A 5G video call quality management system based on network analysis

Abstract

The application discloses a 5G video call quality management system based on network analysis, relates to the technical field of communication network quality monitoring and optimization, and is used for solving the problems that network resources cannot be efficiently allocated and dynamically optimized according to actual demands, and the stability of video call services and the quality of user experience are reduced. Bandwidth utilization, time delay fluctuation and packet loss rate are collected in real time in a communication link, a node load balancing index is calculated, a high-load node is identified and a marking time period is set, network jitter frequency, traffic peak value and terminal performance parameters of the node in the corresponding time period are collected, a first and a second optimization coefficient are calculated, a support vector machine model is constructed to predict optimization execution time, performance data is collected again at this time point, an optimization strategy coefficient is generated in combination with a random forest algorithm, and the optimization strategy coefficient is compared with a user experience threshold value to generate a recommended optimization strategy, and adaptive optimization of network configuration is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of communication network quality monitoring and optimization, more particularly, the present application relates to a 5G video call quality management system based on network analysis. BACKGROUND

[0002] With the wide deployment and application of 5G communication technology, high-definition video calls have become an important way for mobile terminal users to communicate daily. While providing high bandwidth and low latency, the complex link structure and high dynamic network state of 5G network also make video call quality susceptible to various factors, such as uneven base station load, frequent link jitter, strong traffic burst, and terminal device performance difference, etc. These factors may cause user experience problems such as call lag, blurred picture, voice delay, etc., and thus affect the overall quality of 5G video call service.

[0003] The prior art has the following disadvantages:

[0004] At present, most management methods only record static parameters, lack dynamic analysis and joint optimization of network volatility, burst performance consumption and user experience perception dimensions, and cannot realize intelligent prediction and real-time adjustment of video call quality in complex network environment, resulting in that network resources cannot be efficiently allocated and dynamically optimized according to actual demand, and the stability and user experience quality of video call service are reduced. Therefore, a 5G video call quality management system based on network analysis is proposed.

[0005] The above information disclosed in the background section is only used to enhance the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0006] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a 5G video call quality management system based on network analysis, which solves the problems proposed in the above background technology by using multi-source link parameter real-time collection, support vector machine-based optimization execution time determination method, and intelligent strategy generation mechanism based on random forest algorithm.

[0007] To achieve the above-mentioned purpose, the present application provides the following technical scheme, a 5G video call quality management system based on network analysis, comprising the following modules:

[0008] The network state acquisition module acquires 5G link performance parameters in real time, and further acquires the network jitter frequency, traffic peak value distribution and terminal device performance parameters of the marked nodes in the corresponding time period after receiving the marked nodes and time periods transmitted by the data preprocessing module and the optimization execution time transmitted by the dynamic optimization module, and transmitting them to the dynamic optimization module;

[0009] The data preprocessing module receives the link performance parameters, analyzes the node load condition, divides the node load and marks the high-load node, sets multiple time periods as the marking time period, and transmits the network state collection module and the dynamic optimization module;

[0010] The dynamic optimization module calculates the first and second optimization coefficients based on the network jitter, traffic peak value and device performance of the marked node, constructs a support vector machine model to determine the optimization execution time, and feeds back to the network state collection module; combined with the network jitter frequency, traffic peak value distribution and terminal device performance parameters in the optimization execution time, the random forest algorithm is used to generate the optimization strategy coefficient and transmit to the user feedback module;

[0011] The user feedback module receives the optimization strategy coefficient, generates a recommended optimization strategy, and provides the user end to select and adjust the network configuration of the marked node.

[0012] In a preferred embodiment, the current sampling time is taken as the reference, and the fixed time period backtracking forward is taken as the collection time window;

[0013] The ratio of the bandwidth used by a certain network node in the collection time window to the available bandwidth of the node is calculated as the bandwidth utilization rate;

[0014] The arrival time delay difference between consecutive transmission data packets is calculated to obtain the time delay fluctuation;

[0015] The proportion of lost data packets in the data transmission process is calculated to obtain the packet loss rate.

