A network resource scheduling management system and method based on 5G communication

Abstract

The present invention discloses a network resource scheduling management system and method based on 5G communication, which relates to the technical field of network resource scheduling, and includes the following steps: collecting historical network data, collecting current network traffic, bandwidth occupancy, number of user connections, delay and network status information in real time; preprocessing the collected data and extracting key features; using the extracted key feature data to train a feature neural network model, and predicting the demand for various resources used by the received scheduling tasks in the resource scheduling process; scheduling and allocating resources according to the prediction results; real-time monitoring of network performance indicators to further optimize the accuracy of resource demand prediction; and updating historical network data in real time according to feedback data. The present invention proposes to perform resource scheduling by judging the load according to the real-time situation of the data, thereby improving the utilization rate of network resources and ensuring load balancing.

Description

Technical Field

[0001] The present invention relates to the technical field of network resource scheduling, and specifically to a network resource scheduling management system and method based on 5G communication. Background Art

[0002] Traditional network resource scheduling methods are mainly based on static rules or simple threshold strategies. In terms of data collection, some methods simply obtain some network parameters, such as network traffic or the number of user connections, while ignoring other important information, such as the utilization of network resource scheduling, the load of routers and switches, and the transmission characteristics of TCP and UDP data packets. This leads to an incomplete understanding of the network status and an inability to accurately grasp the actual operation of the network. Therefore, the current utilization of network resource scheduling is relatively low, some nodes are overloaded and cannot bear other scheduling tasks, while some nodes have low utilization and are idle for a period of time, and load balancing cannot be achieved. Summary of the invention

[0003] The purpose of the present invention is to provide a network resource scheduling management system and method based on 5G communication to solve the problems raised in the prior art.

[0004] To achieve the above object, the present invention provides the following technical solution: a network resource scheduling management method based on 5G communication, the network resource scheduling management method comprising the following steps:

[0005] Step S1, collect historical network data, including network traffic, bandwidth usage and number of user connections in different time periods; collect current network traffic, bandwidth usage, number of user connections, delay and network status information in real time;

[0006] Step S2, preprocessing the collected data, using the Spearman correlation coefficient method to perform feature selection and extract key features; the preprocessing includes noise processing and outlier replacement;

[0007] Step S3: Use the extracted key feature data to train the feature neural network model to perform demand forecasting on various resources used by the received scheduling tasks in the resource scheduling process;

[0008] Step S4: According to the demand forecast results of various resources, the scheduling tasks are sorted in ascending order according to the demand forecast results of the resources, and resource scheduling and allocation are performed;

[0009] Step S5: monitor network performance indicators in real time, including latency, bandwidth, and throughput, and feed the data back to step S3 to further optimize the accuracy of resource demand prediction; based on the feedback data, update the historical network data in real time.

[0010] In step S1, historical network data is collected, including network traffic, bandwidth usage, number of user connections, etc. in different time periods; the content of real-time collection of current network traffic, bandwidth usage, number of user connections, delay, and network status information is specifically:

[0011] Record network traffic data at different time points through network monitoring tools and server logs hist (t) and F curr ; Among them, F hist (t) represents the historical network traffic size at time point t; F curr Indicates the current network traffic size;

[0012] Obtain bandwidth usage data at different time points through the management interface of network equipment and network monitoring software hist (t) and B curr Among them, B hist (t) represents the historical bandwidth occupancy ratio at time point t; B curr Indicates the current bandwidth usage ratio;

[0013] Obtain the number of user connections U at different time points through the statistical information of network access devices hist (t) and U curr Among them, U hist U (t) represents the number of users historically connected to the network at time point t; curr Indicates the number of users connected to the network;

[0014] Determine network latency by sending test packets and measuring the round trip time curr , D curr Indicates the time required for data to travel from the sender to the receiver;

[0015] The utilization rate S of network resource scheduling is obtained by recording the proportion of used resources;

[0016] Use network management systems and monitoring tools to obtain the load conditions of routers and switches, the number of TCP connections, and the sending and receiving rates of UDP data packets.

