Artificial intelligence prediction performance and dynamic adjustment method and system for 5g virtual private network

Abstract

The application relates to the technical field of data processing, and discloses an artificial intelligence prediction performance and dynamic adjustment method and system for a 5G virtual private network, which comprises the following steps: performing data standardization processing operation according to original index data of target network equipment and network key index types to obtain standardized index data; performing performance prediction operation according to the standardized index data to obtain network future performance prediction results; and performing routing weight value configuration optimization operation on a routing transmission path according to the network future performance prediction results to obtain final routing weight results, so as to realize network traffic balanced distribution and scheduling adjustment. It can be seen that the application can improve the standardization processing accuracy and reliability of index data, and then improve the prediction accuracy and prediction reliability of network future performance, so as to improve the adjustment accuracy and adjustment reliability of routing weight, and further realize automatic rebalancing of network traffic and improve the load balancing and transmission efficiency of the whole network.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for predicting and dynamically adjusting the performance of a 5G virtual private network using artificial intelligence. Background Technology

[0002] With the large-scale commercial deployment of fifth-generation mobile communication technology (5G) and the accelerated digital transformation of vertical industries, 5G industry virtual private networks (VPNs) play a crucial role in several key areas as the core link connecting industry users and 5G public networks. In the actual operation of 5G industry VPNs, routing strategies are the key factors that determine the packet forwarding path and transmission performance.

[0003] Currently, existing 5G industry virtual private networks (VPNs) generally rely on pre-configured, fixed routing paths to allocate dedicated transmission channels for industry services. Their operation depends entirely on a pre-defined network topology and fixed resource configuration, and they can only function normally in scenarios with stable network environments and constant service traffic. However, fixed routing paths cannot cope with dynamic changes in network topology. When base station switching, link failures, or UPF migrations cause topology changes, fixed paths continue to forward data along failed nodes, easily leading to transmission interruptions. Furthermore, fixed routing paths struggle to adapt to the dynamic characteristics of sudden traffic surges in industry services. When actual traffic far exceeds the pre-defined limits, the link bandwidth and node processing capacity on the path quickly become saturated, leading to network congestion, packet loss, or sudden increases in transmission latency. Clearly, existing technologies have shortcomings that urgently need to be addressed. Summary of the Invention

[0004] This invention provides an artificial intelligence-based performance prediction and dynamic adjustment method and system for 5G virtual private networks. It can realize intelligent network performance prediction and routing weight adjustment, improve the accuracy and reliability of network performance status prediction, improve the accuracy and reliability of routing weight adjustment, and thus realize automatic rebalancing of network traffic and improve the overall network load balance and transmission efficiency.

[0005] To address the aforementioned technical problems, the first aspect of this invention discloses a method for predicting and dynamically adjusting the performance of a 5G virtual private network using artificial intelligence, the method comprising: Based on the collected raw indicator data for the target network devices and key network indicator types, perform the corresponding data standardization processing operations to obtain standardized indicator data. Based on the standardized index data, perform the corresponding performance prediction operation to obtain the network future performance prediction results; Based on the network's future performance prediction results, corresponding routing weight configuration optimization operations are performed on the routing transmission paths to obtain the final routing weight results, so as to achieve balanced distribution and scheduling adjustment of network traffic.

[0006] As an optional implementation, in the first aspect of the present invention, the step of performing a corresponding performance prediction operation based on the standardized index data to obtain a future network performance prediction result includes: Based on the standardized indicator data and the supplementary historical performance indicator data, the target data set is determined; Based on the target dataset, perform the corresponding two-parameter exponential average prediction operation to obtain the future performance trend prediction results; Based on the future performance trend prediction results, corresponding network status and performance level prediction operations are performed to obtain the future network performance prediction results, which include the future network performance status and the predicted values ​​of the network key indicator types. And, the step of performing corresponding network state and performance level prediction operations based on the future performance trend prediction results to obtain future network performance prediction results includes: Based on the future performance trend prediction results, determine the first predicted value information of the network key indicator type for the first target time period, and determine the first actual value information of the network key indicator type for the first target time period. Based on the first actual value information and the first predicted value information, perform corresponding absolute error calculation and performance error state division operations to obtain at least two performance error state schemes. Each performance error state scheme includes a performance error state and its corresponding absolute error interval. Based on the determined network performance error state data for the second target time period and the performance error state scheme, the state transition probability matrix result is determined; Based on the predicted future performance trends and the state transition probability matrix, corresponding fusion prediction and error analysis operations are performed to obtain the predicted future performance of the network.

[0007] As an optional implementation, in the first aspect of the present invention, the step of performing a corresponding two-parameter exponential average prediction operation based on the target data set to obtain a future performance trend prediction result includes: Based on the target dataset, determine the network performance data characteristics; Based on the network performance data characteristics and the determined network performance transmission demand information, a first value bias range for the horizontal smoothing parameter and a second value bias range for the trend smoothing parameter are determined. Based on the first value bias range, the second value bias range, and the determined prediction result accuracy requirement information, the optimal parameter combination is determined. The optimal parameter combination includes the first optimal value of the horizontal smoothing parameter and the second optimal value of the trend smoothing parameter. Based on the target data set, determine the actual observed value of the network key indicator type at the first current moment, and based on the actual observed value of the network key indicator type, the first optimal value, and the second optimal value, determine the future performance trend prediction curve of the network key indicator type. Based on the future performance trend prediction curves of the aforementioned key network indicator types, the future performance trend prediction results are determined.

[0008] As an optional implementation, in the first aspect of the present invention, determining the state transition probability matrix result based on the determined network performance error state data for the second target time period and the performance error state scheme includes: Perform corresponding state sequence generation operations on the determined network performance error state data for the second target time period to obtain at least two state sequence results. Each state sequence result includes the target transition time and its corresponding state transition relationship. Based on all the state sequence results, determine the total number of transitions for each performance error state that satisfies the preset initial state condition, and based on all the state sequence results, determine the specific number of transitions for each state transition relationship. Based on the specific number of transitions in each state transition relationship, the transition probability value of each state transition relationship is determined. The state transition probability matrix result is determined based on the number of state types corresponding to all the performance error states and the transition probability value of each state transition relationship.

[0009] As an optional implementation, in the first aspect of the present invention, the step of performing corresponding fusion prediction and error analysis operations based on the future performance trend prediction results and the state transition probability matrix results to obtain the network future performance prediction results includes: Based on the future performance trend prediction results, determine the second predicted value of the network key indicator type for the second current time, and determine the second actual value of the network key indicator type for the second current time; Based on the second predicted value and the second actual value, calculate the first difference, and based on the first difference and all the performance error state schemes, determine the first actual performance error state of the network key indicator type for the second current time. Based on the first actual performance error state and the state transition probability matrix result, the second predicted performance error state of the network key indicator type for the third future time is determined, and based on the second predicted value and the first error correction coefficient of the determined first actual performance error state, the third predicted value of the network key indicator type for the third future time is determined. Based on the network key indicator type, the third predicted value and the second predicted performance error state for the third future time are used to determine the network future performance prediction result for the third future time. Furthermore, the method further includes: Determine the third actual value of the network key indicator type for the third future time, and calculate the second difference based on the third predicted value and the third actual value; Based on the second difference and the second prediction performance error state, it is determined whether the network key indicator type meets the preset performance direct prediction conditions for the fourth future time. When it is determined that the network key indicator type meets the performance direct prediction condition, the second predicted performance error state is determined as the third actual performance error state of the network key indicator type for the third future time. When it is determined that the network key indicator type does not meet the direct performance prediction condition, the third actual performance error state of the network key indicator type for the third future time is determined based on the second difference and all the performance error state schemes. Based on the third actual performance error state and the state transition probability matrix result, the fifth predicted performance error state of the network key indicator type for the fourth future time is determined, and based on the third predicted value and the third error correction coefficient of the determined third actual performance error state, the fourth predicted value of the network key indicator type for the fourth future time is determined. Based on the fourth predicted value for the fourth future time step and the fifth predicted performance error state of the network key indicator type, the network future performance prediction result for the fourth future time step is determined.

[0010] As an optional implementation, in the first aspect of the present invention, the network key indicator type includes one or more of the following: queue length type, packet loss rate type, latency type, bandwidth type, and other transmission-related indicator types. And, the step of performing corresponding routing weight configuration optimization operations on the routing transmission path based on the network future performance prediction results to obtain the final routing weight results includes: Based on the network future performance prediction results and the determined target future adjustment period, determine the average index value of each sub-index type for each route transmission path; Based on the determined network performance transmission requirements, determine the configuration trade-off coefficient for each of the sub-index types; Based on the configuration trade-off coefficient of each sub-index type, the average index value of each sub-index type for each route transmission path, and the preset path comprehensive score determination method, the path comprehensive score result of each route transmission path is determined, and the path comprehensive score result represents the current path congestion status of the route transmission path. Based on the comprehensive path score of each route transmission path, perform corresponding route priority adjustment operations on all route transmission paths to obtain the final route weight result; And, after performing the corresponding performance prediction operation based on the standardized index data to obtain the network future performance prediction result, the method further includes: Determine whether the target network device meets the preset event-triggered route adjustment conditions; when it is determined that the target network device meets the event-triggered route adjustment conditions, execute the step of performing corresponding route weight configuration optimization operations on the route transmission path based on the network future performance prediction results to obtain the final route weight result; and / or, Determine whether the target network device meets the preset timed route adjustment conditions; when it is determined that the target network device meets the timed route adjustment conditions, execute the step of performing corresponding route weight configuration optimization operation on the route transmission path based on the network future performance prediction results to obtain the final route weight result.

[0011] As an optional implementation, in the first aspect of the present invention, the step of performing corresponding data standardization processing operations based on the collected raw indicator data for the target network device and key network indicator types to obtain standardized indicator data includes: Based on the collected raw indicator data for target network devices and key network indicator types, determine the raw dataset for target sub-indicator types, wherein the raw dataset includes at least two raw sample values; Perform the corresponding mean calculation operation on the original dataset to obtain the first mean value, and perform the corresponding standard deviation calculation operation on the original dataset to obtain the first standard deviation value; Based on the first average value and the first standard deviation, a corresponding standardization transformation operation is performed on each of the original sample values ​​in the original dataset to obtain a basic standardized dataset of the target sub-index type. The basic standardized dataset includes basic standard sample values ​​corresponding to each of the original sample values. Perform the corresponding mean calculation operation on the basic standardized dataset to obtain the second mean value, and perform the corresponding standard deviation calculation operation on the basic standardized dataset to obtain the second standard deviation value; Based on the second average value and the second standard deviation, determine whether the target sub-indicator type meets the preset data anomaly conditions; When it is determined that the target sub-indicator type meets the data anomaly conditions, an anomaly judgment value for each of the basic standard sample values ​​is determined based on each of the basic standard sample values, the second average value, and the second standard deviation value; based on the anomaly judgment value for each of the basic standard sample values, a corresponding anomaly sample removal operation is performed on the basic standardized dataset to obtain the final standardized dataset of the target sub-indicator type. When it is determined that the target sub-indicator type does not meet the data anomaly conditions, the basic standardized dataset is determined as the final standardized dataset of the target sub-indicator type. Based on the final standardized dataset of the target sub-indicator type, determine the standardized indicator data.

