A link optimization method, device, equipment, storage medium and program product

Abstract

The application provides a link optimization method and device, equipment, a storage medium and a program product, and relates to the technical field of network communication. The method comprises the following steps: obtaining a plurality of historical data of service link transmission and a plurality of target features extracted from the plurality of historical data; performing feature extraction on the plurality of historical data to obtain feature data of each historical data at each time point; constructing a model according to the feature data of each historical data at each time point to obtain a feature model corresponding to each historical data; predicting prediction data of each feature model in a second preset time period in the future, and adjusting the plurality of target features according to the plurality of prediction data. Individual modeling and prediction are performed on each historical data, the change trend and subtle differences in the data can be flexibly captured, the target features can be effectively adjusted, the prediction accuracy and reliability are improved, the complex link transmission environment can be better adapted to, and the optimization effect of the link is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of network communication, and in particular to a link optimization method, device, equipment, storage medium and program product. BACKGROUND

[0002] Link optimization refers to optimizing data transmission paths and resource configurations through a series of techniques and methods, such as dynamic routing protocols, load balancing, spectrum management, edge caching, and transmission protocol optimization, to improve transmission efficiency, reduce latency, reduce packet loss rate, and improve overall system performance, so as to ensure that data can be transmitted and processed in the network or system in the most optimal way.

[0003] In the prior art, the target features extracted from the historical data of the link are optimized in various ways, including using load balancers to distribute traffic, caching strategies to reduce duplicate data transmission, data compression and encryption to improve transmission efficiency, service discovery and dynamic routing to achieve dynamic traffic management, and network protocol optimization and edge computing to reduce latency. By adjusting the target features of the link, higher bandwidth utilization, lower latency, and lower error rate can be achieved, thereby improving the performance and reliability of network transmission as a whole.

[0004] However, the prior art has slow adjustment speed when facing sudden demand fluctuations or resource constraints, making it difficult to quickly realize the optimization and reorganization of service links, affecting service quality and user experience. At the same time, the prior art relies on preset rules or simple algorithms and lacks the support of artificial intelligence technologies such as deep learning and reinforcement learning, with limited intelligence level, making it difficult to achieve accurate and personalized link optimization. SUMMARY

[0005] The present application provides a link optimization method, device, equipment, storage medium and program product to solve the problem of low accuracy and low personalization of the link optimization method in the prior art.

[0006] In a first aspect, the present application provides a link optimization method, comprising:

[0007] Obtaining a plurality of historical data transmitted by a service link, and a plurality of target features extracted from the plurality of historical data; wherein each historical data refers to data transmitted by the service link in a first preset time period in the past;

[0008] Feature extraction is performed on a plurality of historical data to obtain feature data of each historical data at each time point;

[0009] A model is constructed according to the feature data of each historical data at each time point to obtain a feature model corresponding to each historical data;

[0010] predicting prediction data of each feature model in a second future preset time period, and adjusting the target features according to the prediction data.

[0011] In a possible design, the historical data of the service link transmission and the target features extracted from the historical data include:

[0012] The historical data of the service link transmission are obtained, and original features of each historical data are extracted;

[0013] Each historical data is substituted into a preset regression model to obtain a regression coefficient of each historical data;

[0014] Each original feature corresponding to a regression coefficient greater than a preset value is determined as the target feature.

[0015] In a possible design, the feature extraction on the historical data to obtain feature data of each historical data at each time point includes:

[0016] The historical data are preprocessed to obtain preprocessed data corresponding to each historical data;

[0017] The feature extraction is performed on each preprocessed data at each time point to obtain feature data of each historical data at each time point.

[0018] In a possible design, the adjustment of the target features according to the prediction data includes:

[0019] Anomaly analysis is performed on each prediction data to obtain an improvement point of each prediction data;

[0020] The adjustment amplitude of each prediction data is calculated according to the improvement point of each prediction data;

[0021] The target features are adjusted according to the adjustment amplitudes of the prediction data.

[0022] In a possible design, after the adjustment of the target features according to the prediction data, the method further includes:

[0023] Feedback data are obtained, and each feedback data is data of the service link transmission after a second future preset time period;

[0024] The deviation between each prediction data and each feedback data is calculated;

[0025] According to the plurality of deviations, the plurality of feature models are iteratively optimized to optimize the performance of the plurality of feature models.

[0026] In a possible design, the target model is any one of the plurality of feature models.

[0027] For the target model, the iteratively optimizing the plurality of feature models according to the plurality of deviations comprises:

[0028] According to at least one of the deviations corresponding to the target model, adjusting a parameter of the target model;

[0029] According to the error of the adjusted target model, continuously adjusting the parameter until the error meets a preset criterion.

[0030] In a second aspect, the present application provides a link optimization device, comprising:

[0031] An acquisition module is configured to acquire a plurality of historical data transmitted by a service link and a plurality of target features extracted from the plurality of historical data, wherein each of the historical data refers to data transmitted by the service link in a first preset time period in the past.