[0017] In step S2, the method for performing noise processing on the collected data is specifically: using a standardized transformation method on the acquired data to make the data with different features have the same scale. The specific formula is: ;in, represents the data after the standardization transformation; X represents the data obtained in step S1; μ represents the average value of the data X; σ represents the standard deviation of the data X.

[0018] In step S2, the method for replacing outliers in the collected data is specifically as follows:

[0019] Step S2-1-1, replace outliers on the noise-processed data, and sort the data of the same type in descending order; calculate the upper quartile, median, and lower quartile of the data; the upper quartile, median, and lower quartile of the data are the data corresponding to the upper quartile point Q1, the median point Q2, and the lower quartile point Q3, respectively; ; ; ; Where m represents the number of data, and m is a positive integer; Q1 and Q2 are rounded up; Q3 is rounded down;

[0020] Step S2-1-2, calculating the difference IQR between the upper quartile and the lower quartile; IQR is the absolute value of the difference between the upper quartile and the lower quartile;

[0021] Step S2-1-3, the upper quartile is used as the upper limit, the lower quartile is used as the lower limit, and the data greater than the upper quartile + 1.5 times the IQR and the data less than the lower quartile - 1.5 times the IQR are classified as outliers;

[0022] Step S2-1-4, replace the outliers greater than the upper quartile + 1.5 times the IQR with the value of the upper quartile + 1.5 times the IQR, and replace the outliers less than the lower quartile - 1.5 times the IQR with the value of the lower quartile - 1.5 times the IQR.

[0023] In step S2, feature selection is performed, and the specific steps of extracting key features are:

[0024] Step S2-2-1, arranging the network data and the utilization rate S of the network resource scheduling in descending order, and recording them as sequence R1 and sequence R2;

[0025] Step S2-2-2, calculate the rank of each data point in its respective sequence, that is, find the position of the data point in the sequence, and if there are repeated data points, use the average rank method;

[0026] Step S2-2-3, calculate the rank difference d, that is, the rank of the corresponding data point in R1 minus the rank of the corresponding data point in R2;

[0027] Step S2-2-4, calculate the value of the Spearman correlation coefficient ρ, the specific calculation formula is:

[0028] ; Among them, ρ represents the Spearman correlation coefficient; n represents the number of network data, n is a positive integer; d l Represents the rank difference of the lth network data value;

[0029] Step S2-2-5: The network data corresponding to the ρ value greater than the average Spearman correlation coefficient is determined as the key feature, which is recorded as W1, W2, ..., W K ; Among them, W1, W2, ..., W K represents the 1st, 2nd, ..., Kth key features, where K is a positive integer.

[0030] In step S3, the specific steps of using the extracted key feature data to train the feature neural network model and perform demand forecasting on various resources used by the received scheduling task in the resource scheduling process are:

[0031] Step S3-1, design a feature neural network model, determine the input layer, hidden layer and output layer, input layer: K key features W1, W2, ..., W determined by feature selection K As input layer neurons; Hidden layer: The number of hidden layer neurons is set to H=2K; Nonlinear transformation is performed on the information of the input layer; Output layer: The number of output layer neurons is determined according to the prediction target;

[0032] Step S3-2: perform normalization processing to make the numerical range of the data suitable for neural network training; the normalization formula is: ; Among them, W i ' represents the normalized data; W i represents the i-th key feature data, i represents the neuron number, i is a positive integer, 1≤i≤K; max(W i ) and min(W i ) represent the maximum and minimum values ​​of the i-th key feature data respectively; the normalized key feature data is input into the feature neural network model through the input layer as an input vector to provide a data basis for subsequent training and prediction processes;

[0033] Step S3-3: For the connection weight w from the input layer to the hidden layer ji 1 and the bias b of the hidden layer neurons j 1 Randomly initialize and change the connection weight w from the hidden layer to the output layer j 2 and the bias b of the output layer neurons 2 Random initialization is also performed, and the random initialization range is [0,1]; where i represents the neuron number, i is a positive integer, 1≤i≤K; j represents the hidden layer neuron number, j is a positive integer, 1≤j≤H; hidden layer neuron h j The calculation is as follows: ; Wherein, f represents the activation function, and the present invention adopts the ReLU function f(x)=max(0,x); After the hidden layer is calculated, the calculation of the output layer neuron output y is specifically: ; Among them, the output layer neuron output y represents the resource demand prediction result.