[0012] A second aspect of this invention discloses an artificial intelligence-based performance prediction and dynamic adjustment system for 5G virtual private networks, the system comprising: The data standardization module is used to perform corresponding data standardization processing operations based on the collected raw indicator data for the target network devices and key network indicator types, to obtain standardized indicator data. The performance prediction module is used to perform corresponding performance prediction operations based on the standardized index data to obtain the network future performance prediction results. The routing weight optimization module is used to perform corresponding routing weight configuration optimization operations on the routing transmission path based on the network future performance prediction results, so as to obtain the final routing weight result, so as to realize the balanced distribution and scheduling adjustment of network traffic.

[0013] As an optional implementation, in the second aspect of the present invention, the performance prediction module performs corresponding performance prediction operations based on the standardized index data to obtain the network future performance prediction results in the following specific ways: Based on the standardized indicator data and the supplementary historical performance indicator data, the target data set is determined; Based on the target dataset, perform the corresponding two-parameter exponential average prediction operation to obtain the future performance trend prediction results; Based on the future performance trend prediction results, corresponding network status and performance level prediction operations are performed to obtain the future network performance prediction results, which include the future network performance status and the predicted values ​​of the network key indicator types. Furthermore, the performance prediction module performs corresponding network state and performance level prediction operations based on the future performance trend prediction results, and the specific methods for obtaining the future network performance prediction results include: Based on the future performance trend prediction results, determine the first predicted value information of the network key indicator type for the first target time period, and determine the first actual value information of the network key indicator type for the first target time period. Based on the first actual value information and the first predicted value information, perform corresponding absolute error calculation and performance error state division operations to obtain at least two performance error state schemes. Each performance error state scheme includes a performance error state and its corresponding absolute error interval. Based on the determined network performance error state data for the second target time period and the performance error state scheme, the state transition probability matrix result is determined; Based on the predicted future performance trends and the state transition probability matrix, corresponding fusion prediction and error analysis operations are performed to obtain the predicted future performance of the network.

[0014] As an optional implementation, in the second aspect of the present invention, the performance prediction module performs a corresponding two-parameter exponential average prediction operation based on the target data set to obtain the future performance trend prediction result in the following specific ways: Based on the target dataset, determine the network performance data characteristics; Based on the network performance data characteristics and the determined network performance transmission demand information, a first value bias range for the horizontal smoothing parameter and a second value bias range for the trend smoothing parameter are determined. Based on the first value bias range, the second value bias range, and the determined prediction result accuracy requirement information, the optimal parameter combination is determined. The optimal parameter combination includes the first optimal value of the horizontal smoothing parameter and the second optimal value of the trend smoothing parameter. Based on the target data set, determine the actual observed value of the network key indicator type at the first current moment, and based on the actual observed value of the network key indicator type, the first optimal value, and the second optimal value, determine the future performance trend prediction curve of the network key indicator type. Based on the future performance trend prediction curves of the aforementioned key network indicator types, the future performance trend prediction results are determined.

[0015] As an optional implementation, in the second aspect of the present invention, the method by which the performance prediction module determines the state transition probability matrix result based on the determined network performance error state data for the second target time period and the performance error state scheme specifically includes: Perform corresponding state sequence generation operations on the determined network performance error state data for the second target time period to obtain at least two state sequence results. Each state sequence result includes the target transition time and its corresponding state transition relationship. Based on all the state sequence results, determine the total number of transitions for each performance error state that satisfies the preset initial state condition, and based on all the state sequence results, determine the specific number of transitions for each state transition relationship. Based on the specific number of transitions in each state transition relationship, the transition probability value of each state transition relationship is determined. The state transition probability matrix result is determined based on the number of state types corresponding to all the performance error states and the transition probability value of each state transition relationship.

[0016] As an optional implementation, in the second aspect of the present invention, the performance prediction module performs corresponding fusion prediction and error analysis operations based on the future performance trend prediction results and the state transition probability matrix results to obtain the network future performance prediction results in the following specific ways: Based on the future performance trend prediction results, determine the second predicted value of the network key indicator type for the second current time, and determine the second actual value of the network key indicator type for the second current time; Based on the second predicted value and the second actual value, calculate the first difference, and based on the first difference and all the performance error state schemes, determine the first actual performance error state of the network key indicator type for the second current time. Based on the first actual performance error state and the state transition probability matrix result, the second predicted performance error state of the network key indicator type for the third future time is determined, and based on the second predicted value and the first error correction coefficient of the determined first actual performance error state, the third predicted value of the network key indicator type for the third future time is determined. Based on the network key indicator type, the third predicted value and the second predicted performance error state for the third future time are used to determine the network future performance prediction result for the third future time. Furthermore, the performance prediction module performs corresponding fusion prediction and error analysis operations based on the future performance trend prediction results and the state transition probability matrix results to obtain the network future performance prediction results. Specifically, this method also includes: Furthermore, the method further includes: Determine the third actual value of the network key indicator type for the third future time, and calculate the second difference based on the third predicted value and the third actual value; Based on the second difference and the second prediction performance error state, it is determined whether the network key indicator type meets the preset performance direct prediction conditions for the fourth future time. When it is determined that the network key indicator type meets the performance direct prediction condition, the second predicted performance error state is determined as the third actual performance error state of the network key indicator type for the third future time. When it is determined that the network key indicator type does not meet the direct performance prediction condition, the third actual performance error state of the network key indicator type for the third future time is determined based on the second difference and all the performance error state schemes. Based on the third actual performance error state and the state transition probability matrix result, the fifth predicted performance error state of the network key indicator type for the fourth future time is determined, and based on the third predicted value and the third error correction coefficient of the determined third actual performance error state, the fourth predicted value of the network key indicator type for the fourth future time is determined. Based on the fourth predicted value for the fourth future time step and the fifth predicted performance error state of the network key indicator type, the network future performance prediction result for the fourth future time step is determined.

[0017] As an optional implementation, in the second aspect of the present invention, the network key indicator type includes one or more of the following: queue length type, packet loss rate type, latency type, bandwidth type, and other transmission-related indicator types. Furthermore, the routing weight optimization module performs corresponding routing weight configuration optimization operations on the routing transmission path based on the network future performance prediction results, and the specific methods for obtaining the final routing weight results include: Based on the network future performance prediction results and the determined target future adjustment period, determine the average index value of each sub-index type for each route transmission path; Based on the determined network performance transmission requirements, determine the configuration trade-off coefficient for each of the sub-index types; Based on the configuration trade-off coefficient of each sub-index type, the average index value of each sub-index type for each route transmission path, and the preset path comprehensive score determination method, the path comprehensive score result of each route transmission path is determined, and the path comprehensive score result represents the current path congestion status of the route transmission path. Based on the comprehensive path score of each route transmission path, perform corresponding route priority adjustment operations on all route transmission paths to obtain the final route weight result; The system also includes: The judgment module is used to determine whether the target network device meets preset event-triggered routing adjustment conditions after the performance prediction module performs corresponding performance prediction operations based on the standardized index data to obtain the network future performance prediction result; when it is determined that the target network device meets the event-triggered routing adjustment conditions, it triggers the routing weight optimization module to perform the step of performing corresponding routing weight configuration optimization operations on the routing transmission path based on the network future performance prediction result to obtain the final routing weight result; and / or, determines whether the target network device meets preset timed-triggered routing adjustment conditions; when it is determined that the target network device meets the timed-triggered routing adjustment conditions, it triggers the routing weight optimization module to perform the step of performing corresponding routing weight configuration optimization operations on the routing transmission path based on the network future performance prediction result to obtain the final routing weight result.

[0018] As an optional implementation, in the second aspect of the present invention, the data standardization module performs corresponding data standardization processing operations based on the collected raw indicator data for the target network device and key network indicator types, and the specific methods for obtaining standardized indicator data include: Based on the collected raw indicator data for target network devices and key network indicator types, determine the raw dataset for target sub-indicator types, wherein the raw dataset includes at least two raw sample values; Perform the corresponding mean calculation operation on the original dataset to obtain the first mean value, and perform the corresponding standard deviation calculation operation on the original dataset to obtain the first standard deviation value; Based on the first average value and the first standard deviation, a corresponding standardization transformation operation is performed on each of the original sample values ​​in the original dataset to obtain a basic standardized dataset of the target sub-index type. The basic standardized dataset includes basic standard sample values ​​corresponding to each of the original sample values. Perform the corresponding mean calculation operation on the basic standardized dataset to obtain the second mean value, and perform the corresponding standard deviation calculation operation on the basic standardized dataset to obtain the second standard deviation value; Based on the second average value and the second standard deviation, determine whether the target sub-indicator type meets the preset data anomaly conditions; When it is determined that the target sub-indicator type meets the data anomaly conditions, an anomaly judgment value for each of the basic standard sample values ​​is determined based on each of the basic standard sample values, the second average value, and the second standard deviation value; based on the anomaly judgment value for each of the basic standard sample values, a corresponding anomaly sample removal operation is performed on the basic standardized dataset to obtain the final standardized dataset of the target sub-indicator type. When it is determined that the target sub-indicator type does not meet the data anomaly conditions, the basic standardized dataset is determined as the final standardized dataset of the target sub-indicator type. Based on the final standardized dataset of the target sub-indicator type, determine the standardized indicator data.

[0019] A third aspect of this invention discloses another artificial intelligence-based performance prediction and dynamic adjustment system for 5G virtual private networks, the system comprising: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the artificial intelligence prediction performance and dynamic adjustment method for 5G virtual private networks disclosed in the first aspect of the present invention.

[0020] The fourth aspect of the present invention discloses a computer storage medium storing computer instructions, which, when invoked, are used to execute the artificial intelligence prediction performance and dynamic adjustment method for 5G virtual private networks disclosed in the first aspect of the present invention.

[0021] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: In this embodiment of the invention, based on the collected raw indicator data for the target network device and key network indicator types, corresponding data standardization processing operations are performed to obtain standardized indicator data. Based on this standardized indicator data, corresponding performance prediction operations are performed to obtain network future performance prediction results. Based on these network future performance prediction results, corresponding routing weight configuration optimization operations are performed on the routing transmission path to obtain the final routing weight result, thereby achieving balanced distribution and scheduling adjustment of network traffic. It is evident that this invention can achieve dynamic routing adjustment through data standardization processing, network future performance prediction, and routing weight configuration optimization operations. This improves the accuracy and reliability of indicator data standardization processing, which in turn improves the accuracy and reliability of network future performance prediction, thereby improving the accuracy and reliability of routing weight adjustment. Furthermore, it facilitates automatic rebalancing of network traffic and enhances overall network load balancing and transmission efficiency, flexibly avoiding faulty links or congested areas, and providing stable and efficient communication support for the intelligent operation of industry control systems. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating an artificial intelligence prediction and dynamic adjustment method for 5G virtual private networks disclosed in an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of an artificial intelligence prediction and dynamic adjustment system for a 5G virtual private network disclosed in an embodiment of the present invention; Figure 3 This is a schematic diagram of another 5G virtual private network artificial intelligence prediction performance and dynamic adjustment system disclosed in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of another 5G virtual private network artificial intelligence prediction performance and dynamic adjustment system disclosed in an embodiment of the present invention; Figure 5 This is a schematic diagram of a future performance trend prediction curve disclosed in an embodiment of the present invention. Detailed Implementation

[0024] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or end that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or ends.

