for collecting a set of IP addresses of a global wide area network, matching and fusing the set of IP addresses with the historical network quality detection data to obtain a network data set, and classifying the network data set based on a preset classification rule to obtain a classified network data set; wherein the set of IP addresses includes the IP address information of the target sub-link in the historical network quality detection data, and the network data set includes a plurality of IP address subsets obtained by classifying the set of IP addresses and the classified historical network quality detection data; an extraction module, configured to determine and select a target feature extraction model for the network data set based on the data type of the classified network data set, and extract multi-dimensional feature data corresponding to the network data set based on the target feature extraction model; The prediction module is used to input the target feature data into a preset feature processing model after fusing the multi-dimensional feature data to obtain the target feature data, output corresponding network prediction results for network attributes of different dimensions, determine possible network failures in the future based on the network prediction results, and avoid the network failures; wherein the prediction results represent the prediction results under network attributes of different dimensions.

9. An electronic device, characterized in that: include: A processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate via the bus. When the machine-readable instructions are executed by the processor, the steps of the wide area network quality prediction method according to any one of claims 1 to 7 are performed.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, executes the steps of the method for predicting network quality of a wide area network according to any one of claims 1 to 7.

Images