[0034] In step S4, according to the demand prediction results of various resources, the specific content of scheduling allocation is as follows: set a request priority scheduling list, and schedule the scheduling tasks in the priority scheduling list according to FCFS; the remaining scheduling tasks are allocated as follows: sort the scheduling tasks in ascending order according to the demand prediction results of the resources, start scheduling the scheduling tasks with the smallest demand prediction results, delete the scheduled scheduling tasks in the ascending order, and after each execution, update the data such as network traffic, bandwidth occupancy and number of user connections in real time, extract the data with the lowest load ratio in the data such as network traffic, bandwidth occupancy and number of user connections, find the scheduling tasks corresponding to the data with the lowest load ratio in the ascending order, and perform resource scheduling, and execute them in sequence.

[0035] In step S5, the network performance indicators are monitored in real time, including latency, bandwidth, and throughput, and the data is fed back to step S3 to further optimize the accuracy of resource demand prediction. The specific steps of updating the historical network data in real time according to the feedback data are as follows:

[0036] Step S5-1, real-time monitoring of network performance indicators, feeding back data to step S3, further optimizing the accuracy of resource demand forecasting; the specific method of optimization is: calculating the mean square error, by calculating the error between the predicted output and the target output, using the back propagation algorithm, and adjusting the weight and bias of the neural network according to the error; the mean square error calculation formula is: , where E represents the mean square error; N represents the number of training samples; y represents the resource demand prediction result; y' represents the real-time measurement result; the back propagation algorithm specifically starts from the output layer, reversely calculates the gradient of the error for each weight and bias, and then updates the weight and bias according to the gradient descent method, so that the error gradually decreases. This process is trained for many times until the error converges to a smaller value, indicating that the neural network has learned the mapping relationship between the key input features and the target output.

[0037] Step S5-2: Update the historical network data in real time according to the feedback data.

[0038] A network resource scheduling and management system based on 5G communication, the network resource scheduling and management system comprising a data collection module, a data processing module, a feature selection and model training module, a resource scheduling and allocation module, and a feedback and data update module;

[0039] The data collection module is used to extract traffic data, bandwidth usage, number of connected users, device load, round-trip delay and other data according to time points through various network monitoring tools and server logs; obtain historical data and real-time data of the system to provide basic data support for subsequent data processing and scheduling decisions;

[0040] The data processing module is used to pre-process the collected historical and real-time data, remove noise and outliers, and extract key features using feature selection methods;

[0041] The feature selection and model training module is used to design and train a feature neural network model through key feature data to predict the resource requirements of the scheduling task;

[0042] The resource scheduling and allocation module is used to sort and allocate resources for each scheduling task according to the predicted resource demand;

[0043] The feedback and data update module is used to monitor network performance indicators in real time, optimize the prediction accuracy of the neural network model through a feedback mechanism, and update historical network data in real time based on feedback data.

[0044] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention uses a correlation coefficient analysis method to extract key features, and uses the key features to train a feature neural network model, wherein multiple key features are taken into account and interference feature data are excluded, so the feature neural network model will not be overfitted due to too many features, and does not require self-learning, which reduces the training time to a certain extent and improves the prediction accuracy. Secondly, the present invention proposes a method of determining the load according to the real-time situation of the data for resource scheduling, thereby improving the utilization rate of network resources and ensuring load balancing. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a flowchart of a network resource scheduling management system and method based on 5G communication of the present invention;

[0046] Figure 2 This is a structural schematic diagram of a network resource scheduling management system and method based on 5G communication in the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] Example: Figure 1-Figure 2 As shown, the present invention provides a technical solution, a network resource scheduling management method based on 5G communication, and the network resource scheduling management method comprises the following steps:

[0049] Step S1, collect historical network data, including network traffic, bandwidth usage and number of user connections in different time periods; collect current network traffic, bandwidth usage, number of user connections, delay and network status information in real time;

[0050] Step S2, preprocessing the collected data, using the Spearman correlation coefficient method to perform feature selection and extract key features; the preprocessing includes noise processing and outlier replacement;

[0051] Step S3: Use the extracted key feature data to train the feature neural network model to perform demand forecasting on various resources used by the received scheduling tasks in the resource scheduling process;

[0052] Step S4: According to the demand forecast results of various resources, the scheduling tasks are sorted in ascending order according to the demand forecast results of the resources, and resource scheduling and allocation are performed;

[0053] Step S5: monitor network performance indicators in real time, including latency, bandwidth, and throughput, and feed the data back to step S3 to further optimize the accuracy of resource demand prediction; based on the feedback data, update the historical network data in real time.

[0054] In step S1, historical network data is collected, including network traffic, bandwidth usage, number of user connections, etc. in different time periods; the content of real-time collection of current network traffic, bandwidth usage, number of user connections, delay, and network status information is specifically:

[0055] Record network traffic data at different time points through network monitoring tools and server logs hist (t) and F curr Among them, F hist (t) represents the historical network traffic size at time point t; F curr Indicates the current network traffic size;

[0056] Obtain bandwidth usage data at different time points through the management interface of network equipment and network monitoring software hist (t) and B curr Among them, B hist (t) represents the historical bandwidth occupancy ratio at time point t; B curr Indicates the current bandwidth usage ratio;

[0057] Obtain the number of user connections U at different time points through the statistical information of network access devices hist(t) and U curr Among them, U hist U (t) represents the number of users historically connected to the network at time point t; curr Indicates the number of users connected to the network;

[0058] Determine network latency by sending test packets and measuring the round trip time curr , D curr Indicates the time required for data to travel from the sender to the receiver;

[0059] The utilization rate S of network resource scheduling is obtained by recording the proportion of used resources;

[0060] Use network management systems and monitoring tools to obtain the load conditions of routers and switches, the number of TCP connections, and the sending and receiving rates of UDP data packets.

[0061] Historical network traffic data (F hist (t)): Time point 1: 100Mbps; Time point 2: 150Mbps; Time point 3: 200Mbps;

[0062] Current network traffic data (F curr ): Current traffic: 180Mbps;

[0063] Historical bandwidth usage data (B hist (t)):

[0064] Time point 1: 60%; Time point 2: 75%; Time point 3: 90%;

[0065] Current bandwidth usage data (B curr ): Current bandwidth usage: 80%;

[0066] Historical user connections (U hist (t)):

[0067] Time point 1:50; Time point 2:75; Time point 3:100;

[0068] Current number of user connections (Ucurr): Current number of connections: 90;

[0069] Network delay (D curr ): Current delay: 20ms;

[0070] In step S2, the method for performing noise processing on the collected data is specifically: using a standardized transformation method on the acquired data to make the data with different features have the same scale. The specific formula is: ;in, represents the data after the standardization transformation; X represents the data obtained in step S1; μ represents the average value of the data X; σ represents the standard deviation of the data X.

[0071] F curr :mean μ=150, standard deviation σ=50;

[0072] B curr :mean μ=75, standard deviation σ=15;

[0073] U curr :mean μ=75, standard deviation σ=25;

[0074] Calculate the normalized value:

[0075] F curr '=(180-150) / 50=0.6;

[0076] B curr '=(80-75) / 15=0.33;

[0077] U curr '=(90-75) / 25=0.6;

[0078] In step S2, the method for replacing outliers in the collected data is specifically as follows:

[0079] Step S2-1-1, replace outliers on the noise-processed data, and sort the data of the same type in descending order; calculate the upper quartile, median, and lower quartile of the data; the upper quartile, median, and lower quartile of the data are the data corresponding to the upper quartile point Q1, the median point Q2, and the lower quartile point Q3, respectively; ; ; ; Where m represents the number of data, and m is a positive integer; Q1 and Q2 are rounded up; Q3 is rounded down;

[0080] Step S2-1-2, calculating the difference IQR between the upper quartile and the lower quartile; IQR is the absolute value of the difference between the upper quartile and the lower quartile;

[0081] Step S2-1-3, the upper quartile is used as the upper limit, the lower quartile is used as the lower limit, and the data greater than the upper quartile + 1.5 times the IQR and the data less than the lower quartile - 1.5 times the IQR are classified as outliers;

[0082] Step S2-1-4, replace the outliers greater than the upper quartile + 1.5 times the IQR with the value of the upper quartile + 1.5 times the IQR, and replace the outliers less than the lower quartile - 1.5 times the IQR with the value of the lower quartile - 1.5 times the IQR.

[0083] F hist (t) Perform outlier replacement: Since no value is out of range, the data remains unchanged.

[0084] In step S2, feature selection is performed, and the specific steps of extracting key features are:

[0085] Step S2-2-1, arranging the network data and the utilization rate S of the network resource scheduling in descending order, and recording them as sequence R1 and sequence R2;

[0086] Step S2-2-2, calculate the rank of each data point in its respective sequence, that is, find the position of the data point in the sequence, and if there are repeated data points, use the average rank method;

[0087] Step S2-2-3, calculate the rank difference d, that is, the rank of the corresponding data point in R1 minus the rank of the corresponding data point in R2;

[0088] Step S2-2-4, calculate the value of the Spearman correlation coefficient ρ, the specific calculation formula is:

[0089] ; Among them, ρ represents the Spearman correlation coefficient; n represents the number of network data, n is a positive integer; d l Represents the rank difference of the lth network data value;

[0090] Step S2-2-5: The network data corresponding to the ρ value greater than the average Spearman correlation coefficient is determined as the key feature, which is recorded as W1, W2, ..., W K ; Among them, W1, W2, ..., W K represents the 1st, 2nd, ..., Kth key features, where K is a positive integer.

[0091] In step S3, the specific steps of using the extracted key feature data to train the feature neural network model and perform demand forecasting on various resources used by the received scheduling task in the resource scheduling process are:

[0092] Step S3-1, design a feature neural network model, determine the input layer, hidden layer and output layer, input layer: K key features W1, W2, ..., W determined by feature selection K As input layer neurons; Hidden layer: The number of hidden layer neurons is set to H=2K; Nonlinear transformation is performed on the information of the input layer; Output layer: The number of output layer neurons is determined according to the prediction target;

[0093] Step S3-2: perform normalization processing to make the numerical range of the data suitable for neural network training; the normalization formula is: ; Among them, W i' represents the normalized data; W i represents the i-th key feature data, i represents the neuron number, i is a positive integer, 1≤i≤K; max(W i ) and min(W i ) represent the maximum and minimum values ​​of the i-th key feature data respectively; the normalized key feature data is input into the feature neural network model through the input layer as an input vector to provide a data basis for subsequent training and prediction processes;

[0094] Step S3-3: For the connection weight w from the input layer to the hidden layer ji 1 and the bias b of the hidden layer neurons j 1 Randomly initialize and change the connection weight w from the hidden layer to the output layer j 2 and the bias b of the output layer neurons 2 Random initialization is also performed, and the random initialization range is [0,1]; where i represents the neuron number, i is a positive integer, 1≤i≤K; j represents the hidden layer neuron number, j is a positive integer, 1≤j≤H; hidden layer neuron h j The calculation is as follows: ; Wherein, f represents the activation function, and the present invention adopts the ReLU function f(x)=max(0,x); After the hidden layer is calculated, the calculation of the output layer neuron output y is specifically: ; Among them, the output layer neuron output y represents the resource demand prediction result.