[0016] In a preferred embodiment, the bandwidth utilization rate, time delay fluctuation and packet loss rate are standardized to obtain a standardized feature vector;

[0017] Define the node load state variable and construct a sample set;

[0018] The frequency of each load category in the sample set is counted, and the prior probability of the node load state is calculated;

[0019] The one-dimensional Gaussian distribution is used to model and calculate the conditional probability distribution of the bandwidth utilization rate, time delay fluctuation and packet loss rate under different load states;

[0020] Combined with the feature vector, the posterior probability of the node belonging to the high-load and low-load states is calculated;

[0021] If the posterior probability of the node belonging to the high-load state is greater than the preset judgment threshold, the node is determined to be a high-load node and a marked node.

[0022] In a preferred embodiment, when the node is determined to be a high-load node and a marked node, the current time is taken as the reference to backtrace the preset marking time period;

[0023] collecting network jitter frequency and traffic peak value distribution of the marked node in the marked time period;

[0024] collecting CPU utilization, memory occupancy and image encoding compression frame rate as terminal device performance parameters.

[0025] In a preferred embodiment, the network jitter frequency and traffic peak value distribution are weighted and averaged to obtain a first optimization coefficient;

[0026] The CPU utilization, memory occupancy and image encoding compression frame rate are used to construct a linear logistic regression to obtain a second optimization coefficient.

[0027] In a preferred embodiment, a support vector machine model is constructed by combining the first optimization coefficient and the second optimization coefficient to obtain an optimized execution time;

[0028] The first and second optimization coefficients are combined into a coefficient feature vector;

[0029] A radial basis kernel function is used on the coefficient feature vector to calculate the optimal hyperplane that maximizes the classification boundary by minimizing the Lagrange objective function;

[0030] A support vector set and corresponding coefficients are obtained after training;

[0031] The support vector set and corresponding coefficients are used to calculate the classification score at each time point, and the time point with the highest score is selected as the optimized execution time;

[0032] The network jitter frequency, traffic peak value distribution and terminal device performance parameters of the marked node are collected at the optimized execution time.

[0033] In a preferred embodiment, the network jitter frequency, traffic peak value distribution and terminal device performance parameters of the marked node are normalized to construct an optimization feature vector;

[0034] The feature vector is used as input to calculate the optimization strategy coefficient using a random forest model;

[0035] Multiple groups of training samples are extracted from the optimization feature vector using the bootstrap sampling method, and each group builds a decision tree;

[0036] Each decision tree recursively selects the feature with the maximum Gini gain to divide the data, forming a complete decision tree;

[0037] The new node features are input into all decision trees at the same time, and the prediction scores of each tree are collected;

[0038] The optimization strategy coefficient is obtained by averaging all prediction scores.

[0039] In a preferred embodiment, the optimization strategy coefficient of each marked node is compared with the user experience threshold to obtain an optimization recommendation level;

[0040] If the optimization strategy coefficient is greater than and equal to the user experience threshold, the optimization recommendation level is a strong recommendation for optimization.

[0041] If the optimization strategy coefficient is less than the user experience threshold, the optimization recommendation level is a strong recommendation for optimization.

[0042] In a preferred embodiment, according to the optimization recommendation level, the corresponding network configuration parameter adjustment suggestion is extracted to generate a recommended optimization strategy.

[0043] The recommended optimization strategy is sent to the user terminal system to adjust the access configuration and scheduling strategy of the network environment in which the current node is located.

[0044] Technical effects and advantages of the present application:

[0045] The present application collects link performance parameters in real time in the communication link, including bandwidth utilization, delay fluctuation and packet loss rate, and calculates the load balancing index of each network node to identify high-load nodes. Then, a time window is set as a marked time period, and the network jitter frequency, traffic peak distribution and terminal device performance parameters of the high-load nodes in the marked time period are collected. Based on these data, a first optimization coefficient and a second optimization coefficient are calculated, and a support vector machine model is constructed to predict the best optimization execution time. At this time point, relevant performance parameters are further collected, and a random forest algorithm is used to generate an optimization strategy coefficient. Finally, the optimization strategy coefficient is compared with the preset user experience threshold to generate a recommended optimization strategy for the user to select and adjust the network configuration, thereby realizing adaptive optimization of the network configuration. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 The implementation flowchart of the 5G video call quality management system based on network analysis of the present application.