[0032] A feature extraction module is configured to perform feature extraction on the plurality of historical data to obtain feature data of each of the historical data at each time point.

[0033] A model construction module is configured to perform model construction according to the feature data of each of the historical data at each time point to obtain a feature model corresponding to each of the historical data.

[0034] An adjustment module is configured to predict prediction data of each of the feature models in a second preset time period in the future, and adjust the plurality of target features according to the plurality of prediction data to optimize the performance of the service link.

[0035] In a possible design, the acquisition module comprises:

[0036] An acquisition unit is configured to acquire the plurality of historical data transmitted by the service link and extract original features of each of the historical data.

[0037] A first calculation unit is configured to substitute each of the historical data into a preset regression model to obtain a regression coefficient of each of the historical data.

[0038] A determination unit is configured to determine each of the original features with a corresponding regression coefficient greater than a preset value as the target feature.

[0039] In a possible design, the feature extraction module comprises:

[0040] The preprocessing unit is configured to preprocess the historical data to obtain preprocessing data corresponding to each of the historical data.

[0041] The feature extraction unit is configured to extract features of each of the preprocessing data at each of the time points to obtain feature data of each of the historical data at each of the time points.

[0042] In a possible design, the adjustment module comprises:

[0043] The anomaly analysis unit is configured to perform anomaly analysis on each of the prediction data to obtain an improvement point of each of the prediction data.

[0044] The second calculation unit is configured to calculate an adjustment amplitude of each of the prediction data according to the improvement point of each of the prediction data.

[0045] The adjustment unit is configured to adjust the target features according to the adjustment amplitudes of the prediction data.

[0046] In a possible design, the link optimization apparatus further comprises:

[0047] The second acquisition module is configured to acquire a plurality of feedback data, each of the feedback data being data transmitted by the service link after a second preset time period in the future.

[0048] The third calculation module is configured to calculate a deviation between each of the prediction data and each of the feedback data.

[0049] The optimization module is configured to iteratively optimize the feature models according to the deviations, so as to optimize the performance of the feature models.

[0050] In a possible design, the target model is any one of the feature models.

[0051] For the target model, the optimization module comprises:

[0052] The first adjustment unit is configured to adjust a parameter of the target model according to at least one of the deviations corresponding to the target model.

[0053] The second adjustment unit is configured to continuously adjust the parameter according to an error of the adjusted target model until the error meets a preset standard.

[0054] In a third aspect, the present application provides an electronic device, comprising a processor and a memory connected with the processor in communication.

[0055] The memory stores computer-executed instructions;

[0056] When the processor executes computer execution instructions stored in the memory, it is used to implement a link optimization method according to the first aspect of the invention.

[0057] Fourthly, this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement a link optimization method according to the first aspect of the invention.

[0058] Fifthly, this application provides a computer program product, including a computer program, which, when executed by a processor, is used to implement a link optimization method according to the first aspect of the invention.

[0059] This application provides a link optimization method, apparatus, device, storage medium, and program product. The method includes: acquiring multiple historical data transmitted via a service link, and extracting multiple target features from the multiple historical data; wherein each historical data refers to data transmitted by the service link in the past first preset time period; extracting features from the multiple historical data to obtain feature data of each historical data at each time point; constructing a model based on the feature data of each historical data at each time point to obtain a feature model corresponding to each historical data; predicting the predicted data of each feature model in the future second preset time period, and adjusting the multiple target features based on the multiple predicted data to optimize the performance of the service link. The link optimization method of this application achieves more refined and accurate predictions by constructing a feature model for each historical data and using these feature models to predict future data. This method can fully utilize the unique features of each historical data, avoiding the generalization error and inaccurate prediction problems that may occur when using a single feature model in traditional methods. By performing personalized modeling and prediction for each historical data, it is possible to more flexibly capture the changing trends and subtle differences in the data, thereby more effectively adjusting the target features and optimizing the performance of the service link. This method improves the accuracy and reliability of predictions, better adapts to complex and ever-changing link transmission environments, and ultimately enhances the overall link optimization effect. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0061] Figure 1 An application scenario of the link optimization method provided in the embodiments of the present application is shown in the figure.

[0062] Figure 2 A flowchart of the link optimization method provided in the embodiments of the present application is shown in the figure. Figure 1 ;

[0063] Figure 3 A flowchart of the link optimization method provided in the embodiments of the present application is shown in the figure. Figure 2 ;

[0064] Figure 4 A structure diagram of the link optimization device provided in the embodiments of the present application is shown in the figure.

[0065] Figure 5 An electronic device provided in the embodiments of the present application. DETAILED DESCRIPTION

[0066] The exemplary embodiments will be described in detail herein with reference to the attached drawings. When the description below refers to the drawings, the same numbers in different drawings refer to the same or similar elements unless otherwise described. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of apparatuses and methods consistent with some aspects of the present application as detailed in the appended claims.