[0095] In step S4, according to the demand prediction results of various resources, the specific content of scheduling allocation is as follows: set a request priority scheduling list, and schedule the scheduling tasks in the priority scheduling list according to FCFS; the remaining scheduling tasks are allocated as follows: sort the scheduling tasks in ascending order according to the demand prediction results of the resources, start scheduling the scheduling tasks with the smallest demand prediction results, delete the scheduled scheduling tasks in the ascending order, and after each execution, update the data such as network traffic, bandwidth occupancy and number of user connections in real time, extract the data with the lowest load ratio in the data such as network traffic, bandwidth occupancy and number of user connections, find the scheduling tasks corresponding to the data with the lowest load ratio in the ascending order, and perform resource scheduling, and execute them in sequence.

[0096] In step S5, the network performance indicators are monitored in real time, including latency, bandwidth, and throughput, and the data is fed back to step S3 to further optimize the accuracy of resource demand prediction. The specific steps of updating the historical network data in real time according to the feedback data are as follows:

[0097] Step S5-1, real-time monitoring of network performance indicators, feeding back data to step S3, further optimizing the accuracy of resource demand forecasting; the specific method of optimization is: calculating the mean square error, by calculating the error between the predicted output and the target output, using the back propagation algorithm, and adjusting the weight and bias of the neural network according to the error; the mean square error calculation formula is: , where E represents the mean square error; N represents the number of training samples; y represents the resource demand prediction result; y' represents the real-time measurement result; the back propagation algorithm specifically starts from the output layer, reversely calculates the gradient of the error for each weight and bias, and then updates the weight and bias according to the gradient descent method, so that the error gradually decreases. This process is trained for many times until the error converges to a smaller value, indicating that the neural network has learned the mapping relationship between the key input features and the target output.

[0098] Step S5-2: Update the historical network data in real time according to the feedback data.

[0099] like Figure 2 As shown, a network resource scheduling and management system based on 5G communication, the network resource scheduling and management system includes a data collection module, a data processing module, a feature selection and model training module, a resource scheduling and allocation module and a feedback and data update module;

[0100] The data collection module is used to extract traffic data, bandwidth usage, number of connected users, device load, round-trip delay and other data according to time points through various network monitoring tools and server logs; obtain historical data and real-time data of the system to provide basic data support for subsequent data processing and scheduling decisions;

[0101] The data processing module is used to pre-process the collected historical and real-time data, remove noise and outliers, and extract key features using feature selection methods;

[0102] The feature selection and model training module is used to design and train a feature neural network model through key feature data to predict the resource requirements of the scheduling task;

[0103] The resource scheduling and allocation module is used to sort and allocate resources for each scheduling task according to the predicted resource demand;

[0104] The feedback and data update module is used to monitor network performance indicators in real time, optimize the prediction accuracy of the neural network model through a feedback mechanism, and update historical network data in real time based on feedback data.

[0105] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A network resource scheduling management method based on 5G communication, characterized in that: The network resource scheduling management method comprises the following steps: Step S1, collect historical network data, including network traffic, bandwidth usage and number of user connections in different time periods; collect current network traffic, bandwidth usage, number of user connections, delay and network status information in real time; Step S2: pre-process the collected data, use the Spearman correlation coefficient method to perform feature selection, and extract key features; The preprocessing includes noise processing and outlier replacement; Step S3: Use the extracted key feature data to train the feature neural network model to perform demand forecasting on various resources used by the received scheduling tasks in the resource scheduling process; Step S4: According to the demand forecast results of various resources, the scheduling tasks are sorted in ascending order according to the demand forecast results of the resources, and resource scheduling and allocation are performed; Step S5: monitor network performance indicators in real time, including latency, bandwidth, and throughput, and feed the data back to step S3 to further optimize the accuracy of resource demand prediction; based on the feedback data, update the historical network data in real time.