[0047] Figure 2 The step schematic diagram of the 5G video call quality management system based on network analysis of the present application. DETAILED DESCRIPTION

[0048] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0049] Embodiment 1, a network analysis-based 5G video call quality management system, as shown in Figure 1 includes a network state acquisition module, a data preprocessing module, a dynamic optimization module and a user feedback module, and the modules are connected through electrical signals;

[0050] The network state acquisition module collects the link performance parameters in the 5G network environment in real time, and transmits the collected data into the data preprocessing module; after detecting the marked nodes and marked time periods transmitted by the data preprocessing module and the optimization execution time transmitted by the dynamic optimization module, further acquire the network jitter frequency, traffic peak distribution and terminal device performance parameters of the marked nodes in the marked time period and the optimization execution time into the dynamic optimization module;

[0051] The data preprocessing module receives the link performance parameters transmitted by the network state acquisition module, analyzes the load balancing of each network node according to the link performance parameters, divides the network nodes into high-load nodes and low-load nodes, and selects the high-load nodes for marking; set multiple time intervals as marking time periods, and transmit the marked nodes and marking time periods to the network state acquisition module and the dynamic optimization module;

[0052] The dynamic optimization module receives the network jitter frequency, traffic peak distribution and terminal device performance parameters of the marked nodes in the marked time period, calculates the first and second optimization coefficients, constructs a support vector machine model by comprehensively considering the first and second optimization coefficients, determines the optimization execution time in the marked time period, and sends the optimization execution time to the network state acquisition module; comprehensively consider the network jitter frequency, traffic peak distribution and terminal device performance parameters of each marked node in the optimization execution time, generate optimization strategy coefficients by using random forest algorithm, and transmit the optimization strategy coefficients to the user feedback module;

[0053] The user feedback module receives the optimization strategy coefficients transmitted by the dynamic optimization module, performs operation with the preset user experience threshold, generates a recommended optimization strategy, and sends the recommended optimization strategy to the user terminal for the user to select and adjust the network configuration of the corresponding marked node.

[0054] The specific implementation is as follows:

[0055] In the network state acquisition module, the data flow transmitted in real time in the communication link is monitored and counted, and the current sampling time is taken as the benchmark, a fixed time period of is backtracked as a collection time window, the link performance parameters in the 5G network environment are collected in real time in the collection time window, and the link performance parameters include bandwidth utilization, delay fluctuation and packet loss rate.

[0056] Bandwidth utilization is the ratio of the bandwidth used by a network node in a time window to the available bandwidth of the node, and the calculation formula is ;

[0057] wherein, is the bandwidth utilization, represents the actual bandwidth usage of node n at time point t, represents the total bandwidth that can be allocated to node n at the same time point.

[0058] Delay fluctuation represents the difference in arrival delay between consecutive transmitted data packets, and the calculation formula is ;

[0059] wherein, is the delay fluctuation, represents the one-way transmission delay of the i-th data packet of node n at time point t, is the average transmission delay of node n in this time period, and M is the number of statistical samples.

[0060] Packet loss rate represents the proportion of lost data packets in the data transmission process, and the calculation formula is ;

[0061] wherein, is the packet loss rate, represents the number of data packets lost by node n at time point t, represents the total number of data packets sent by the node in the same time period.

[0062] It should be noted that the communication link refers to a physical channel or logical path used for data transmission in a communication system.

[0063] The network state acquisition module transmits the above link performance parameters to the data preprocessing module to provide original support data for subsequent load analysis and network optimization strategy generation.

[0064] In the data preprocessing module, a Bayesian classification model is used to identify the load state of the network node, and the network node is divided into high-load nodes and low-load nodes.

[0065] In order to eliminate the dimension effect and unify the value range, first, the bandwidth utilization, delay fluctuation and packet loss rate are standardized to obtain the standardized feature vector:

[0066] ;

[0067] wherein, , , , are the mean values of bandwidth utilization, delay fluctuation and packet loss rate, respectively, For its corresponding standard deviation.