[0067] In the embodiments of the present application, the terms such as "first", "second", etc. are used to distinguish the same or similar items with substantially the same functions and effects. Those skilled in the art can understand that the terms such as "first", "second", etc. do not limit the quantity and execution order, and the terms such as "first", "second", etc. do not necessarily mean different. It should be noted that the terms such as "exemplary" or "for example" in the embodiments of the present application are used to represent an example, illustration or description. Any embodiment or design scheme described as "exemplary" or "for example" in the present application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. Rather, the terms such as "exemplary" or "for example" are intended to present the relevant concept in a specific manner. In the embodiments of the present application, "at least one" means one or more, and "multiple" means two or more.

[0068] It should be noted that "at" in the embodiments of the present application can be the moment when a certain condition occurs, or a period of time after a certain condition occurs, which is not specifically limited in the embodiments of the present application. In addition, the link optimization method provided in the embodiments of the present application is only an example, and the link optimization method can include more or less content.

[0069] For the sake of clear description of the technical solutions of the embodiments of the present application, the following will briefly introduce some terms and technologies involved in the embodiments of the present application:

[0070] Link: refers to a physical or logical channel used to connect two or more devices or nodes in a communication system. The main function of the link is to transmit data, so that different devices or nodes can communicate with each other. The performance and quality of the link directly affect the speed, reliability and stability of data transmission.

[0071] Link optimization method: a technology for improving the performance and efficiency of network communication. It involves adjusting and optimizing various parameters of network links to achieve higher transmission speed, lower delay, higher reliability and better resource utilization.

[0072] Regression model: an analysis method used to study the relationship between the dependent variable and one or more independent variables. The main purpose of the regression model is to predict the value of the dependent variable through the independent variable, or to understand the influence of the independent variable on the dependent variable.

[0073] Regression coefficient: a parameter in the regression model, used to quantify the influence of the independent variable on the dependent variable. In regression analysis, the regression coefficient represents the degree of contribution of each independent variable to the dependent variable. Specifically, the regression coefficient reflects the expected change in the dependent variable when an independent variable increases by one unit.

[0074] Hereinafter, exemplary embodiments will be described in detail, examples of which are shown in the accompanying drawings. In the following description, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0075] The technical solutions of the present application will be described in detail below with specific examples. The following specific examples can be combined with each other, and the same or similar concepts or processes may not be described again in some examples. The embodiments of the present application will be described below with reference to the accompanying drawings.

[0076] In order to clearly understand the technical solutions of the present application, first, the prior art solutions will be described in detail. Link refers to a path or channel used to transmit data in a communication network, and the link can be wired or wireless. During the link data transmission process, the link may exhibit various phenomena, such as bandwidth limitation, delay, jitter, data packet loss and signal attenuation, etc. These phenomena will affect the efficiency and quality of data transmission, resulting in unstable communication or reduced speed. Therefore, the link needs to be optimized.

[0077] Currently, the optimization method of the link is usually to optimize the link by using load balancer, cache strategy, data compression and other methods. However, the prior art has slow adjustment speed when facing sudden demand fluctuation or resource constraint, and it is difficult to quickly realize the optimization and reorganization of the service link, which affects the service quality and user experience. At the same time, the prior art relies on preset rules or simple algorithms, lacks the support of artificial intelligence technologies such as deep learning and reinforcement learning, and the intelligent level is limited, so it is difficult to realize accurate and personalized link optimization. Therefore, the current link optimization method has the problems of low accuracy and low personalization.

[0078] Therefore, in order to solve the problem of low accuracy and low personalization of the link optimization method in the prior art, it is found in research that it is necessary to adjust the optimization strategy according to different application scenarios and user needs: ① by analyzing a plurality of historical data, extracting key features, and constructing a plurality of feature models; ② using artificial intelligence technologies such as deep learning and reinforcement learning to dynamically adjust link parameters to realize intelligent and adaptive optimization; ③ not all features in the plurality of features of the link need to be optimized, and the features that have a greater impact on the link can be calculated to adjust only these features.

[0079] Specifically:

[0080] A feature model can be constructed for each historical data to predict the trend of the data using the feature model, and the feature corresponding to the historical data is adjusted according to the trend; after adjusting the features of the link, the feature model can be iteratively optimized using the current data of the link to enhance the performance of the feature model; the importance or correlation of each feature of the link can be calculated, and the feature with high importance or correlation is taken as the target feature, and only the target feature of the link is optimized.

[0081] The link optimization method of the embodiment of the application realizes more fine and accurate prediction by constructing a unique feature model for each historical data and predicting future data using these models. This method makes full use of the unique features of each historical data, avoids the generalization error and inaccurate prediction problem that may be caused by a single model. Personalized modeling and prediction can flexibly capture the trend and subtle differences in the data, thereby more effectively adjusting the target feature and improving the accuracy and personalization of the prediction, and optimizing the performance of the link.

[0082] Based on the above creative findings, the technical scheme of the application is proposed.

[0083] The application scenario of the link optimization method provided by the embodiment of the application is introduced below. Figure 1 The application scenario of the link optimization method provided by the embodiment of the application is introduced below. Figure 1As shown in the figure, 101 is a server, 102 is an electronic device, 103 is a link, and the server 101 and the electronic device 102 transmit data through the link 103. During the data transmission process, the electronic device 102 optimizes the link 102 according to the transmitted data.