[0026] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0027] This invention discloses an artificial intelligence-based performance prediction and dynamic adjustment method and system for 5G virtual private networks. It enables dynamic routing adjustment through data standardization processing, network future performance prediction, and routing weight configuration optimization. This improves the accuracy and reliability of standardized data processing, thereby enhancing the accuracy and reliability of network future performance prediction. Consequently, it improves the accuracy and reliability of routing weight adjustment, further facilitating automatic network traffic rebalancing and improving overall network load balancing and transmission efficiency. It also allows for flexible avoidance of faulty links or congested areas, providing stable and efficient communication support for the intelligent operation of industry control systems. Detailed descriptions follow.

[0028] Example 1 Please see Figure 1 , Figure 1 This is a flowchart illustrating an artificial intelligence-based performance prediction and dynamic adjustment method for a 5G virtual private network disclosed in an embodiment of the present invention. Wherein, Figure 1The described method can be applied to an AI-based performance prediction and dynamic adjustment system for 5G virtual private networks. This system may include a server, which may be a local server or a cloud server; however, this embodiment of the invention does not limit the scope of the application. Figure 1 As shown, the AI-based performance prediction and dynamic adjustment method for this 5G virtual private network includes the following operations: 101. Based on the collected raw indicator data for the target network devices and key network indicator types, perform the corresponding data standardization processing operations to obtain standardized indicator data.

[0029] Optionally, the network key indicator types may include, but are not limited to, one or more of the following: queue length type, packet loss rate type, latency type, bandwidth type, and other transmission-related indicator types. This embodiment of the invention does not impose any limitations.

[0030] Optionally, the target network device can be understood as various network element devices in the network, such as routers, switches, etc., but this embodiment of the invention does not limit it.

[0031] Optionally, the above data standardization processing operations can be illustrated by, for example, cleaning and removing outliers from historical network performance data. Performance data of 5G industry virtual private networks may contain outliers due to equipment failure (such as false port reports), collection errors (such as sensor malfunctions), instantaneous interference (such as sudden increases in latency caused by electromagnetic pulses), etc. It is necessary to detect, judge, and then remove outlier samples. This embodiment of the invention does not limit this.

[0032] 102. Based on the standardized indicator data, perform the corresponding performance prediction operations to obtain the network's future performance prediction results.

[0033] Optionally, the above performance prediction operation is specifically implemented by predicting key network indicators through two-parameter exponential average prediction, error calculation, and state partitioning. Furthermore, the error calculation and state partitioning operations are specifically implemented through performance error state setting, Markov chain modeling, fusion prediction and error analysis, and future state and performance prediction, etc. The embodiments of the present invention are not limited.

[0034] Optionally, the network future performance prediction results can be understood as the predicted values ​​of the network future performance status and indicators, which is not limited in this embodiment of the invention.

[0035] 103. Based on the network's future performance prediction results, perform corresponding routing weight configuration optimization operations on the routing transmission paths to obtain the final routing weight results, so as to achieve balanced distribution and scheduling adjustment of network traffic.

[0036] Optionally, based on the collected real-time data, the algorithm uses a weighted average method to dynamically adjust the weights in the routing table. Specifically, for each optional path, a comprehensive score reflecting the current congestion status is calculated according to the congestion level of its associated network devices. This score is used to adjust the priority of the path. The lower the score, the lower the current congestion level of the path, and the higher the routing priority will be assigned. This embodiment of the invention does not limit this.

[0037] Optionally, the routing transmission path can be understood as the currently available path. Further, the target routing transmission path to be finally adapted is selected from the currently available paths. This embodiment of the invention does not impose any limitations.

[0038] Alternatively, the method may further include the following operations: Based on the final routing weight result, the target routing path with the highest priority is determined from all routing paths and used as the current selected routing path.

[0039] As can be seen, the AI-based performance prediction and dynamic adjustment method for 5G virtual private networks described in this embodiment of the invention can achieve dynamic routing adjustment through data standardization processing, network future performance prediction, and routing weight configuration optimization. This is beneficial for improving the accuracy and reliability of standardized processing of indicator data, thereby improving the accuracy and reliability of network future performance prediction, which in turn improves the accuracy and reliability of routing weight adjustment. Furthermore, it is beneficial for achieving automatic rebalancing of network traffic and improving the overall network load balance and transmission efficiency, flexibly avoiding faulty links or congested areas, and providing stable and efficient communication support for the intelligent operation of industry control systems.

[0040] In an optional embodiment, the above-mentioned performance prediction operation based on standardized index data to obtain the network future performance prediction result may include: The target dataset was determined based on standardized indicator data and supplemented historical performance indicator data. Based on the target dataset, perform the corresponding two-parameter exponential average prediction operation to obtain the future performance trend prediction results; Based on the future performance trend prediction results, the corresponding network status and performance level prediction operations are performed to obtain the future network performance prediction results, which include the predicted values ​​of the future network performance status and key network indicator types.

[0041] Optionally, based on standardized indicator data, supplementary historical indicator data can be collected and integrated to form a complete predictive analysis dataset (i.e., target data set) to complete the initialization preparation of the predictive model. This embodiment of the invention does not limit this.

[0042] Optionally, based on the target dataset, the horizontal smoothing parameter α and the trend smoothing parameter β are initialized. The current level and long-term trend of the data are analyzed using the formula of predicted value, level value and trend value of the two-parameter exponential average method to generate a preliminary future performance trend prediction curve as the future performance trend prediction result. This embodiment of the invention does not limit the scope of the invention.

[0043] As can be seen, this optional embodiment can achieve network future performance prediction through target data set determination operation, two-parameter exponential average prediction operation, and network state and performance level prediction operation. This is beneficial to improving the comprehensiveness, rationality, and progressiveness of the method for determining network future performance prediction results, thereby improving the accuracy and reliability of the determined network future performance prediction results, and thus improving the prediction accuracy and reliability of network future performance.

[0044] In another optional embodiment, the above-mentioned network state and performance level prediction operation based on the future performance trend prediction results to obtain the network future performance prediction results may include: Based on the future performance trend prediction results, determine the first predicted value information of the network key indicator type for the first target time period, and determine the first actual value information of the network key indicator type for the first target time period. Based on the first actual value information and the first predicted value information, perform the corresponding absolute error calculation and performance error state division operation to obtain at least two performance error state schemes. Each performance error state scheme includes a performance error state and its corresponding absolute error interval. Based on the determined network performance error state data and performance error state scheme for the second target time period, the state transition probability matrix result is determined; Based on the future performance trend prediction results and the state transition probability matrix results, corresponding fusion prediction and error analysis operations are performed to obtain the future performance prediction results of the network.

[0045] Optionally, based on the preliminary prediction curve (i.e., the prediction result of future performance trends) generated by the two-parameter exponential averaging method, real-time network performance data is continuously collected and compared with the predicted value to calculate the absolute error; the absolute error over a period of time is statistically analyzed, and the range of the absolute error is recorded. It is then divided into several non-overlapping sub-intervals according to the value, with each interval representing a state level of network performance, such as "predicted value is severely high", "predicted value is generally high", "predicted value is normal", "predicted value is generally low", "predicted value is severely low", etc. (i.e., performance error state), thus completing the mapping from error to performance state. This embodiment of the invention does not impose any limitations.

[0046] Optionally, based on the performance error state division results within a historical period, the transition patterns between different performance error states are statistically analyzed, the state transition probability matrix is ​​calculated, and the evolutionary relationship between states is quantified. This embodiment of the invention does not impose any limitations.

[0047] Optionally, the preliminary prediction results can be corrected by using the future performance trend prediction curve of the two-parameter exponential averaging method and the state transition probability matrix of the Markov chain as dual bases, combined with real-time error data, to improve the flexibility and accuracy of the prediction. Based on the above-mentioned corrected prediction results, the current network performance status and state transition probability, the most likely performance status of the network in the future time period and the corresponding queue length, packet loss rate, latency and bandwidth prediction values ​​are deduced to form the final network future performance prediction result. This embodiment of the invention does not limit the scope of the prediction.

[0048] Optionally, based on the number of packets in the port queue of a certain device, the first actual value information, the first predicted value information, and the absolute error can be illustrated with, but are not limited to, Table 1-1 below:

[0049] Based on Table 1-1 above, for example, the performance error states and their corresponding absolute error ranges are as follows: State 1 (predicted value is severely high): [-20, -10); State 2 (predicted value is generally high): [-10, 0); State 3 (predicted value is generally low): [0, 10); State 4 (predicted value is severely low): [10, 20]. This embodiment of the invention does not limit the specific cases.

[0050] Optionally, the second target time period can be a time period determined according to actual needs, such as the past 7 days, etc., and this embodiment of the invention does not limit it.

[0051] As can be seen, this optional embodiment can obtain a performance error state scheme through absolute error calculation and performance error state division operation. Based on the network performance error state data and the performance error state scheme, the state transition probability matrix result is determined, and then the network future performance prediction result is obtained through fusion prediction and error analysis operation. This is conducive to improving the comprehensiveness, rationality and progressiveness of the network state and performance level prediction method, and thus conducive to improving the pertinence, flexibility and fit of the parameters involved in determining the network future performance prediction result, thereby improving the accuracy and reliability of the determined network future performance prediction result.

[0052] In another optional embodiment, the above-mentioned execution of a corresponding two-parameter exponential average prediction operation based on the target dataset to obtain the future performance trend prediction result may include: Based on the target dataset, determine the characteristics of network performance data; Based on the characteristics of network performance data and the determined network performance transmission demand information, a first value bias range for the horizontal smoothing parameter and a second value bias range for the trend smoothing parameter are determined. Based on the first value bias range, the second value bias range, and the determined prediction result accuracy requirements, the optimal parameter combination is determined. The optimal parameter combination includes the first optimal value of the horizontal smoothing parameter and the second optimal value of the trend smoothing parameter. Based on the target dataset, determine the actual observed values ​​of the network key indicator types at the first current moment, and based on the actual observed values, the first optimal value, and the second optimal value of the network key indicator types, determine the future performance trend prediction curves of the network key indicator types. The future performance trend prediction results are determined based on the future performance trend prediction curves of key network indicator types.

[0053] Optionally, the two-parameter exponential averaging method is a time series forecasting method, particularly suitable for time series data with obvious trends. It adds consideration of trends to the simple exponential smoothing by introducing two smoothing parameters, α (horizontal smoothing parameter) and β (trend smoothing parameter), which correspond to the weight between the current observation and the smoothed value, and the weight between the current trend and the smoothed trend, respectively. The parameter values ​​of α and β are not fixed constants, but are dynamically adapted and determined based on network performance data characteristics, service QoS requirements (i.e., network performance transmission requirements information) and prediction accuracy targets (i.e., prediction result accuracy requirements information). This embodiment of the invention does not limit these values.