[0068] Definition of node load state variable When the node is in high load state, When the node is in low load state, ;

[0069] By collecting historical data in the historical database to build a sample set , the frequency of each load category in the sample set is counted, and the prior probability P(y) of the node load state is calculated, as follows:

[0070] ;

[0071] Where, represents the number of high load nodes in the sample, represents the number of low load nodes, .

[0072] For each dimension feature , the conditional probability distribution under different load states is calculated respectively . For easy calculation, a one-dimensional Gaussian distribution is used for modeling, and the expression is ;

[0073] Where, represents the mean of feature under load state , and is its standard deviation.

[0074] Based on Bayes' theorem, combined with the feature vector , the posterior probability of belonging to high load and low load state is calculated, and the calculation formula is as follows: ;

[0075] Where the numerator is the joint probability of the current load state , and the denominator is the sum of the joint probability under all possible states.

[0076] Compare the posterior probabilities of the two states, and the decision rule is as follows:

[0077] If , the node is judged to be a high load node;

[0078] If , the node is judged to be a low load node;

[0079] Where, is the threshold value, which is obtained by the experiment of the professional personnel, and is not described here.

[0080] Let the set of nodes identified as high-load status be denoted as:

[0081] ;

[0082] Label each network node in the above set of nodes;

[0083] To support multi-scale time domain analysis, when a node is determined to be a high-load node and serves as a labeled node, further preset K labeled time periods of different lengths ;

[0084] wherein, each labeled time period represents a time window that is traced back from the current time t.

[0085] It should be noted that the Bayesian classification model is a supervised learning model based on Bayesian theorem for probability inference, which is calculated by combining prior probability and conditional probability; the historical database is used to record the running state data and load annotation information of the network node in the historical multi-time period; the prior probability refers to the known probability of the occurrence of a certain event or category itself without observed sample data; the one-dimensional Gaussian distribution modeling refers to a method of modeling the normal distribution of a certain dimension continuous variable; the posterior probability is the conditional probability of the occurrence of a certain category given the observation sample.

[0086] The data preprocessing module transmits the above screened labeled nodes and labeled time periods to the network state acquisition module and the dynamic optimization module.

[0087] After receiving the labeled nodes and labeled time periods transmitted by the data preprocessing module, the network state acquisition module acquires the network jitter frequency, traffic peak distribution and terminal device performance parameters of the labeled nodes in the labeled time period.

[0088] The network jitter frequency is the delay change of the continuous video call data packets in the labeled time period , and the calculation formula is wherein, is the network jitter frequency, represents the network delay of the i-th data packet, and N represents the total number of data packets collected in the labeled time period.

[0089] The traffic peak distribution represents the maximum uplink or downlink throughput value in the labeled time period , and the calculation expression is wherein, is the traffic peak distribution, represents the transmission byte number at the sampling time point , and represents the window time length.

[0090] terminal device performance parameter, denoted as The CPU utilization rate, memory occupancy rate and image encoding compression frame rate of the terminal device of the node are collected, and the following feature vector is formed:

[0091] ;

[0092] wherein, is the CPU utilization rate, is the memory occupancy rate, is the image encoding compression frame rate;

[0093] After the network state collection module completes the collection of the three types of performance parameters, the data set is input to the dynamic optimization module.

[0094] It should be noted that the uplink throughput value refers to the amount of data successfully transmitted by the terminal device to the base station in a unit of time; the downlink throughput value refers to the amount of data successfully transmitted by the base station to the terminal device in a unit of time.

[0095] The dynamic optimization module calculates the first optimization coefficient and the second optimization coefficient based on the input feature set using weighted average, which are used to represent short-term jitter adaptability and device load margin.

[0096] The first optimization coefficient is calculated from the network jitter frequency and traffic peak distribution, and the calculation formula is as follows:

[0097] ;

[0098] wherein, is the first optimization coefficient, is the network jitter frequency, is the traffic peak distribution, the network jitter frequency and the traffic peak distribution are greater, the first optimization coefficient is greater.