[0084] The embodiments of the present application will be described below with reference to the accompanying drawings.

[0085] Figure 2 The flowchart of the link optimization method provided by the embodiments of the present application Figure 1 . As shown in the figure, in this embodiment, the execution subject of the embodiments of the present application is an electronic device. The link optimization method provided by this embodiment includes the following steps: Figure 2

[0086] S201, obtaining a plurality of historical data transmitted by a service link and a plurality of target features extracted from the plurality of historical data; wherein each historical data refers to data transmitted by the service link in a past first preset time period.

[0087] Specifically, a data collector is set on the service link to collect the transmission data of the link in a certain specific time period (i.e. the first preset time period) regularly. These data can include but are not limited to delay, bandwidth usage, packet loss rate and other key performance indicators. Store these transmission data in a database, and use data analysis and processing technology to extract a plurality of target features from these historical data. These target features can be the results obtained after analyzing, feature engineering or clustering analysis of the original data, such as average delay, maximum bandwidth usage, packet loss rate distribution, etc.

[0088] S202, performing feature extraction on the plurality of historical data to obtain feature data of each historical data at each time point.

[0089] Specifically, the collected plurality of historical data is usually time series data, which can include timestamp, delay, bandwidth usage and packet loss rate, etc. Use feature extraction algorithm to process these raw data. For the data at each time point, calculate its features in a certain window, such as average delay, maximum bandwidth usage and packet loss rate in the past one minute, etc. Through these feature extraction methods, the original time series data can be converted into a set of feature data, which can better describe the state of the service link at each time point.

[0090] S203, constructing a model according to the feature data of each historical data at each time point to obtain a feature model corresponding to each historical data.

[0091] ​Specifically, appropriate modeling methods and algorithms are selected, such as time series analysis models, machine learning models, or deep learning models. Feature data at each time point is taken as input, and these modeling methods are used to train the data. During the training process, techniques such as cross-validation and grid search can be used to optimize model parameters, ensuring the generalization ability and prediction accuracy of the model. For each historical data set, a feature model is constructed that can capture the patterns and rules in the data. Finally, the trained and validated model can accurately describe the feature relationship of each historical data at different time points, forming a feature model corresponding to each historical data.

[0092] S204, predicting the prediction data of each feature model in the future second preset time period, and adjusting the target features according to the multiple prediction data to optimize the performance of the service link.

[0093] Specifically, input the time points in the future second preset time period into the feature model, and the model will make predictions based on its internal rules and patterns. The model will generate prediction data for the future second preset time period. These prediction data may include future delay, bandwidth usage, and packet loss rate, among other key performance indicators. Analyze the prediction data to identify key factors and potential problems that may affect the performance of the service link, such as predicted high latency or bandwidth bottlenecks. Based on these analysis results, develop appropriate optimization strategies, which may include dynamically adjusting bandwidth allocation, optimizing routing paths, and adjusting service priorities. Through simulation and verification, the optimal adjustment scheme is selected and applied to the actual service link. Through the above steps, the target features can be effectively adjusted using prediction data, thereby optimizing the overall performance of the service link.

[0094] The embodiment provides a link optimization method, which comprises the following steps: obtaining a plurality of historical data transmitted by a service link and a plurality of target features extracted from the plurality of historical data; wherein each historical data refers to data transmitted by the service link in a first preset time period in the past; feature extraction is performed on the plurality of historical data to obtain feature data of each historical data at each time point; a feature model corresponding to each historical data is obtained by performing model construction according to the feature data of each historical data at each time point; prediction data of each feature model in a second preset time period in the future is predicted, and the plurality of target features are adjusted according to the plurality of prediction data, so as to optimize the performance of the service link. The link optimization method realizes the following technical effects: by constructing a feature model for each historical data respectively and predicting future data by using the models respectively, more refined and accurate prediction is realized. The method fully utilizes the unique features of each historical data, avoids the generalization error and inaccurate prediction caused by using a single feature model in the traditional method. The method can more effectively adjust the target features, optimize the performance of the service link, improve the prediction accuracy and reliability, better adapt to the complex and changeable link transmission environment, and improve the overall link optimization effect.

[0095] In a possible design, S201, obtaining a plurality of historical data transmitted by a service link and a plurality of target features extracted from the plurality of historical data, comprises the following steps:

[0096] S2011, obtaining a plurality of historical data transmitted by a service link and extracting original features of each historical data.

[0097] Specifically, by using a network monitoring tool or a log system, transmission data of the link in a first preset time period is continuously recorded, and the data can include timestamps, delays, bandwidth usage rates and packet loss rates and other key performance indicators. By using a data processing tool or writing a script, original features of each data point are extracted from the stored historical data, and the original features can be basic indicators directly obtained from the original data, such as delay values, bandwidth usage rate values and packet loss rate values at each time point.