[0054] Optionally, the above network performance data characteristics, network performance transmission demand information, the first value bias range for the horizontal smoothing parameter, and the second value bias range for the trend smoothing parameter are illustrated as follows: (1) For scenarios with significant trends: if historical data (such as queue length, latency) shows a significant long-term growth / decline trend (such as the bandwidth occupancy rate continuously increasing during the centralized backhaul period of power metering automation services), then the second value bias range for the trend smoothing parameter is 0.3~0.5 (i.e., enhancing the trend capture capability), and the first value bias range for the horizontal smoothing parameter is 0.2~0.4 (i.e., weakening the impact of immediate fluctuations on prediction and focusing on long-term trends); (2) For scenarios with gentle trends: if the data has no obvious trend and is mainly characterized by random fluctuations (such as the packet loss rate during normal periods), then the second value bias range for the trend smoothing parameter is 0.05~0.2 (i.e., reducing the weight of the trend term and avoiding overfitting), and the first value bias range for the horizontal smoothing parameter is 0.05~0.2 (i.e., reducing the weight of the trend term and avoiding overfitting). The first value bias range of the smoothing parameter is 0.5~0.7 (that is, to strengthen the weight of the real-time observation and improve the response speed to short-term fluctuations); (3) For high real-time services (such as power distribution automation, remote control of relay opening and closing): the real-time requirements of prediction are high and the network state changes need to be responded to quickly. Therefore, the first value bias range of the horizontal smoothing parameter is 0.6~0.8 (that is, to prioritize following the changes of the current observation). The second value bias range of the trend smoothing parameter is 0.1~0.3 (that is, to moderately capture the trend); (4) For high stability services (such as video surveillance, equipment status monitoring, etc.): a certain prediction delay is allowed, but the prediction result is required to be stable without violent fluctuations. Therefore, the first value bias range of the horizontal smoothing parameter is 0.1~0.3 (that is, to smooth the real-time fluctuations). The second value bias range of the trend smoothing parameter is 0.3~0.6 (that is, to focus on tracking the long-term trend). This embodiment of the invention does not limit the scope.

[0055] Optionally, the optimal parameter combination is determined based on the first value bias range, the second value bias range, and the determined prediction result accuracy requirements. For example, with the optimization objective of "minimizing the absolute error (MAE) between the predicted value and the actual value", the optimal parameter combination is determined through trial calculation or algorithm iteration to ensure that the prediction accuracy meets the routing adjustment requirements (usually requiring MAE ≤ 5%, and the specific threshold can be adjusted according to business needs). This embodiment of the invention does not impose any limitations.

[0056] Optionally, by adapting to long-term growth or decline trends in the data, predicted values ​​for key network indicator types for future periods can be provided. Further, based on the actual observed values, the first optimal value, and the second optimal value of the key network indicator type, a future performance trend prediction curve for that type is determined. For example, the formula for calculating the predicted value is: ; The formula for calculating the level value is: ; The formula for calculating the trend value is: ; in, It is the actual observed value at the current moment; This represents the smoothing level of the indicator at the current moment, reflecting the current stable state of the indicator after real-time fluctuation correction; This represents the smooth trend of the indicator at the current moment, reflecting the long-term growth or decline pattern of the indicator (such as the trend of continuous increase in queue length, continuous sufficiency of bandwidth, etc.). This embodiment of the invention does not limit this.

[0057] Optionally, since different key network metrics need to be predicted separately, a two-parameter exponential average calculation will be performed independently for each metric to obtain the future performance trend prediction result. This can be referenced, but is not limited to, the two-parameter exponential average prediction operation described above, i.e., the y-value corresponding to the queue length. t The y-value corresponds to the "current standardized queue length observation" and the packet loss rate. t Let y be the "current standardized packet loss rate observation" and the corresponding delay. t For the "current standardized delay observation" and the bandwidth corresponding to y t The "current standardized time-bandwidth measurement" ensures that the prediction of each indicator is based on its own preprocessed real observation data, and the embodiments of the present invention do not impose any limitations.

[0058] Optionally, future performance trend prediction curves can be referenced, but are not limited to, those attached to the instruction manual. Figure 5 As shown, the embodiments of the present invention are not limited.

[0059] Optionally, the smoothing parameters α and β of the two-parameter exponential averaging method are initialized. Generally, α is responsible for smoothing the current level, while β is used to capture trend changes. The two-parameter exponential averaging method is applied for time series analysis. This method uses two smoothing parameters to simultaneously consider the current level and long-term trend of the data. In specific operation, α is used to weight the most recent observations to reflect immediate changes, while β is used to adjust the trend term and predict possible future growth or decline trends. This process generates a future performance trend prediction curve based on historical data. This embodiment of the invention is not limited.

[0060] As can be seen, this optional embodiment can determine the value bias range of the horizontal smoothing parameter and the trend smoothing parameter based on the network performance data characteristics and network performance transmission demand information, and determine the optimal values ​​of the horizontal smoothing parameter and the trend smoothing parameter based on the accuracy requirements of the prediction results, thereby determining the future performance trend prediction curve of the network key indicator type. This is beneficial to improving the comprehensiveness, rationality, and progressiveness of the method for determining the future performance trend prediction results, and thus beneficial to improving the diversity, flexibility, and pertinence of the parameters involved in determining the future performance trend prediction curve, thereby improving the accuracy and reliability of the determined future performance trend prediction results. In addition, it is beneficial to improve the comprehensiveness, rationality, and progressiveness of the method for determining the optimal values ​​of the horizontal smoothing parameter and the trend smoothing parameter, thereby improving the accuracy and reliability of the determined optimal values.

[0061] In yet another optional embodiment, the determination of the state transition probability matrix result based on the determined network performance error state data and performance error state scheme for the second target time period may include: Perform corresponding state sequence generation operations on the determined network performance error state data for the second target time period to obtain at least two state sequence results. Each state sequence result includes the target transition time and its corresponding state transition relationship. Based on all state sequence results, determine the total number of transitions for each performance error state that satisfies the preset initial state conditions, and based on all state sequence results, determine the specific number of transitions for each state transition relationship. The transition probability value of each state transition relationship is determined based on the specific number of transitions in each state transition relationship. The state transition probability matrix is ​​determined based on the number of state types corresponding to all performance error states and the transition probability value of each state transition relationship.

[0062] Optionally, the above network performance error state data and state sequence results can be illustrated as follows: Select network performance error state data from the past 7 days and arrange them in chronological order to form a continuous state sequence (unit time granularity is 1 minute, a total of 7×24×60=10080 moments, corresponding to 10079 state transitions). Further, for the sake of simplification, extract the state sequence of 20 consecutive transitions (the actual calculation requires complete data): Moment 1 → Moment 2: State 2 → State 3; Moment 2 → Moment 3: State 3 → State 3; Moment 3 → Moment 4: State 3 → State 2; Moment 4 → Moment 5: State 2 → State 2; Moment 5 → Moment 6: State 2 → State 3; Moment 6 → Moment 7: State 3 → State 4; Moment 7 → Moment 8 The following timeframes are defined as follows: State 4 → State 3; Time 8 → Time 9: State 3 → State 3; Time 9 → Time 10: State 3 → State 2; Time 10 → Time 11: State 2 → State 3; Time 11 → Time 12: State 3 → State 3; Time 12 → Time 13: State 3 → State 3; Time 13 → Time 14: State 3 → State 4; Time 14 → Time 15: State 4 → State 3; Time 15 → Time 16: State 3 → State 2; Time 16 → Time 17: State 2 → State 2; Time 17 → Time 18: State 2 → State 3; Time 18 → Time 19: State 3 → State 3; Time 19 → Time 20: State 3 → State 3; Time 20 → Time 21: State 3 → State 2. This embodiment of the invention is not limited in its scope.

[0063] Optionally, the total number of transitions for the aforementioned performance error state to satisfy the preset starting state condition can be understood as the total number of transitions when the performance error state is the "starting state," that is, the total number of transitions that occur at the next moment after the performance error state appears. Further, based on the above example sequence, for example: state 1 does not appear in the example sequence, so the total number of transitions for state 1 is 0; the state sequence where state 2 appears as the starting state is at times 1, 4, 5, 10, 16, 17, so the total number of transitions for state 2 is 6; the state sequence where state 3 appears as the starting state is at times 2, 3, 6, 8, 9, 11, 12, 13, 15, 18, 19, 20, so the total number of transitions for state 3 is 12; the state sequence where state 4 appears as the starting state is at times 7, 14, so the total number of transitions for state 4 is 2. This embodiment of the invention does not limit the scope of the invention.

[0064] Optionally, the specific number of transitions in the above state transition relationship can be understood as counting the corresponding number of "starting state → target state" one by one (such as the number of transitions from state 2 to state 3, the number of transitions from state 3 to state 2, etc.). Further, based on the above example sequence, refer to Table 1-2 below:

[0065] Optional, the transition probability value of the state transition relation. It can be obtained through the following calculation formula:

[0066] Furthermore, if the total number of transitions in state i is 0, then the transition probability value is uniformly recorded as 0, which is not limited in this embodiment of the invention.

[0067] Optionally, the state transition probability matrix M above is an n×n matrix (n is the number of states, here n=4), where the matrix element M[i][j]=P(i→j), the row index is the starting state, and the column index is the target state. For example, the final matrix is ​​as follows:

[0068] As can be seen, this optional embodiment can determine the transition probability value of each state transition relationship based on the specific number of transitions in each state transition relationship, thereby determining the state transition probability matrix result. This is beneficial to improving the comprehensiveness and rationality of the method for determining the state transition probability matrix result, and to improving the diversity and flexibility of the parameters involved in determining the state transition probability matrix result. In turn, it is beneficial to improving the accuracy and reliability of the determined state transition probability matrix result.

[0069] In another optional embodiment, the above-mentioned fusion prediction and error analysis operations performed based on the future performance trend prediction results and the state transition probability matrix results to obtain the network future performance prediction results may include: Based on the future performance trend prediction results, determine the second predicted value of the network key indicator type for the second current time, and determine the second actual value of the network key indicator type for the second current time. Based on the second predicted value and the second actual value, calculate the first difference, and based on the first difference and all performance error state schemes, determine the network key indicator type for the first actual performance error state at the second current moment; Based on the first actual performance error state and the state transition probability matrix results, the second predicted performance error state of the network key indicator type for the third future time is determined, and based on the second predicted value and the first error correction coefficient of the determined first actual performance error state, the third predicted value of the network key indicator type for the third future time is determined. Based on the network key indicator types, the third predicted value for the third future time and the second predicted performance error state, the network future performance prediction result for the third future time is determined.

[0070] Optionally, based on the first actual performance error state and the state transition probability matrix results, the network key indicator type is determined for the second predicted performance error state at the third future time. For example: Suppose that for the current time t (i.e., the second current time), the trend data (preliminary predicted value) is output through the future performance trend prediction results, and the second predicted value for the second current time is F. t =100, while the actual value measured at the current time (i.e., the second actual value for the second current moment) is y. t =103, and the Markov chain transition probability matrix is ​​consistent with the above M; furthermore, since y t and F t The difference is 103-100=3 (the first difference), and the current error state belongs to state 3 (i.e., the first actual performance error state). Therefore, with state 3 as the starting state, the most likely error state at time t+1 (i.e., the third future time) is predicted through the state transition probability matrix result. For state 3, among the transition probabilities of the state transition probability matrix result, P(3→3)=0.5 (the highest probability). Therefore, the error state at time t+1 (i.e., the second predicted performance error state) is predicted to be "state 3 (the predicted value is generally low)". This embodiment of the invention does not limit this.