[0099] The second optimization coefficient is calculated from the terminal device performance parameter, and the calculation formula is as follows:

[0100] ;

[0101] wherein, is the second optimization coefficient, is the CPU utilization rate in the terminal device performance parameter, is the memory occupancy rate in the terminal device performance parameter, is the image encoding compression frame rate in the terminal device performance parameter, The maximum encoding frame rate supported by the system, The weight parameter.

[0102] The dynamic optimization module combines the first optimization coefficient and the second optimization coefficient to construct a support vector machine model, adopts a supervised learning method to train and predict the optimization execution time in the marked time period ;

[0103] First, the first optimization coefficient and the second optimization coefficient form a joint feature vector , defined as ;

[0104] Based on the above feature vector set , a support vector machine model is constructed for predicting the optimization execution time in a given marked time period. The support vector machine model adopts a radial basis kernel function as the kernel function form, defined as ;

[0105] Wherein, is the kernel width parameter, represents any two training samples, and the training process is performed by minimizing the Lagrange objective function to solve the optimal hyperplane, and the calculation formula is ;

[0106] The constraint condition is: ;

[0107] Wherein, is the weight vector, is the bias term, is the feature mapping function, is the soft margin variable, is the penalty factor, indicates the sample classification label, and in this case the classification label is optimization adaptation or non-adaptation.

[0108] The support vector set and the corresponding coefficient obtained after training are used to construct the final discriminant function , wherein, is the Lagrange multiplier corresponding to the support vector.

[0109] The dynamic optimization module inputs the corresponding of the discrete time points in the time period one by one based on the above model , obtains the optimization execution time , and the is the time point with the highest score in the positive class of the discrimination result, and the calculation formula is ;

[0110] It should be noted that the support vector machine model is a supervised learning method for classification and regression tasks; the radial basis kernel function is a commonly used support vector machine kernel function for controlling the scale of nonlinear mapping; the Lagrange objective function is used to solve the Lagrange multiplier, and the minimization of the function is the optimization objective of model training.

[0111] The dynamic optimization module obtains the optimized execution time transmitted to the network state acquisition module to guide the subsequent data acquisition enhancement and scheduling behavior of the node.

[0112] The network state acquisition module receives the optimized execution time transmitted by the dynamic optimization module, and then collects the network jitter frequency, traffic peak distribution, and terminal device performance parameters of the marked node according to the optimized execution time.

[0113] After the dynamic optimization module receives the normalized parameters, for each marked node n, a feature vector is constructed: ;

[0114] Then, the random forest algorithm is used to calculate the optimization strategy coefficient. The random forest model is composed of independent classification and regression trees , which are used to fit and predict the scheduling strategy of the node state. Each tree constructs a bootstrap sample set from the training set data in the training stage through replacement sampling, and constructs a tree structure in the feature subspace. The specific division is based on the maximum information gain: ;

[0115] wherein, is the current node sample set, represents the subset after division according to the value of the feature , is the proportion of samples belonging to the th class.

[0116] In the model prediction stage, the dynamic optimization module inputs the current feature vector into all sub-trees to obtain the individual prediction result of each sub-tree, and the final optimization strategy coefficient is obtained by averaging the prediction values of each sub-tree: ;

[0117] wherein, represents the priority of the node in the current optimization execution time period. The higher the optimization strategy coefficient value, the more the node needs to be scheduled and adjusted under the network fluctuation and load scenario, and the higher the priority.

[0118] It should be noted that random forest is an ensemble learning algorithm consisting of multiple decision trees; sampling with replacement is a statistical resampling method that randomly selects samples from the original dataset to form a new sample set, and the samples can be drawn again after each sampling.

[0119] The user feedback module receives the optimization strategy coefficients from the dynamic optimization module. Then, for each marked node Optimization strategy coefficients With user experience threshold Comparative analysis is performed, and a recommendation optimization strategy is generated. User experience threshold. These are preset parameters that satisfy... This represents the minimum acceptable quality of service score for a user, and was derived experimentally by those skilled in the art.

[0120] For each node Optimization suggestion level Conduct an assessment;

[0121] like ,but Optimized for strong recommendations;

[0122] like ,but Optimize for weak recommendations.