[0098] S2012, substituting each historical data into a preset regression model to obtain a regression coefficient of each historical data.

[0099] Specifically, an appropriate regression model is selected and constructed, which can be linear regression or ridge regression. The original features of each historical data point are extracted and used as input variables, then substituted into the pre-defined regression model for training. During training, the model estimates parameters based on the input original feature data and the corresponding target variable (e.g., link performance metrics), calculating the regression coefficient for each feature. The regression coefficients reflect the degree of influence of each feature on the target variable. Through iterative optimization algorithms (such as gradient descent), the model continuously adjusts the regression coefficients to minimize prediction error. Finally, after training is complete, the model outputs the regression coefficient for each historical data point.

[0100] S2013. Each original feature whose corresponding regression coefficient is greater than the preset value is identified as the target feature.

[0101] Specifically, after training the regression model and obtaining the regression coefficients for each historical data point, a preset value is set as a threshold. This threshold is used to filter out features that have a significant impact on the target variable (such as link performance metrics). The regression coefficients for each historical data point are iterated through, and each coefficient is compared to the preset value. If the regression coefficient of a certain original feature is greater than the preset value, then that feature is considered to have a significant impact on the target variable and is identified as the target feature. In this way, a set of features most critical to link performance optimization can be selected, while features with less impact or irrelevant information are removed, thereby simplifying the model and improving the accuracy and efficiency of prediction and optimization.

[0102] The technical effect of this solution in this embodiment is that it filters out features that have a significant impact on link performance through regression analysis, removes irrelevant or less influential features, thereby simplifying the model, improving the accuracy and efficiency of prediction and optimization, and ultimately optimizing the performance of the service link.

[0103] In one possible design, S202, feature extraction is performed on multiple historical data sets to obtain feature data for each historical data set at each time point, including:

[0104] S2021. Preprocess multiple historical data to obtain preprocessed data corresponding to each historical data.

[0105] Specifically, the collected historical data may contain noise, missing values or outliers, which need to be addressed in the preprocessing stage. The data is cleaned to remove or correct missing values and outliers; the data is normalized or standardized to eliminate differences between different feature dimensions and ensure that the data is compared on the same scale; if the data has a timestamp, time alignment and synchronization processing can be performed to ensure the time consistency of the data. In addition, data smoothing processing can also be performed, such as using a moving average or filter to reduce random fluctuations in the data. Through the above preprocessing steps, the original historical data can be converted into pretreated data with higher quality and uniform format, providing a reliable data basis for subsequent feature extraction and model construction.

[0106] S2022, respectively, at each time point, feature extraction is performed on each pretreated data to obtain feature data of each historical data at each time point.

[0107] Specifically, first, the pretreated historical data is segmented according to time points, and each time point corresponds to a group of data. For the data at each time point, a feature extraction method is applied to extract key features. Through these methods, feature data that can reflect the link performance can be extracted from the data at each time point. Finally, the feature data extracted at each time point is summarized to form complete feature data, which provides input for subsequent model construction and prediction.

[0108] The technical effect of this scheme in this embodiment is that the data quality is improved through the preprocessing step, noise and outliers are eliminated, and key features are extracted at each time point to ensure that the feature data accurately reflects the link state, thereby providing high-quality input data for subsequent model construction and prediction, and improving the accuracy and effect of link optimization

[0109] Figure 3 Flowchart of the link optimization method provided by the embodiments of the present application Figure 2 In this embodiment, in the link optimization method provided by the embodiments of the present application, Figure 2 Based on the embodiments provided, the link optimization method is further explained. The link optimization method comprises:

[0110] S301, obtaining a plurality of historical data transmitted by a service link, and a plurality of target features extracted from the plurality of historical data; wherein each historical data refers to data transmitted by the service link in a first preset time period in the past.

[0111] S302, feature extraction is performed on the plurality of historical data to obtain feature data of each historical data at each time point.

[0112] S303, constructing a model according to the feature data of each historical data at each time point to obtain a feature model corresponding to each historical data.

[0113] S304, predicting prediction data of each feature model in a future second preset time period.

[0114] S301-S304 are similar to S201-S204, and the embodiment will not be described again.

[0115] S305, performing anomaly analysis on each prediction data to obtain an improvement point of each prediction data.

[0116] Specifically, first, define the standards and indicators of anomaly analysis, such as setting a threshold range or a normal range of historical data. Compare each prediction data with these standards to identify abnormal data points that exceed the normal range or do not meet expectations. For each detected abnormal data point, further analyze its specific causes and influencing factors to determine the aspects that need to be improved, i.e., improvement points, which may include performance bottlenecks in specific time periods or abnormal fluctuations in specific features, etc. In this way, anomalies in prediction data can be systematically identified and located, providing targeted improvement suggestions to optimize the performance of service links.

[0117] S306, calculating an adjustment amplitude of each prediction data according to the improvement point of each prediction data.