[0071] Optionally, based on the second predicted value and the first error correction coefficient determined for the first actual performance error state, the third predicted value for the network key indicator type at the third future time is determined. For example, the "correction rule" for each performance error state is determined based on historical error statistics. If the historical average error for state 3 is 5, it indicates that the predicted value is too low. Therefore, its correction coefficient is defined as +5 (i.e., the first error correction coefficient for the first actual performance error state). This means the actual value is usually 5 higher than the predicted value. Furthermore, the initial predicted value is adjusted to ensure that the correction logic conforms to the actual error pattern. Therefore, the corrected predicted value at time t+1 is: F t+1 =100+5=105 (i.e., the third predicted value at the third future time), which is not limited in the embodiments of the present invention.

[0072] It is evident that this optional embodiment can provide a corresponding method for determining the future network performance prediction results based on the preliminary prediction values ​​determined by the future performance trend prediction results. This is beneficial to improving the comprehensiveness, rationality, and pertinence of the method for determining the future network performance prediction results, and thus to improving the accuracy and reliability of the determined network performance prediction results.

[0073] In yet another alternative embodiment, the method may further include the following operations: The network key indicator type is determined for the third actual value at the third future time, and the second difference is calculated based on the third predicted value and the third actual value. Based on the second difference and the second prediction performance error state, determine whether the network key indicator types meet the preset performance direct prediction conditions for the fourth future time. When it is determined that the network key indicator type meets the direct performance prediction conditions, the second predicted performance error state is determined as the third actual performance error state of the network key indicator type for the third future time. When it is determined that the network key indicator type does not meet the direct performance prediction conditions, the third actual performance error state of the network key indicator type for the third future time is determined based on the second difference and all performance error state schemes. Based on the results of the third actual performance error state and the state transition probability matrix, the fifth predicted performance error state of the network key indicator type for the fourth future time is determined, and based on the third predicted value and the third error correction coefficient of the determined third actual performance error state, the fourth predicted value of the network key indicator type for the fourth future time is determined. Based on the network key indicator types, the fourth predicted value and the fifth predicted performance error state for the fourth future time are used to determine the network future performance prediction result for the fourth future time.

[0074] Optional, for example: When time t+1 arrives, suppose we collect the third actual value y at the third future time. t+1 =110, the corrected prediction error (i.e., the second difference) is 110-105=5 (still in state 3), indicating that the state prediction is accurate, so continue to predict F. t+2 =105+5=110; Furthermore, assume that the third actual value y is collected at the third future time. t+1 =122, the actual error (i.e., the second difference) is 122-108=14, corresponding to the third actual performance error state at the third future time as state 4 = the predicted value is seriously low [10,20]. Then adjust the state prediction at time t+2 (i.e., the fourth future time) (based on the actual state 4 at time t+1, query matrix P(4→3)=1, predict the fifth prediction performance error state at the fourth future time as state 3), and adjust according to the correction coefficient of state 4 (e.g., the historical average error of state 4 = 15, then the correction coefficient is 15). Then the predicted fourth prediction value at the fourth future time is F. t+2 =105+15=120, and the same logic applies to other cases. This invention does not limit the specific implementation.

[0075] Further optionally, the above-mentioned determination of whether the network key indicator type meets the preset performance direct prediction conditions for the fourth future time step based on the second difference and the second prediction performance error state may include: Based on the second difference and all performance error state schemes, determine the target performance error state; determine whether the target performance error state matches the second predicted performance error state; if the determination result is yes, determine that the network key indicator type meets the preset direct performance prediction conditions for the fourth future time; if the determination result is no, determine that the network key indicator type does not meet the preset direct performance prediction conditions for the fourth future time; and / or, Based on the second predicted performance error state and all performance error state schemes, determine the target difference range; determine whether the second difference is within the target difference range; when the determination result is yes, determine that the network key indicator type meets the preset direct performance prediction conditions for the fourth future time; when the determination result is no, determine that the network key indicator type does not meet the preset direct performance prediction conditions for the fourth future time.

[0076] Optionally, the method for determining the network future performance prediction results for other future moments after the fourth future moment can refer to, but is not limited to, the methods for determining the network future performance prediction results for the third and fourth future moments described above; the network future performance prediction results for other future moments after the fourth future moment can be determined based on the network future performance prediction results for the third and fourth future moments described above, and this embodiment of the invention does not impose any limitations.

[0077] Optionally, the network may simulate the state sequence and corresponding performance indicators that it may experience in the future time period; based on the current state, the predicted trend direction and the state transition probability, the network is most likely to reach the state and its performance level at a certain point in the future. From the perspective of taking into account the impact of short-term fluctuations and long-term trends, in order to enhance the comprehensiveness and practicality of the prediction, the embodiments of the present invention are not limited.

[0078] As can be seen, this optional embodiment can provide a corresponding method for determining the future network performance prediction result based on the error correction coefficient determined by the performance error state. This is beneficial to improving the comprehensiveness and rationality of the method for determining the future network performance prediction result, as well as its diversity, flexibility, and relevance. Consequently, it is beneficial to improve the accuracy and reliability of the determined future network performance prediction result. In addition, it can match the corresponding actual performance error state determination method for those that meet the direct performance prediction conditions and those that do not meet the direct performance prediction conditions. This is beneficial to improving the diversity, flexibility, and relevance of the actual performance error state determination method, as well as the accuracy and reliability of the determined actual performance error state. This, in turn, is beneficial to improving the accuracy and reliability of the error correction based on the actual performance error state, and further beneficial to improving the accuracy and reliability of the determined prediction value.

[0079] In another optional embodiment, the above-mentioned optimization of routing weight configuration on the routing transmission path based on the network future performance prediction results to obtain the final routing weight result may include: Based on the network future performance prediction results and the determined target future adjustment period, determine the average value of each sub-index type for each route transmission path. Based on the determined network performance transmission requirements, determine the configuration trade-off coefficient for each sub-index type; Based on the configuration trade-off coefficient of each sub-index type, the average index value of each sub-index type for each route transmission path, and the preset path comprehensive score determination method, the path comprehensive score result of each route transmission path is determined. The path comprehensive score result indicates the current path congestion status of the route transmission path. Based on the comprehensive path score of each route transmission path, the corresponding route priority adjustment operation is performed on all route transmission paths to obtain the final route weight result.

[0080] Optionally, the network future performance prediction result includes the predicted value corresponding to the target future adjustment period. Further, the future time of the aforementioned operation to determine the network future performance prediction result is related to the target future adjustment period, and the integration of future times includes the target future adjustment period and is greater than or equal to the target future adjustment period. Further, it is possible to predetermine the future time for which network future performance prediction needs to be performed based on the target future adjustment period, and then determine the network future performance prediction result for that future time, and determine the predicted value corresponding to the target future adjustment period based on the network future performance prediction result for that future time. Alternatively, after the network future performance prediction result of the integrated future time including the target future adjustment period and greater than or equal to the target future adjustment period is determined, the predicted value corresponding to the target future adjustment period is determined based on the network future performance prediction result of the integrated future time. This embodiment of the invention does not limit the scope of the prediction.

[0081] Optionally, based on the network future performance prediction results, extract the predicted values ​​of four sub-indicator types (i.e., queue length prediction, packet loss rate prediction, latency prediction, and bandwidth prediction) for each optional path within the "target future adjustment period (e.g., the next 1 minute)," and calculate the mean value for each sub-indicator type. For example, calculate the mean queue length n every 10 seconds within the next 1 minute. q Mean packet loss rate n p Mean delay n t Bandwidth average n b The embodiments of the present invention are not limited.

[0082] Furthermore, for example: In a certain 5G power virtual private network, there are two optional routing transmission paths (path 1 and path 2). The average indicators for the next minute are shown in Table 1-3 below:

[0083] Optionally, based on the determined network performance transmission requirements, the configuration trade-off coefficient for each sub-index type is determined. This can be understood as the differentiated QoS requirements for services, and a trade-off coefficient is configured for each sub-index. The configuration trade-off coefficient for queue length is... A q The configuration trade-off coefficient for packet loss rate is: A p The latency configuration trade-off coefficient is A t The bandwidth configuration trade-off factor is A b The coefficients range from (0,1], and their sum is 1. Furthermore, for example: for high real-time services (such as power distribution automation), the weights of latency and packet loss rate are increased, while the weights of queue length and bandwidth are decreased; for high bandwidth-demand services (such as video surveillance), the weight of bandwidth is increased, while the weights of other indicators are decreased; further, for example, in this case, the service is power distribution automation. A q =0.2、 A p =0.3、 A t =0.4、 A b =0.1, which is not limited in the embodiments of the present invention.

[0084] Optionally, the comprehensive route score can be determined by referring to, but is not limited to, the following formula:

[0085] Optionally, for the overall path score of the route transmission path, for example: the overall path score of path 1 is s1 = 0.2 × 25 + 0.3 × 0.3 + 0.4 × 35 0.1 × 800 = 60.91; The overall path score for path 2 is s² = 0.2 × 40 + 0.3 × 0.8 + 0.4 × 60 0.1 × 500 = 17.76; Furthermore, the lower the score, the lower the current congestion of the path, and the higher the routing priority will be assigned. In this case, path 1 will be selected for new traffic transmission. In addition, the transmission traffic of path 2 can be reduced and the transmission traffic of path 1 can be increased. This embodiment of the invention does not limit this.

[0086] Optionally, based on the comprehensive path score, higher routing priority is assigned to low-congestion paths with lower scores, and lower priority is assigned to high-congestion paths with higher scores, dynamically adjusting the routing table weights; simultaneously, the Pre-Label Reattachment (PRL) algorithm is triggered to calculate the globally optimal path based on the adjusted weights, guiding service traffic to redirect to low-congestion paths, achieving automatic rebalancing of network traffic, and ensuring service transmission performance; prompting the redistribution of data flows to avoid excessive concentration on a few highly congested links, thereby achieving automatic rebalancing of network traffic, improving the overall network load balance and transmission efficiency, and this embodiment of the invention is not limited.

[0087] As can be seen, this optional embodiment can determine the comprehensive path score of each route transmission path based on the configuration trade-off coefficient of each sub-index type and the average index of each route transmission path, and then perform route priority adjustment operation to obtain the final route weight result. This is beneficial to improving the comprehensiveness and rationality of the final route weight result determination method, and to improving the diversity and flexibility of the parameters considered in the route priority adjustment. This, in turn, is beneficial to improving the accuracy and reliability of the route priority adjustment, and thus to improving the accuracy and reliability of the determined final route weight result.

[0088] In yet another optional embodiment, after performing the corresponding performance prediction operation based on the standardized index data to obtain the network future performance prediction result, the method may further include the following operations: Determine whether the target network device meets the preset event-triggered route adjustment conditions; when it is determined that the target network device meets the event-triggered route adjustment conditions, execute the above-mentioned steps of optimizing the route transmission path according to the network future performance prediction results to obtain the final route weight result; and / or, Determine whether the target network device meets the preset timed route adjustment conditions; when it is determined that the target network device meets the timed route adjustment conditions, execute the above steps of optimizing the route transmission path according to the network future performance prediction results to obtain the final route weight result.

[0089] Optionally, route weight configuration optimization is triggered every 1 minute by default (configurable to 30 seconds / 5 minutes according to business needs) to adapt to the smooth changes in network status; when any indicator in the future network performance prediction results exceeds the preset threshold (such as prediction latency ≥ 50ms, prediction packet loss rate ≥ 1%), or when a link failure occurs in the network or a sudden surge in business traffic (such as traffic increase ≥ 50%), route weight configuration optimization is triggered immediately. This embodiment of the invention does not impose any limitations.