[0123] After completing the classification of optimization suggestion levels for all nodes, the user feedback module will classify the optimization suggestion levels according to each category. Extract corresponding network configuration parameter adjustment suggestions from the optimization strategy rule base and generate recommended optimization strategies. ;

[0124] in, The recommended network parameter configuration for node n is determined by the optimization policy function, i.e. Strategy function Based on the combination relationship between the node optimization strategy coefficient and the suggestion level, the matching parameter template in the optimization strategy rule base is called to generate the recommendation strategy content.

[0125] The user feedback module will recommend optimization strategies. After encapsulation, the data is sent to the user terminal system and presented to the user through the interface control interface. The user can choose whether to apply the recommended configuration based on their actual experience. If the user confirms the application, the terminal system will adjust the access configuration and scheduling strategy of the current node's network environment according to the recommended parameters, completing the adaptive adjustment process of network configuration for the marked node.

[0126] It should be noted that the optimization strategy rule base is a structured strategy knowledge set, which defines the network state characteristics and recommended configuration behaviors corresponding to each rule, and will not be described here.

[0127] The above formulas are dimensionless values calculated, and the formulas are obtained by collecting a large amount of data to simulate a formula of the most recent real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0128] The above embodiments can be realized wholly or partially by software, hardware, firmware or any other combination. When realized by software, the above embodiments can be realized wholly or partially in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are wholly or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable devices. The computer instructions can be stored in a computer readable storage medium or transferred from one computer readable storage medium to another, for example, the computer instructions can be transferred from one website, computer, server or data center to another website, computer, server or data center through wired or wireless (such as infrared, wireless, microwave, etc.) mode. The computer readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center and the like containing one or more available medium sets. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid state disk.

[0129] It should be understood that in various embodiments of the present application, the size of the sequence number of each process described above does not mean the order of execution, and the execution order of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0130] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. A person 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 present application.

[0131] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system, device and unit described above can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0132] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0133] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on a plurality of network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0134] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit.

[0135] If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the essential part or part of the technical solutions that make contributions to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program code storage media.

[0136] The above merely provides the specific implementation of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of the changes or replacements within the technical range disclosed by the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A network analysis based 5G video call quality management system, characterized in that: The system comprises the following modules: The network state acquisition module acquires 5G link performance parameters in real time, and further acquires the network jitter frequency, traffic peak value distribution and terminal device performance parameters of the marked node within the corresponding time period after receiving the marked node and marked time period transmitted by the data preprocessing module and the optimization execution time transmitted by the dynamic optimization module, and transmits them to the dynamic optimization module; The data preprocessing module receives the link performance parameters, analyzes the node load condition, divides the nodes according to the load and marks the high-load nodes, sets multiple time periods as the marked time periods, and transmits them to the network state acquisition module and the dynamic optimization module; The dynamic optimization module calculates the first and second optimization coefficients based on the network jitter, traffic peak value and device performance of the marked node, constructs a support vector machine model to determine the optimization execution time, and feeds back to the network state acquisition module; The optimization strategy coefficient is generated by combining the network jitter frequency, traffic peak value distribution and terminal device performance parameters within the optimization execution time, and is transmitted to the user feedback module; The first optimization coefficient is calculated from the network jitter frequency and traffic peak value distribution, and the calculation formula is as follows: ； wherein, is a first optimization coefficient, is a network jitter frequency, is a traffic peak distribution, the network jitter frequency and the traffic peak distribution is greater, the first optimization coefficient is greater; The second optimization coefficient is calculated from the terminal device performance parameters, and the calculation formula is as follows: ； wherein, is a second optimization coefficient, is a CPU utilization rate in the terminal device performance parameter, is a memory occupancy rate in the terminal device performance parameter, is an image encoding compression frame rate in the terminal device performance parameter, is a maximum encoding frame rate supported by the system, is a weight parameter; The dynamic optimization module combines the first optimization coefficient and the second optimization coefficient to construct a support vector machine model, and adopts a supervised learning mode to train and predict the optimization execution time in the marked time period ; First, the first optimization coefficient and the second optimization coefficient are constituted into a joint eigenvector , defined as ; based on the set of joint feature vectors , a support vector machine model is constructed for predicting the optimized execution time within a given labeling time period; the support vector machine model adopts a radial basis kernel function as a kernel function form, defined as ; wherein, is a kernel width parameter, denotes any two training samples, the training process is performed by minimizing a Lagrangian objective function for optimal hyperplane solving, and the calculation formula is ; The constraints are: ; wherein, is a weight vector, is a bias term, is a feature mapping function, is a soft margin variable, is a penalty factor, denotes a sample classification label, the classification label being optimized to be either adapted or not adapted; The final discriminant function is constructed by the support vector set and the corresponding coefficients obtained after training as , wherein is the Lagrange multiplier corresponding to the support vector; x is the training sample at the time point to be predicted. The dynamic optimization module obtains the discrete time points in the time period based on the above model corresponding to each of the time points , and obtains the optimized execution time , the The time point with the highest score in the positive class is the discrimination result, and the calculation formula is ; The user feedback module receives the optimization strategy coefficient, generates a recommended optimization strategy, and provides the user end with the option to adjust the network configuration of the marked node.