[0118] Specifically, for each prediction data improvement point, determine its specific abnormal type and severity, such as excessively high delay, insufficient bandwidth, or abnormal packet loss rate, etc. Set the calculation rules of the adjustment amplitude, and then use these rules to quantitatively analyze each improvement point to calculate the specific amplitude that needs to be adjusted. For example, if the delay at a certain time point exceeds the normal range by 10%, the amount of delay that needs to be reduced can be calculated. In this way, a specific adjustment amplitude can be determined for each prediction data improvement point, providing clear guidance for subsequent optimization measures, ultimately achieving effective adjustment of multiple target features and optimizing the performance of service links.

[0119] S307, adjusting multiple target features according to the adjustment amplitudes of multiple prediction data.

[0120] Specifically, aggregate the adjustment amplitudes of each prediction data to determine the specific values that each target feature needs to adjust. Then, based on these adjustment amplitudes, develop specific optimization strategies, such as adjusting network configuration parameters, allocating more bandwidth, optimizing routing paths, or adjusting service quality settings, etc. Then, gradually implement these optimization strategies and monitor the adjusted link performance in real time to ensure that the adjustment measures are effective and do not introduce new problems. In this way, multiple target features can be accurately adjusted according to the adjustment amplitudes of prediction data, thereby optimizing the overall performance of service links.

[0121] S308, obtain a plurality of feedback data, each feedback data is used to indicate the data transmitted by the service link after a second preset time period in the future.

[0122] Specifically, after the optimization and adjustment of the service link, the transmission data of the link after a second preset time period in the future is continuously monitored and recorded. These data include the actual running conditions of the link, such as delay, bandwidth utilization and packet loss rate and other performance indicators. In this way, the actual performance of the link after adjustment can be obtained, thereby providing a real and reliable data basis for subsequent deviation analysis and model optimization.

[0123] S309, respectively calculate the deviation between each prediction data and each feedback data.

[0124] Specifically, the feedback data actually transmitted after a second preset time period in the future is collected and sorted, which should correspond to the previously predicted data in time and features. Each prediction data is matched with its corresponding feedback data. Then, the difference between each pair of prediction data and feedback data is calculated, which can be realized by simple subtraction operation, i.e. deviation = feedback data - prediction data. For different features, their deviation values can be calculated respectively, such as delay, bandwidth utilization and packet loss rate. In this way, the difference between the prediction data and the actual data can be quantified, which provides the basis for subsequent model optimization.

[0125] S310, according to a plurality of deviations, iteratively optimize a plurality of feature models to optimize the performance of the plurality of feature models.

[0126] Specifically, the deviation between each prediction data and its corresponding feedback data is collected and sorted, which reflects the difference between the model prediction and the actual situation. Analyze the distribution and trend of these deviations to identify the shortcomings of the model in certain features or time points. Then, use these deviation information to iteratively optimize the feature model by adjusting the model parameters, introducing new features or improving the feature extraction method. Optimization algorithms in machine learning can be used to gradually reduce the deviation and improve the prediction accuracy of the model. Through multiple iterations and optimization, the performance of the feature model is gradually improved, so that it can more accurately reflect the actual situation of the service link, thereby realizing more effective link optimization.

[0127] The technical effect of this scheme in this embodiment is: by systematically identifying and quantifying the anomalies in the prediction data, clear adjustment guidance is provided, thereby realizing precise optimization of the target feature. This not only improves the stability and efficiency of the link performance, but also responds and corrects potential problems in a timely manner, ensuring the reliability and efficient operation of the service link; by introducing feedback data and bias analysis, errors in the prediction model can be identified and corrected, gradually improving the accuracy and reliability of the model, thereby realizing more accurate link optimization and improving the overall performance of the service link.

[0128] In a possible design, the target model is any one of the plurality of feature models.

[0129] For the target model, S310 iteratively optimizes the plurality of feature models according to the plurality of biases, including:

[0130] S3101, according to at least one bias corresponding to the target model, adjusting the parameters of the target model.

[0131] Specifically, by calculating the gradient of the bias on the model parameters, the direction and amplitude of the parameter adjustment are determined, and then the model parameters are updated to reduce the prediction error. Repeat this process to gradually adjust the model parameters until the prediction error of the model meets the preset standard or reaches a minimum value. Through this iterative optimization, the accuracy and performance of the target model can be significantly improved, thereby more effectively optimizing the service link.

[0132] S3102, according to the error of the adjusted target model, continuously adjusting the parameters until the error meets the preset standard.

[0133] Specifically, after each parameter adjustment, the error of the target model is calculated, which is usually achieved by comparing the model prediction value with the actual feedback data, and the error can use indicators such as mean square error and mean absolute error. Calculate the gradient of the error on the model parameters using optimization algorithms to determine the direction and amplitude of the parameter adjustment. Then, update the model parameters according to the gradient information to reduce the error. Repeat this process to continuously monitor and calculate the error of the adjusted model, gradually optimize the model parameters. This iterative process continues until the model error falls below the preset standard or reaches a minimum value, thereby ensuring that the prediction accuracy and performance of the model meet the optimization requirements.