[0090] Further, optionally, the above determination of whether the target network device meets the preset event-triggered routing adjustment conditions includes: Based on the network future performance prediction results, determine the predicted value for each sub-index type; determine whether the predicted value for each sub-index type is greater than or equal to the preset corresponding predicted value threshold; if the determination result is yes, determine that the target network device meets the preset event-triggered routing adjustment conditions; if the determination result is no, determine that the target network device does not meet the preset event-triggered routing adjustment conditions; and / or, Determine whether the target network device meets the preset abnormal operation conditions (such as network link failure conditions, sudden increase in service traffic conditions, etc.); if the determination result is yes, determine that the target network device meets the preset event-triggered routing adjustment conditions; if the determination result is no, determine that the target network device does not meet the preset event-triggered routing adjustment conditions.

[0091] As can be seen, this optional embodiment can trigger the execution of route weight configuration optimization operation when the target network device meets the event-triggered route adjustment condition and / or the timed route adjustment condition. This is beneficial to improving the diversity, flexibility and pertinence of the triggering and execution evaluation method of route weight configuration optimization operation, thereby improving the accuracy and timeliness of the triggering and execution of route weight configuration optimization operation and reducing unnecessary resource consumption of route weight configuration optimization operation.

[0092] In another optional embodiment, the above-mentioned data standardization processing operation, performed on the collected raw indicator data for the target network device and key network indicator types to obtain standardized indicator data, may include: Based on the collected raw indicator data for the target network devices and key network indicator types, determine the raw dataset for the target sub-indicator types. The raw dataset includes at least two raw sample values. Perform the corresponding mean calculation operation on the original dataset to obtain the first mean value, and perform the corresponding standard deviation calculation operation on the original dataset to obtain the first standard deviation value; Based on the first mean and the first standard deviation, perform the corresponding standardization transformation operation on each original sample value in the original dataset to obtain the basic standardized dataset of the target sub-index type. The basic standardized dataset includes the basic standard sample value corresponding to each original sample value. Perform the corresponding mean calculation operation on the basic standardized dataset to obtain the second mean, and perform the corresponding standard deviation calculation operation on the basic standardized dataset to obtain the second standard deviation. Based on the second average value and the second standard deviation, determine whether the target sub-indicator type meets the preset data anomaly conditions; When it is determined that the target sub-indicator type meets the data anomaly conditions, the anomaly judgment value of each basic standard sample value is determined based on each basic standard sample value, the second average value, and the second standard deviation value; based on the anomaly judgment value of each basic standard sample value, the corresponding anomaly sample removal operation is performed on the basic standardized dataset to obtain the final standardized dataset of the target sub-indicator type. When it is determined that the target sub-indicator type does not meet the data anomaly conditions, the basic standardized dataset is determined as the final standardized dataset of the target sub-indicator type. Based on the final standardized dataset of the target sub-indicator type, determine the standardized indicator data.

[0093] Optionally, the performance data of 5G industry virtual private networks may contain outliers due to equipment failure (such as port false alarms), collection errors (such as sensor malfunctions), or transient interference (such as sudden increases in latency caused by electromagnetic pulses). It is necessary to clean the data and remove outliers to ensure data quality. This embodiment of the invention does not limit the scope of the outlier.

[0094] Optionally, based on the collected raw indicator data for the target network devices and key network indicator types, the original dataset for the target sub-indicator type is determined. For example, all raw data for the target sub-indicator type (such as queue length data at all historical moments) is extracted from the collected historical network performance data and denoted as the original dataset. X raw ={ X raw1 ,X raw2 ,...,X rawm},in X rawi Let m be the i-th original sample value of the target sub-index type, and m be the total number of samples of the target sub-index type. This embodiment of the invention does not limit the value of m.

[0095] Optionally, the above first average value E(X) raw It can be calculated using the following formula: ; Wherein, E(X) raw ) is the expected value E(X) of the original dataset for this target sub-index type. raw The mean of the original dataset is not limited to this.

[0096] Optionally, the first standard deviation σ mentioned above can be calculated using the following formula: ; Wherein, σ is the overall standard deviation of the original dataset of the target sub-index type, which is not limited in the embodiments of the present invention.

[0097] Optionally, the above standardization transformation operation specifically involves: transforming the original dataset X of the target sub-indicator type... raw Each sample value in the dataset is calculated using the following formula to obtain the basic standard sample value, thus forming the basic standardized dataset. The specific standardization formula is as follows: ; in, X i The base standard sample values ​​corresponding to the original sample values. X Based on the standardized dataset, the embodiments of this invention are not limited.

[0098] Optionally, the formula for calculating the second average value can refer to, but is not limited to, the formula for calculating the first average value, and the formula for calculating the second standard deviation can refer to, but is not limited to, the formula for calculating the first standard deviation. This embodiment of the invention does not impose any limitations.

[0099] Further optionally, the above determination of whether the target sub-indicator type meets the preset data anomaly conditions based on the second average value and the second standard deviation may include: Determine whether the second average value matches the preset first verification threshold, and determine whether the second standard deviation matches the preset second verification threshold; When it is determined that the second average value matches the first verification threshold and the second standard deviation matches the second verification threshold, it is determined that the target sub-indicator type does not meet the preset data anomaly conditions. When it is determined that the second average value does not match the first verification threshold and / or the second standard deviation does not match the second verification threshold, the target sub-indicator type is determined to meet the preset data anomaly conditions.

[0100] Optionally, the second average value matches the first verification threshold, for example: the second average value ≈ 0; the second standard deviation matches the second verification threshold, for example: the second standard deviation ≈ 1; the second average value does not match the first verification threshold, for example: the second average value is not approximately equal to 0 and the deviation value > 0.05; the second standard deviation does not match the second verification threshold, for example: the second standard deviation is not approximately equal to 1 and the deviation value > 0.05. This embodiment of the invention does not impose any limitations.

[0101] Optionally, the anomaly assessment value of the baseline standard sample value can be calculated using the following formula: ; Where, x i Based on the standard sample value, Z i The abnormality assessment value of the basic standard sample value is E(X), which is the second average value and σ(X) is the second standard deviation value. The embodiments of the present invention are not limited.

[0102] Further, optionally, the above-mentioned outlier removal operation performed on the basic standardized dataset based on the outlier assessment value of each basic standard sample value to obtain the final standardized dataset of the target sub-indicator type may include: The absolute value of the anomaly assessment values ​​in the basic standardized dataset must be greater than or equal to a preset anomaly assessment value threshold (e.g., ...). The target basic standard sample values ​​(etc.) are identified as abnormal samples, and all abnormal samples in the basic standardized dataset are removed to obtain the final standardized dataset of the target sub-indicator type.

[0103] Optionally, the standardized data should satisfy "mean ≈ 0, standard deviation ≈ 1". If the deviation is large (absolute value > 0.05), the original data can be backtracked to check whether there are any unprocessed outliers. This embodiment of the invention does not limit this.

[0104] As can be seen, this optional embodiment can provide a specific data standardization processing method. By performing basic standardized data processing operations, data anomaly detection and removal operations, the final standardized dataset is obtained. This is beneficial to improving the comprehensiveness, rationality and progressiveness of the data standardization processing method, and to improving the diversity, flexibility and pertinence of the data standardization sub-processing methods. In turn, it is beneficial to improve the accuracy and reliability of the determined standardized indicator data. In addition, providing specific methods for determining and removing abnormal data samples is beneficial to improving the accuracy and reliability of removing abnormal data samples.

[0105] Example 2 Please see Figure 2 , Figure 2 This is a schematic diagram of the structure of an artificial intelligence prediction and dynamic adjustment system for a 5G virtual private network disclosed in an embodiment of the present invention. Figure 2 The described system may include a server, which may be a local server or a cloud server; this embodiment of the invention does not limit the scope. Figure 2 As shown, the AI-based performance prediction and dynamic adjustment system for this 5G virtual private network may include: The data standardization module 301 is used to perform corresponding data standardization processing operations based on the collected raw indicator data for the target network devices and key network indicator types to obtain standardized indicator data.

[0106] The performance prediction module 302 is used to perform corresponding performance prediction operations based on standardized index data to obtain the future performance prediction results of the network.

[0107] The routing weight optimization module 303 is used to perform corresponding routing weight configuration optimization operations on the routing transmission path based on the network future performance prediction results, so as to obtain the final routing weight result and realize the balanced distribution and scheduling adjustment of network traffic.

[0108] It is evident that implementation Figure 2 The described 5G virtual private network's AI-based performance prediction and dynamic adjustment system can achieve dynamic routing adjustments through data standardization processing, network future performance prediction, and routing weight configuration optimization. This improves the accuracy and reliability of standardized data processing, thereby enhancing the accuracy and reliability of network future performance prediction. Consequently, it improves the accuracy and reliability of routing weight adjustment, further facilitating automatic network traffic rebalancing and improving overall network load balancing and transmission efficiency. It also flexibly avoids faulty links or congested areas, providing stable and efficient communication support for the intelligent operation of industry control systems.

[0109] In an optional embodiment, the performance prediction module 302 performs corresponding performance prediction operations based on standardized index data to obtain the network future performance prediction results in the following specific ways: The target dataset was determined based on standardized indicator data and supplemented historical performance indicator data. Based on the target dataset, perform the corresponding two-parameter exponential average prediction operation to obtain the future performance trend prediction results; Based on the future performance trend prediction results, the corresponding network status and performance level prediction operations are performed to obtain the future network performance prediction results, which include the predicted values ​​of the future network performance status and key network indicator types.

[0110] It is evident that implementation Figure 3 The described system can predict the future performance of a network through operations such as target data set determination, two-parameter exponential average prediction, and network state and performance level prediction. This helps to improve the comprehensiveness, rationality, and progressiveness of the method for determining the future performance prediction results, thereby improving the accuracy and reliability of the determined future performance prediction results.

[0111] In another optional embodiment, the performance prediction module 302 performs corresponding network state and performance level prediction operations based on the future performance trend prediction results, and the specific methods for obtaining the future network performance prediction results include: Based on the future performance trend prediction results, determine the first predicted value information of the network key indicator type for the first target time period, and determine the first actual value information of the network key indicator type for the first target time period. Based on the first actual value information and the first predicted value information, perform the corresponding absolute error calculation and performance error state division operation to obtain at least two performance error state schemes. Each performance error state scheme includes a performance error state and its corresponding absolute error interval. Based on the determined network performance error state data and performance error state scheme for the second target time period, the state transition probability matrix result is determined; Based on the future performance trend prediction results and the state transition probability matrix results, corresponding fusion prediction and error analysis operations are performed to obtain the future performance prediction results of the network.

[0112] It is evident that implementation Figure 3 The described system can also obtain a performance error state scheme through absolute error calculation and performance error state division operations. Based on the network performance error state data and the performance error state scheme, the system determines the state transition probability matrix result and then obtains the network future performance prediction result through fusion prediction and error analysis operations. This is conducive to improving the comprehensiveness, rationality and progressiveness of the network state and performance level prediction method, and further conducive to improving the pertinence, flexibility and fit of the parameters involved in determining the network future performance prediction result, thereby improving the accuracy and reliability of the determined network future performance prediction result.