2. The 5G video call quality management system based on network analysis according to claim 1, characterized in that: The current sampling time is taken as the benchmark, and a fixed time period backtracked from the current time is taken as the collection time window; The ratio of the bandwidth used by a certain network node in the collection time window to the available bandwidth of the node is calculated as the bandwidth utilization rate; The arrival time delay difference between consecutive data packets is calculated to obtain the time delay fluctuation; The proportion of the lost data packets in the data transmission process is calculated to obtain the packet loss rate.

3. The 5G video call quality management system based on network analysis according to claim 2, characterized in that: The bandwidth utilization rate, time delay fluctuation and packet loss rate are standardized to obtain a standardized feature vector; A node load state variable is defined, and a sample set is constructed; The frequency of each load category in the sample set is counted to calculate the prior probability of the node load state; A one-dimensional Gaussian distribution model is used to calculate the conditional probability distribution of the bandwidth utilization rate, time delay fluctuation and packet loss rate under different load states; The posterior probability of the node belonging to the high-load and low-load states is calculated based on the feature vector; If the posterior probability of the node belonging to the high-load state is greater than a preset judgment threshold, the node is determined to be a high-load node and a marked node.

4. The 5G video call quality management system based on network analysis according to claim 3, characterized in that: When the node is determined to be a high-load node and a marked node, the current time is taken as the benchmark and a preset marked time period is backtracked from the current time; The network jitter frequency and traffic peak value distribution of the marked node within the marked time period are collected; The CPU utilization rate, memory occupancy rate and image encoding compression frame rate are collected as terminal device performance parameters.

5. The 5G video call quality management system based on network analysis according to claim 4, characterized in that: The network jitter frequency, traffic peak distribution and terminal device performance parameters of the marked nodes are normalized to construct an optimization feature vector; The feature vector is taken as input, and a random forest model is used to calculate the optimization strategy coefficient; Multiple groups of training samples are extracted from the optimization feature vector using the bootstrap sampling method, and each group builds a decision tree; Each decision tree recursively selects the feature with the maximum Gini gain to divide the data, forming a complete decision tree; The new node features are input into all decision trees, and the prediction scores of each tree are collected; The optimization strategy coefficient is obtained by averaging all prediction scores.

6. The 5G video call quality management system based on network analysis according to claim 5, characterized in that: The optimization strategy coefficient of each marked node is compared with the user experience threshold to obtain the optimization recommendation level; If the optimization strategy coefficient is greater than and equal to the user experience threshold, the optimization recommendation level is strong recommendation optimization; If the optimization strategy coefficient is less than the user experience threshold, the optimization recommendation level is strong recommendation optimization.

7. The 5G video call quality management system based on network analysis according to claim 6, characterized in that: According to the optimization recommendation level, the corresponding network configuration parameter adjustment suggestion is extracted to generate a recommended optimization strategy; The recommended optimization strategy is sent to the user terminal system to adjust the access configuration and scheduling strategy of the network environment where the current node is located.

Images