2. A network resource scheduling management method based on 5G communication according to claim 1, characterized in that: In step S1, historical network data is collected, including network traffic, bandwidth usage, and number of user connections in different time periods; the content of real-time collection of current network traffic, bandwidth usage, number of user connections, delay, and network status information is specifically: Record network traffic data at different time points through network monitoring tools and server logs hist (t) and F curr Among them, F hist (t) represents the historical network traffic size at time point t; F curr Indicates the current network traffic size; Obtain bandwidth usage data at different time points through the management interface of network equipment and network monitoring software hist (t) and B curr Among them, B hist (t) represents the historical bandwidth occupancy ratio at time point t; B curr Indicates the current bandwidth usage ratio; Obtain the number of user connections U at different time points through the statistical information of network access devices hist (t) and U curr Among them, U hist U (t) represents the number of users historically connected to the network at time point t; curr Indicates the number of users connected to the network; Determine network latency by sending test packets and measuring the round trip time curr , D curr Indicates the time required for data to travel from the sender to the receiver; The utilization rate S of network resource scheduling is obtained by recording the proportion of used resources; Use network management systems and monitoring tools to obtain the load conditions of routers and switches, the number of TCP connections, and the sending and receiving rates of UDP data packets.

3. A network resource scheduling management method based on 5G communication according to claim 2, characterized in that: In step S2, the method for performing noise processing on the collected data is specifically: using a standardized transformation method on the acquired data to make the data with different features have the same scale. The specific formula is: ;in, represents the data after the standardization transformation; X represents the data obtained in step S1; μ represents the average value of the data X; σ represents the standard deviation of the data X.

4. A network resource scheduling and management method based on 5G communication according to claim 3, characterized in that: In step S2, the method for replacing outliers in the collected data is specifically as follows: Step S2-1-1, replace outliers on the noise-processed data, and sort the data of the same type in descending order; calculate the upper quartile, median, and lower quartile of the data; the upper quartile, median, and lower quartile of the data are the data corresponding to the upper quartile point Q1, the median point Q2, and the lower quartile point Q3, respectively; ; ; ; Where m represents the number of data, and m is a positive integer; Q1 and Q2 are rounded up; Q3 is rounded down; Step S2-1-2, calculating the difference IQR between the upper quartile and the lower quartile; IQR is the absolute value of the difference between the upper quartile and the lower quartile; Step S2-1-3, the upper quartile is used as the upper limit, the lower quartile is used as the lower limit, and the data greater than the upper quartile + 1.5 times the IQR and the data less than the lower quartile - 1.5 times the IQR are classified as outliers; Step S2-1-4, replace the outliers greater than the upper quartile + 1.5 times the IQR with the value of the upper quartile + 1.5 times the IQR, and replace the outliers less than the lower quartile - 1.5 times the IQR with the value of the lower quartile - 1.5 times the IQR.

5. A network resource scheduling and management method based on 5G communication according to claim 4, characterized in that: In step S2, feature selection is performed, and the specific steps of extracting key features are: Step S2-2-1, arranging the network data and the utilization rate S of the network resource scheduling in descending order, and recording them as sequence R1 and sequence R2; Step S2-2-2, calculate the rank of each data point in its respective sequence, that is, find the position of the data point in the sequence, and if there are repeated data points, use the average rank method; Step S2-2-3, calculate the rank difference d, that is, the rank of the corresponding data point in R1 minus the rank of the corresponding data point in R2; Step S2-2-4, calculate the value of the Spearman correlation coefficient ρ, the specific calculation formula is: ; Among them, ρ represents the Spearman correlation coefficient; n represents the number of network data, n is a positive integer; d l Represents the rank difference of the lth network data value; Step S2-2-5: The network data corresponding to the ρ value greater than the average Spearman correlation coefficient is determined as the key feature, which is recorded as W1, W2, ..., W K ; Among them, W1, W2, ..., W K represents the 1st, 2nd, ..., Kth key features, where K is a positive integer.