[0134] The technical effect of this scheme in this embodiment is: by continuously iterating and optimizing the model parameters, gradually reducing the prediction error, and improving the accuracy and reliability of the model. This iterative optimization process can significantly improve the performance of the feature model, thereby realizing more accurate link optimization.

[0135] Figure 4 The structure diagram of the link optimization device provided by the embodiment of the present application is shown in FIG. 1. Figure 4As shown, the link optimization apparatus comprises:

[0136] The acquisition module 401 is configured to acquire a plurality of historical data transmitted by the service link, and extract a plurality of target features from the plurality of historical data; each historical data is data transmitted by the service link in a first preset time period in the past.

[0137] The feature extraction module 402 is configured to perform feature extraction on the plurality of historical data to obtain feature data of each historical data at each time point.

[0138] The model construction module 403 is configured to construct a model according to the feature data of each historical data at each time point to obtain a feature model corresponding to each historical data.

[0139] The adjustment module 404 is configured to predict prediction data of each feature model in a second preset time period in the future, and adjust the plurality of target features according to the plurality of prediction data to optimize the performance of the service link.

[0140] In a possible design, the acquisition module 401 comprises:

[0141] The acquisition unit is configured to acquire a plurality of historical data transmitted by the service link, and extract original features of each historical data.

[0142] The first calculation unit is configured to substitute each historical data into a preset regression model to obtain a regression coefficient of each historical data.

[0143] The determination unit is configured to determine each original feature corresponding to a regression coefficient greater than a preset value as a target feature.

[0144] In a possible design, the feature extraction module 402 comprises:

[0145] The preprocessing unit is configured to perform preprocessing on the plurality of historical data to obtain preprocessing data corresponding to each historical data.

[0146] The feature extraction unit is configured to perform feature extraction on each preprocessing data at each time point to obtain feature data of each historical data at each time point.

[0147] In a possible design, the adjustment module 404 comprises:

[0148] The anomaly analysis unit is configured to perform anomaly analysis on each prediction data to obtain an improvement point of each prediction data.

[0149] The second calculation unit is configured to calculate an adjustment amplitude of each prediction data according to the improvement point of each prediction data.

[0150] The adjusting unit is configured to adjust the target features according to adjustment amplitudes of the plurality of prediction data.

[0151] In a possible design, the link optimization apparatus further includes:

[0152] The second obtaining module is configured to obtain a plurality of feedback data, each of the feedback data being indicative of data transmitted by the service link after a second preset time period in the future.

[0153] The third calculating module is configured to calculate a deviation between each prediction data and each feedback data, respectively.

[0154] The optimization module is configured to iteratively optimize the plurality of feature models according to the plurality of deviations, so as to optimize performance of the plurality of feature models.

[0155] In a possible design, the target model is any one of the plurality of feature models.

[0156] For the target model, the optimization module includes:

[0157] The first adjusting unit is configured to adjust a parameter of the target model according to at least one deviation corresponding to the target model.

[0158] The second adjusting unit is configured to continuously adjust the parameter according to an error of the adjusted target model until the error meets a preset criterion.

[0159] The link optimization apparatus provided in this embodiment can perform Figure 2-3 The technical solutions in the method embodiments shown in the table are similar to those in the apparatus embodiments shown in the table, and thus, repeated descriptions are not given here. Figure 2-3 The technical solutions in the method embodiments shown in the table are similar to those in the apparatus embodiments shown in the table, and thus, repeated descriptions are not given here.

[0160] Figure 5 An electronic device is provided for the embodiments of this application. As shown in the table, the electronic device includes at least one processor 510 and a memory 520. The electronic device further includes a communication component 530. The processor 510, the memory 520, and the communication component 530 are connected through a bus 540. Figure 5 In the specific implementation process, the at least one processor 510 executes computer execution instructions stored in the memory 520, so that the at least one processor 510 is configured to implement the link optimization method of the above-mentioned embodiments.

[0161] The specific implementation process of the processor 510 can refer to the method embodiments described above, and the implementation principles and technical effects are similar, which will not be described here again in this embodiment.

[0162]

[0163] ​In the above embodiments, it should be understood that the processor 510 can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in combination with the application can be directly embodied as hardware processor execution, or executed by a combination of hardware and software modules in the processor.

[0164] The memory 520 can include a high-speed RAM memory, and can also include a non-volatile storage NVM, such as at least one disk memory.

[0165] The bus 540 can be an industry standard architecture (ISA) bus, a peripheral component (PCI) bus, or an extended industry standard architecture (EISA) bus, etc. The bus 540 can be divided into an address bus, a data bus, a control bus, etc. For the sake of representation, the bus 540 in the drawings of the present application does not limit to only one bus or one type of bus.

[0166] The functions realized by the electronic device and the master device described above are introduced for the scheme provided by the embodiments of the present application. It can be understood that the electronic device or the master device contains the hardware structure and / or software module corresponding to the execution of each function in order to realize the above functions. The units and algorithm steps of each example described in combination with the embodiments disclosed in the embodiments of the present application can be realized in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed by hardware or computer software driven hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the technical solution of the embodiments of the present application.