[0113] In another optional embodiment, the performance prediction module 302 performs a corresponding two-parameter exponential average prediction operation based on the target dataset to obtain the future performance trend prediction result in the following ways: Based on the target dataset, determine the characteristics of network performance data; Based on the characteristics of network performance data and the determined network performance transmission demand information, a first value bias range for the horizontal smoothing parameter and a second value bias range for the trend smoothing parameter are determined. Based on the first value bias range, the second value bias range, and the determined prediction result accuracy requirements, the optimal parameter combination is determined. The optimal parameter combination includes the first optimal value of the horizontal smoothing parameter and the second optimal value of the trend smoothing parameter. Based on the target dataset, determine the actual observed values ​​of the network key indicator types at the first current moment, and based on the actual observed values, the first optimal value, and the second optimal value of the network key indicator types, determine the future performance trend prediction curves of the network key indicator types. The future performance trend prediction results are determined based on the future performance trend prediction curves of key network indicator types.

[0114] It is evident that implementation Figure 3The described system can also determine the value bias range of the horizontal smoothing parameter and the trend smoothing parameter based on network performance data characteristics and network performance transmission demand information, and determine the optimal values ​​of the horizontal smoothing parameter and the trend smoothing parameter based on the accuracy requirements of the prediction results, thereby determining the future performance trend prediction curves for key network indicators. This is beneficial to improving the comprehensiveness, rationality, and progressiveness of the method for determining future performance trend prediction results, and further beneficial to improving the diversity, flexibility, and specificity of the parameters involved in determining the future performance trend prediction curves, thus improving the accuracy and reliability of the determined future performance trend prediction results. In addition, it is beneficial to improve the comprehensiveness, rationality, and progressiveness of the method for determining the optimal values ​​of the horizontal smoothing parameter and the trend smoothing parameter, thereby improving the accuracy and reliability of the determined optimal values.

[0115] In another optional embodiment, the performance prediction module 302 determines the state transition probability matrix result based on the determined network performance error state data and performance error state scheme for the second target time period in the following specific ways: Perform corresponding state sequence generation operations on the determined network performance error state data for the second target time period to obtain at least two state sequence results. Each state sequence result includes the target transition time and its corresponding state transition relationship. Based on all state sequence results, determine the total number of transitions for each performance error state that satisfies the preset initial state conditions, and based on all state sequence results, determine the specific number of transitions for each state transition relationship. The transition probability value of each state transition relationship is determined based on the specific number of transitions in each state transition relationship. The state transition probability matrix is ​​determined based on the number of state types corresponding to all performance error states and the transition probability value of each state transition relationship.

[0116] It is evident that implementation Figure 3 The described system can also determine the transition probability value of each state transition relationship based on the specific number of transitions in each state transition relationship, thereby determining the state transition probability matrix result. This is beneficial to improving the comprehensiveness and rationality of the method for determining the state transition probability matrix result, and to improving the diversity and flexibility of the parameters involved in determining the state transition probability matrix result. In turn, it is beneficial to improving the accuracy and reliability of the determined state transition probability matrix result.

[0117] In another optional embodiment, the performance prediction module 302 performs corresponding fusion prediction and error analysis operations based on the future performance trend prediction results and the state transition probability matrix results to obtain the network future performance prediction results in the following ways: Based on the future performance trend prediction results, determine the second predicted value of the network key indicator type for the second current time, and determine the second actual value of the network key indicator type for the second current time. Based on the second predicted value and the second actual value, calculate the first difference, and based on the first difference and all performance error state schemes, determine the network key indicator type for the first actual performance error state at the second current moment; Based on the first actual performance error state and the state transition probability matrix results, the second predicted performance error state of the network key indicator type for the third future time is determined, and based on the second predicted value and the first error correction coefficient of the determined first actual performance error state, the third predicted value of the network key indicator type for the third future time is determined. Based on the network key indicator types, the third predicted value for the third future time and the second predicted performance error state, the network future performance prediction result for the third future time is determined.

[0118] It is evident that implementation Figure 3 The described system can also provide a method for determining the future network performance prediction results based on the preliminary prediction values ​​determined by the future performance trend prediction results. This is beneficial to improving the comprehensiveness, rationality and pertinence of the method for determining the future network performance prediction results, and thus improving the accuracy and reliability of the determined network performance prediction results.

[0119] In another optional embodiment, the performance prediction module 302 performs corresponding fusion prediction and error analysis operations based on the future performance trend prediction results and the state transition probability matrix results to obtain the network future performance prediction results in the following ways: The network key indicator type is determined for the third actual value at the third future time, and the second difference is calculated based on the third predicted value and the third actual value. Based on the second difference and the second prediction performance error state, determine whether the network key indicator types meet the preset performance direct prediction conditions for the fourth future time. When it is determined that the network key indicator type meets the direct performance prediction conditions, the second predicted performance error state is determined as the third actual performance error state of the network key indicator type for the third future time. When it is determined that the network key indicator type does not meet the direct performance prediction conditions, the third actual performance error state of the network key indicator type for the third future time is determined based on the second difference and all performance error state schemes. Based on the results of the third actual performance error state and the state transition probability matrix, the fifth predicted performance error state of the network key indicator type for the fourth future time is determined, and based on the third predicted value and the third error correction coefficient of the determined third actual performance error state, the fourth predicted value of the network key indicator type for the fourth future time is determined. Based on the network key indicator types, the fourth predicted value and the fifth predicted performance error state for the fourth future time are used to determine the network future performance prediction result for the fourth future time.

[0120] It is evident that implementation Figure 3 The described system can also provide corresponding methods for determining future network performance prediction results based on the error correction coefficients determined by the performance error state. This is beneficial for improving the comprehensiveness and rationality of the methods for determining future network performance prediction results, as well as their diversity, flexibility, and relevance. Consequently, it helps to improve the accuracy and reliability of the determined future network performance prediction results. Furthermore, it can match corresponding methods for determining actual performance error states for those that meet and those that do not meet the direct performance prediction conditions. This is beneficial for improving the diversity, flexibility, and relevance of the methods for determining actual performance error states, as well as their accuracy and reliability. This, in turn, helps to improve the accuracy and reliability of error correction based on actual performance error states, and further helps to improve the accuracy and reliability of the determined prediction values.

[0121] In yet another optional embodiment, the network key indicator types include one or more of the following: queue length type, packet loss rate type, latency type, bandwidth type, and other transmission-related indicator types; Furthermore, the routing weight optimization module 303 performs corresponding routing weight configuration optimization operations on the routing transmission path based on the network future performance prediction results, and the specific methods for obtaining the final routing weight result include: Based on the network future performance prediction results and the determined target future adjustment period, determine the average value of each sub-index type for each route transmission path. Based on the determined network performance transmission requirements, determine the configuration trade-off coefficient for each sub-index type; Based on the configuration trade-off coefficient of each sub-index type, the average index value of each sub-index type for each route transmission path, and the preset path comprehensive score determination method, the path comprehensive score result of each route transmission path is determined. The path comprehensive score result indicates the current path congestion status of the route transmission path. Based on the comprehensive path score of each route transmission path, the corresponding route priority adjustment operation is performed on all route transmission paths to obtain the final route weight result.

[0122] It is evident that implementation Figure 3 The described system can also determine the comprehensive path score of each route transmission path based on the configuration trade-off coefficient of each sub-index type and the average index of each route transmission path, and then perform route priority adjustment operation to obtain the final route weight result. This is beneficial to improving the comprehensiveness and rationality of the final route weight result determination method, and to improving the diversity and flexibility of the parameters considered in the route priority adjustment. This, in turn, is beneficial to improving the accuracy and reliability of the route priority adjustment, and thus to improving the accuracy and reliability of the determined final route weight result.

[0123] In yet another alternative embodiment, such as Figure 3 As shown, the system may also include: The judgment module 304 is used to determine whether the target network device meets the preset event-triggered routing adjustment conditions after the performance prediction module 302 performs the corresponding performance prediction operation based on standardized indicator data to obtain the network future performance prediction result; when it is determined that the target network device meets the event-triggered routing adjustment conditions, the routing weight optimization module 303 is triggered to perform the above-mentioned step of performing the corresponding routing weight configuration optimization operation on the routing transmission path based on the network future performance prediction result to obtain the final routing weight result; and / or, to determine whether the target network device meets the preset timed-triggered routing adjustment conditions; when it is determined that the target network device meets the timed-triggered routing adjustment conditions, the routing weight optimization module 303 is triggered to perform the above-mentioned step of performing the corresponding routing weight configuration optimization operation on the routing transmission path based on the network future performance prediction result to obtain the final routing weight result.

[0124] It is evident that implementation Figure 3 The described system can also trigger route weight configuration optimization operations when the target network device meets the event-triggered route adjustment conditions and / or the timed route adjustment conditions. This helps to improve the diversity, flexibility, and targeting of the triggering and execution evaluation methods for route weight configuration optimization operations, thereby improving the accuracy and timeliness of triggering and execution of route weight configuration optimization operations and reducing unnecessary resource consumption of route weight configuration optimization operations.

[0125] In another optional embodiment, the data standardization module 301 performs corresponding data standardization processing operations based on the collected raw indicator data for the target network device and key network indicator types, and the specific methods for obtaining standardized indicator data include: Based on the collected raw indicator data for the target network devices and key network indicator types, determine the raw dataset for the target sub-indicator types. The raw dataset includes at least two raw sample values. Perform the corresponding mean calculation operation on the original dataset to obtain the first mean value, and perform the corresponding standard deviation calculation operation on the original dataset to obtain the first standard deviation value; Based on the first mean and the first standard deviation, perform the corresponding standardization transformation operation on each original sample value in the original dataset to obtain the basic standardized dataset of the target sub-index type. The basic standardized dataset includes the basic standard sample value corresponding to each original sample value. Perform the corresponding mean calculation operation on the basic standardized dataset to obtain the second mean, and perform the corresponding standard deviation calculation operation on the basic standardized dataset to obtain the second standard deviation. Based on the second average value and the second standard deviation, determine whether the target sub-indicator type meets the preset data anomaly conditions; When it is determined that the target sub-indicator type meets the data anomaly conditions, the anomaly judgment value of each basic standard sample value is determined based on each basic standard sample value, the second average value, and the second standard deviation value; based on the anomaly judgment value of each basic standard sample value, the corresponding anomaly sample removal operation is performed on the basic standardized dataset to obtain the final standardized dataset of the target sub-indicator type. When it is determined that the target sub-indicator type does not meet the data anomaly conditions, the basic standardized dataset is determined as the final standardized dataset of the target sub-indicator type. Based on the final standardized dataset of the target sub-indicator type, determine the standardized indicator data.

[0126] It is evident that implementation Figure 3 The described system can also provide specific data standardization processing methods. Through basic standardized data processing operations, data anomaly detection and removal operations, the final standardized dataset is obtained. This is conducive to improving the comprehensiveness, rationality and progressiveness of data standardization processing methods, and to improving the diversity, flexibility and pertinence of data standardization sub-processing methods. In turn, it is conducive to improving the accuracy and reliability of the determined standardized indicator data. In addition, it provides specific methods for identifying and removing abnormal data samples, which is conducive to improving the accuracy and reliability of removing abnormal data samples.