6. A network resource scheduling management method based on 5G communication according to claim 5, characterized in that: In step S3, the specific steps of using the extracted key feature data to train the feature neural network model and perform demand forecasting on various resources used by the received scheduling task in the resource scheduling process are: Step S3-1, design a feature neural network model, determine the input layer, hidden layer and output layer, input layer: K key features W1, W2, ..., W determined by feature selection K As input layer neurons; Hidden layer: The number of neurons in the hidden layer is set to H=2K; nonlinear transformation is performed on the information of the input layer; Output layer: The number of neurons in the output layer is determined according to the prediction target; Step S3-2: perform normalization processing to make the numerical range of the data suitable for neural network training; the normalization formula is: ; Among them, W i ' represents the normalized data; W i represents the i-th key feature data, i represents the neuron number, i is a positive integer, 1≤i≤K; max(W i ) and min(W i ) represent the maximum and minimum values ​​of the i-th key feature data respectively; the normalized key feature data is input into the feature neural network model through the input layer as an input vector to provide a data basis for subsequent training and prediction processes; Step S3-3: For the connection weight w from the input layer to the hidden layer ji 1 and the bias b of the hidden layer neurons j 1 Randomly initialize and change the connection weight w from the hidden layer to the output layer j 2 and the bias b of the output layer neurons 2 Random initialization is also performed, and the random initialization range is [0,1]; where i represents the neuron number, i is a positive integer, 1≤i≤K; j represents the hidden layer neuron number, j is a positive integer, 1≤j≤H; hidden layer neuron h j The calculation is as follows: ; Wherein, f represents the activation function, and the present invention adopts the ReLU function f(x)=max(0,x); After the hidden layer is calculated, the calculation of the output layer neuron output y is specifically: ; Among them, the output layer neuron output y represents the resource demand prediction result.

7. A network resource scheduling management method based on 5G communication according to claim 6, characterized in that: In step S4, according to the demand prediction results of various resources, the specific content of scheduling allocation is as follows: set a request priority scheduling list, and schedule the scheduling tasks in the priority scheduling list according to FCFS; the remaining scheduling tasks are allocated as follows: sort the scheduling tasks in ascending order according to the demand prediction results of the resources, start scheduling the scheduling tasks with the smallest demand prediction results, delete the scheduled scheduling tasks in the ascending order, and after each execution, update the network traffic, bandwidth occupancy and number of user connections in real time, extract the data with the lowest load ratio in network traffic, bandwidth occupancy and number of user connections, find the scheduling tasks corresponding to the data with the lowest load ratio in the ascending order, and perform resource scheduling, and execute them in sequence.

8. A network resource scheduling management method based on 5G communication according to claim 7, characterized in that: In step S5, the network performance indicators are monitored in real time, including latency, bandwidth and throughput, and the data is fed back to step S3 to further optimize the accuracy of resource demand prediction; the specific steps of updating the historical network data in real time according to the feedback data are as follows: Step S5-1, real-time monitoring of network performance indicators, feeding back data to step S3, further optimizing the accuracy of resource demand prediction; the specific method of optimization is: calculating the mean square error, by calculating the error between the predicted output and the target output, using the back propagation algorithm, and adjusting the weight and bias of the neural network according to the error; Step S5-2: Update the historical network data in real time according to the feedback data.

9. A network resource scheduling and management system based on 5G communication, characterized in that: The network resource scheduling and management system includes a data collection module, a data processing module, a feature selection and model training module, a resource scheduling and allocation module, and a feedback and data update module; The data collection module is used to extract traffic data, bandwidth usage, number of connected users, device load and round-trip delay data according to time points through various network monitoring tools and server logs; obtain historical data and real-time data of the system to provide basic data support for subsequent data processing and scheduling decisions; The data processing module is used to pre-process the collected historical and real-time data, remove noise and outliers, and extract key features using feature selection methods; The feature selection and model training module is used to design and train a feature neural network model through key feature data to predict the resource requirements of the scheduling task; The resource scheduling and allocation module is used to sort and allocate resources for each scheduling task according to the predicted resource demand; The feedback and data update module is used to monitor network performance indicators in real time, optimize the prediction accuracy of the neural network model through a feedback mechanism, and update historical network data in real time based on feedback data.

Images