[0167] The embodiments of the present application also provide a computer readable storage medium, and the computer readable storage medium stores computer execution instructions. When the computer execution instructions are executed by a processor, a link optimization method of the above embodiments is implemented. In the specific implementation of the above multi-protocol label switching packet transmission method, each module can be implemented as a processor.

[0168] The readable storage medium can be implemented by any type of volatile or nonvolatile storage devices or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk. The readable storage medium can be any available medium that can be accessed by a general or special purpose computer.

[0169] An exemplary readable storage medium is coupled to the processor, so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be an integral part of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (ASIC). Of course, the processor and the readable storage medium can also exist as discrete components in an electronic device or a host device.

[0170] The embodiments of the present application also provide a computer program product, comprising a computer program, which is executed by a processor to implement the link optimization method of the above-mentioned embodiments.

[0171] The computer program is stored in a readable storage medium, and at least one processor can read the computer program from the readable storage medium, and execute the computer program to perform the scheme provided by any of the above-mentioned embodiments.

[0172] Those skilled in the art can understand that all or part of the steps of the above-mentioned method embodiments can be completed by program instruction related hardware. The foregoing program can be stored in a computer readable storage medium. The program is executed to perform the steps of the above-mentioned method embodiments; and the foregoing storage medium includes ROM, RAM, magnetic disk or optical disk and various storage media that can store program codes.

[0173] So far, the technical scheme of the present application has been described in combination with the preferred embodiments shown in the drawings, but those skilled in the art can easily understand that the protection scope of the present application is obviously not limited to these specific embodiments, and the above embodiments are only used to illustrate the technical scheme of the present application, but not to limit it; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical scheme recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical scheme deviate from the scope of the technical scheme of the embodiments of the present application.

Claims

1. A link optimization method, characterized in that, include: Acquire multiple historical data points transmitted through the service link and extract the original features of each historical data point; Substitute each historical data point into a preset regression model to obtain the regression coefficient for each historical data point; identify each original feature whose corresponding regression coefficient is greater than a preset value as a target feature; wherein, each historical data point refers to the data transmitted by the service link in the past first preset time period; Feature extraction is performed on multiple historical data sets to obtain feature data for each historical data set at each time point; Based on the feature data of each historical data point at each time point, a model is constructed to obtain the feature model corresponding to each historical data point; Predict the predicted data for each of the feature models within a second preset time period in the future, and adjust the multiple target features based on the multiple predicted data to optimize the performance of the service link.

2. The link optimization method according to claim 1, characterized in that, The step of extracting features from multiple historical data sets to obtain feature data for each historical data set at each time point includes: The historical data is preprocessed to obtain preprocessed data corresponding to each historical data. At each of the aforementioned time points, feature extraction is performed on each of the preprocessed data to obtain the feature data of each of the historical data at each time point.

3. The link optimization method according to claim 1, characterized in that, The step of adjusting the multiple target features based on the multiple predicted data includes: Anomaly analysis is performed on each of the predicted data to identify areas for improvement. Calculate the adjustment range for each of the predicted data points based on the improvement points of each predicted data point; The target features are adjusted based on the adjustment magnitude of the predicted data.

4. The link optimization method according to any one of claims 1 to 3, characterized in that, After adjusting the multiple target features based on the multiple predicted data, the method further includes: Multiple feedback data are obtained, each of which refers to the data transmitted by the service link after a second preset time period in the future; Calculate the deviation between each of the predicted data and each of the feedback data; Based on the multiple deviations, the multiple feature models are iteratively optimized to improve the performance of the multiple feature models.

5. The link optimization method according to claim 4, characterized in that, The target model is any one of the multiple feature models; For the target model, the iterative optimization of the multiple feature models based on the multiple deviations includes: The parameters of the target model are adjusted based on at least one of the deviations corresponding to the target model. Based on the error of the adjusted target model, the parameters are continuously adjusted until the error meets the preset standard.

6. A link optimization device, characterized in that, include: The acquisition module is used to acquire multiple historical data transmitted through the service link and extract the original features of each historical data. Substitute each historical data point into a preset regression model to obtain the regression coefficient for each historical data point; identify each original feature whose corresponding regression coefficient is greater than a preset value as a target feature; wherein, each historical data point refers to the data transmitted by the service link in the past first preset time period; The feature extraction module is used to extract features from multiple historical data to obtain feature data of each historical data at each time point; The model building module is used to build a model based on the feature data of each historical data point at each time point, so as to obtain the feature model corresponding to each historical data. An adjustment module is used to predict the predicted data of each feature model in the future second preset time period, and adjust the multiple target features based on the multiple predicted data to optimize the performance of the service link.

7. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; When the processor executes the computer execution instructions stored in the memory, it is used to implement the link optimization method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the link optimization method as described in any one of claims 1 to 5.

9. A computer program product comprising a computer program, which, when executed by a processor, is used to implement the link optimization method as described in any one of claims 1 to 5.

Images