[0127] Example 3 Please see Figure 4 , Figure 4 This is a schematic diagram of the structure of another 5G virtual private network artificial intelligence prediction performance and dynamic adjustment system disclosed in an embodiment of the present invention. Wherein, Figure 4The described system may include a server, which may be a local server or a cloud server; this embodiment of the invention does not limit the scope. Figure 4 As shown, the system may include: Memory 401 storing executable program code; Processor 402 coupled to memory 401; Furthermore, it may also include an input interface 403 coupled to the processor 402 and an output interface 404; The processor 402 calls the executable program code stored in the memory 401 to execute the steps in the artificial intelligence prediction performance and dynamic adjustment method for 5G virtual private network described in Embodiment 1.

[0128] Example 4 This invention discloses a computer storage medium that stores a computer program for electronic data interchange, wherein the computer program causes a computer to execute the steps in the artificial intelligence prediction performance and dynamic adjustment method for 5G virtual private networks described in Embodiment 1.

[0129] Example 5 This invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to perform the steps in the artificial intelligence prediction performance and dynamic adjustment method for 5G virtual private networks described in Embodiment 1.

[0130] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0131] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0132] Finally, it should be noted that the artificial intelligence prediction performance and dynamic adjustment method and system for 5G virtual private networks disclosed in the embodiments of this invention are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this invention.

Claims

1. A method for predicting and dynamically adjusting the performance of a 5G virtual private network using artificial intelligence, characterized in that, The method includes: Based on the collected raw indicator data for the target network devices and key network indicator types, perform the corresponding data standardization processing operations to obtain standardized indicator data; Based on the standardized index data, perform the corresponding performance prediction operation to obtain the network future performance prediction results; Based on the network's future performance prediction results, corresponding routing weight configuration optimization operations are performed on the routing transmission paths to obtain the final routing weight results, so as to achieve balanced distribution and scheduling adjustment of network traffic.

2. The method for predicting and dynamically adjusting the performance of a 5G virtual private network using artificial intelligence according to claim 1, characterized in that, The step of performing corresponding performance prediction operations based on the standardized index data to obtain the network future performance prediction results includes: Based on the standardized indicator data and the supplementary historical performance indicator data, the target data set is determined; Based on the target dataset, perform the corresponding two-parameter exponential average prediction operation to obtain the future performance trend prediction results; Based on the future performance trend prediction results, corresponding network status and performance level prediction operations are performed to obtain the future network performance prediction results, which include the future network performance status and the predicted values ​​of the network key indicator types. And, the step of performing corresponding network state and performance level prediction operations based on the future performance trend prediction results to obtain future network performance prediction results includes: Based on the future performance trend prediction results, determine the first predicted value information of the network key indicator type for the first target time period, and determine the first actual value information of the network key indicator type for the first target time period. Based on the first actual value information and the first predicted value information, perform corresponding absolute error calculation and performance error state division operations to obtain at least two performance error state schemes. Each performance error state scheme includes a performance error state and its corresponding absolute error interval. Based on the determined network performance error state data for the second target time period and the performance error state scheme, the state transition probability matrix result is determined; Based on the predicted future performance trends and the state transition probability matrix, corresponding fusion prediction and error analysis operations are performed to obtain the predicted future performance of the network.

3. The method for predicting and dynamically adjusting the performance of a 5G virtual private network using artificial intelligence according to claim 2, characterized in that, The step of performing a corresponding two-parameter exponential average prediction operation based on the target data set to obtain future performance trend prediction results includes: Based on the target dataset, determine the network performance data characteristics; Based on the network performance data characteristics and the determined network performance transmission demand information, a first value bias range for the horizontal smoothing parameter and a second value bias range for the trend smoothing parameter are determined. Based on the first value bias range, the second value bias range, and the determined prediction result accuracy requirement information, the optimal parameter combination is determined. The optimal parameter combination includes the first optimal value of the horizontal smoothing parameter and the second optimal value of the trend smoothing parameter. Based on the target data set, determine the actual observed value of the network key indicator type at the first current moment, and based on the actual observed value of the network key indicator type, the first optimal value, and the second optimal value, determine the future performance trend prediction curve of the network key indicator type. Based on the future performance trend prediction curves of the aforementioned key network indicator types, the future performance trend prediction results are determined.

4. The method for predicting and dynamically adjusting the performance of a 5G virtual private network using artificial intelligence according to claim 2, characterized in that, The step of determining the state transition probability matrix result based on the determined network performance error state data for the second target time period and the performance error state scheme includes: Perform corresponding state sequence generation operations on the determined network performance error state data for the second target time period to obtain at least two state sequence results. Each state sequence result includes the target transition time and its corresponding state transition relationship. Based on all the state sequence results, determine the total number of transitions for each performance error state that satisfies the preset initial state condition, and based on all the state sequence results, determine the specific number of transitions for each state transition relationship. Based on the specific number of transitions in each state transition relationship, the transition probability value of each state transition relationship is determined. The state transition probability matrix result is determined based on the number of state types corresponding to all the performance error states and the transition probability value of each state transition relationship.

5. The method for predicting and dynamically adjusting the performance of a 5G virtual private network using artificial intelligence according to claim 2, characterized in that, The step of performing corresponding fusion prediction and error analysis operations based on the future performance trend prediction results and the state transition probability matrix results to obtain the network future performance prediction results includes: Based on the future performance trend prediction results, determine the second predicted value of the network key indicator type for the second current time, and determine the second actual value of the network key indicator type for the second current time; Based on the second predicted value and the second actual value, calculate the first difference, and based on the first difference and all the performance error state schemes, determine the first actual performance error state of the network key indicator type for the second current time. Based on the first actual performance error state and the state transition probability matrix result, the second predicted performance error state of the network key indicator type for the third future time is determined, and based on the second predicted value and the first error correction coefficient of the determined first actual performance error state, the third predicted value of the network key indicator type for the third future time is determined. Based on the network key indicator type, the third predicted value and the second predicted performance error state for the third future time are used to determine the network future performance prediction result for the third future time. Furthermore, the method further includes: Determine the third actual value of the network key indicator type for the third future time, and calculate the second difference based on the third predicted value and the third actual value; Based on the second difference and the second prediction performance error state, it is determined whether the network key indicator type meets the preset performance direct prediction conditions for the fourth future time. When it is determined that the network key indicator type meets the performance direct prediction condition, the second predicted performance error state is determined as the third actual performance error state of the network key indicator type for the third future time. When it is determined that the network key indicator type does not meet the direct performance prediction condition, the third actual performance error state of the network key indicator type for the third future time is determined based on the second difference and all the performance error state schemes. Based on the third actual performance error state and the state transition probability matrix result, the fifth predicted performance error state of the network key indicator type for the fourth future time is determined, and based on the third predicted value and the third error correction coefficient of the determined third actual performance error state, the fourth predicted value of the network key indicator type for the fourth future time is determined. Based on the fourth predicted value for the fourth future time step and the fifth predicted performance error state of the network key indicator type, the network future performance prediction result for the fourth future time step is determined.

6. The method for predicting and dynamically adjusting the performance of a 5G virtual private network using artificial intelligence according to any one of claims 1-5, characterized in that, The network key indicator types include one or more of the following: queue length type, packet loss rate type, latency type, bandwidth type, and other transmission-related indicator types; And, the step of performing corresponding routing weight configuration optimization operations on the routing transmission path based on the network future performance prediction results to obtain the final routing weight results includes: Based on the network future performance prediction results and the determined target future adjustment period, determine the average index value of each sub-index type for each route transmission path; Based on the determined network performance transmission requirements, determine the configuration trade-off coefficient for each of the sub-index types; Based on the configuration trade-off coefficient of each sub-index type, the average index value of each sub-index type for each route transmission path, and the preset path comprehensive score determination method, the path comprehensive score result of each route transmission path is determined, and the path comprehensive score result represents the current path congestion status of the route transmission path. Based on the comprehensive path score of each route transmission path, perform corresponding route priority adjustment operations on all route transmission paths to obtain the final route weight result; And, after performing the corresponding performance prediction operation based on the standardized index data to obtain the network future performance prediction result, the method further includes: Determine whether the target network device meets the preset event-triggered route adjustment conditions; when it is determined that the target network device meets the event-triggered route adjustment conditions, execute the step of performing corresponding route weight configuration optimization operations on the route transmission path based on the network future performance prediction results to obtain the final route weight result; and / or, Determine whether the target network device meets the preset timed route adjustment conditions; when it is determined that the target network device meets the timed route adjustment conditions, execute the step of performing corresponding route weight configuration optimization operation on the route transmission path based on the network future performance prediction results to obtain the final route weight result.

7. The method for predicting and dynamically adjusting the performance of a 5G virtual private network using artificial intelligence according to any one of claims 1-5, characterized in that, The process involves performing corresponding data standardization operations based on the collected raw indicator data for the target network devices and key network indicator types to obtain standardized indicator data, including: Based on the collected raw indicator data for target network devices and key network indicator types, determine the raw dataset for target sub-indicator types, wherein the raw dataset includes at least two raw sample values; Perform the corresponding mean calculation operation on the original dataset to obtain the first mean value, and perform the corresponding standard deviation calculation operation on the original dataset to obtain the first standard deviation value; Based on the first average value and the first standard deviation, a corresponding standardization transformation operation is performed on each of the original sample values ​​in the original dataset to obtain a basic standardized dataset of the target sub-index type. The basic standardized dataset includes basic standard sample values ​​corresponding to each of the original sample values. Perform the corresponding mean calculation operation on the basic standardized dataset to obtain the second mean value, and perform the corresponding standard deviation calculation operation on the basic standardized dataset to obtain the second standard deviation value; Based on the second average value and the second standard deviation, determine whether the target sub-indicator type meets the preset data anomaly conditions; When it is determined that the target sub-indicator type meets the data anomaly conditions, an anomaly judgment value for each of the basic standard sample values ​​is determined based on each of the basic standard sample values, the second average value, and the second standard deviation value; based on the anomaly judgment value for each of the basic standard sample values, a corresponding anomaly sample removal operation is performed on the basic standardized dataset to obtain the final standardized dataset of the target sub-indicator type. When it is determined that the target sub-indicator type does not meet the data anomaly conditions, the basic standardized dataset is determined as the final standardized dataset of the target sub-indicator type. Based on the final standardized dataset of the target sub-indicator type, determine the standardized indicator data.

8. An artificial intelligence-based performance prediction and dynamic adjustment system for a 5G virtual private network, characterized in that, The system includes: The data standardization module is used to perform corresponding data standardization processing operations based on the collected raw indicator data for the target network devices and key network indicator types, to obtain standardized indicator data. The performance prediction module is used to perform corresponding performance prediction operations based on the standardized index data to obtain the network future performance prediction results. The routing weight optimization module is used to perform corresponding routing weight configuration optimization operations on the routing transmission path based on the network future performance prediction results, so as to obtain the final routing weight result, so as to realize the balanced distribution and scheduling adjustment of network traffic.

9. A system for predicting and dynamically adjusting the performance of a 5G virtual private network using artificial intelligence, characterized in that, The system includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the artificial intelligence prediction performance and dynamic adjustment method for 5G virtual private networks as described in any one of claims 1-7.

10. A computer storage medium, characterized in that, The computer storage medium stores computer instructions, which, when invoked, are used to execute the artificial intelligence prediction performance and dynamic adjustment method for the 5G virtual private network as described in any one of claims 1-7.

Images