Satellite-based network traffic prediction method and electronic equipment

Abstract

The present application provides a satellite-based network traffic prediction method and electronic device, the method comprising obtaining network data corresponding to each network node in a historical time period; inputting the network data corresponding to all network nodes into a trained prediction model, and constructing a function similarity matrix and a data similarity matrix corresponding to each network node through the prediction model; determining the spatial features corresponding to the network nodes based on the function similarity matrix and the data similarity matrix; determining the time features corresponding to the network nodes based on the network data; fusing the spatial features and time features of all network nodes to obtain a fused feature value, calculating the network traffic for the time period to be predicted based on the fused feature value, and outputting the network traffic via the prediction model, thereby solving the technical problem of inaccurate satellite traffic prediction in the prior art and improving the accuracy of satellite traffic prediction.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a satellite-based network traffic prediction method and electronic equipment. Background Art

[0002] With the development of communication technology, the integrated space-air network has emerged. The integrated space-air network is a communication network that makes comprehensive use of satellites, aircraft and ground infrastructure. The satellite network is one of the important components of the integrated space-air network, which realizes global data transmission and communication through the satellite system. The satellite network provides a seamless connection solution for global communication. The accurate prediction of the network traffic of the satellite network can help optimize network performance, resource allocation and security management. Therefore, it is very necessary to predict the network traffic of the satellite.

[0003] In the prior art, traditional methods for network traffic prediction mainly rely on the analysis of satellite network data, such as ARIMA and Holt-Winters. Since one of the characteristics of satellite network traffic is its strong spatiotemporal correlation, traffic prediction needs to consider the impact of different times and locations, which increases the complexity of prediction. Traditional methods are generally difficult to effectively capture the complex spatiotemporal dependencies in network data. Most existing network traffic prediction methods mainly consider spatial similarity, that is, they assume that network nodes with closer distances have higher similarity and are more likely to establish connections in space. However, this spatial relationship may ignore the hidden relationship between network nodes, resulting in reduced prediction accuracy. Summary of the invention

[0004] In view of this, the purpose of this application is to propose a satellite-based network traffic prediction method and electronic device to overcome all or part of the deficiencies in the prior art.

[0005] Based on the above purpose, the present application provides a satellite-based network traffic prediction method, including: obtaining network data corresponding to each network node in a historical time period; inputting the network data corresponding to all network nodes into a trained prediction model, and constructing a function similarity matrix and a data similarity matrix corresponding to each network node through the prediction model; determining the spatial characteristics corresponding to the network node based on the function similarity matrix and the data similarity matrix; determining the time characteristics corresponding to the network node based on the network data; fusing the spatial characteristics and time characteristics of all network nodes to obtain a fused feature value, calculating the network traffic for the time period to be predicted based on the fused feature value, and outputting the network traffic via the prediction model.

[0006] Optionally, constructing the functional similarity matrix corresponding to each network node includes: performing binary conversion on the network data corresponding to each network node to obtain a character string; based on all the character strings, respectively calculating the Hamming distance between the network node and each other network node among all the network nodes except the network node; based on the Hamming distance and a predetermined distance, determining the functional similarity between the network node and each other network node; and constructing the functional similarity matrix based on all the functional similarities.

[0007] Optionally, determining the functional similarity between the network node and each other network node based on the Hamming distance and the predetermined distance includes: in response to determining that the Hamming distance is greater than or equal to the predetermined distance, determining the functional similarity between the network node and the other network nodes corresponding to the Hamming distance as 0; or, in response to determining that the Hamming distance is less than the predetermined distance, determining the functional similarity between the network node and the other network nodes corresponding to the Hamming distance as 1.

[0008] Optionally, constructing the data similarity matrix corresponding to each network node includes: based on all network data, determining the data similarity between the network node and each other network node among all network nodes except the network node; and constructing the data similarity matrix based on all data similarities.

[0009] Optionally, determining the data similarity between the network node and each other network node among all network nodes except the network node based on all network data includes: determining the data similarity coefficient between the network node and each other network node among all network nodes except the network node based on all network data; in response to determining that the data similarity coefficient is greater than or equal to a predetermined coefficient, determining the data similarity between the network node and the other network nodes corresponding to the data similarity coefficient as 1; or, in response to determining that the data similarity coefficient is less than a predetermined coefficient, determining the data similarity between the network node and the other network nodes corresponding to the data similarity coefficient as 0.

[0010] Optionally, determining the spatial features corresponding to the network nodes based on the functional similarity matrix and the data similarity matrix includes: constructing a first diagonal matrix corresponding to the functional similarity matrix based on the functional similarity matrix; constructing a second diagonal matrix corresponding to the data similarity matrix based on the data similarity matrix; determining the product of the functional similarity matrix, the first diagonal matrix, the data similarity matrix and the second diagonal matrix to obtain a fusion matrix corresponding to the network nodes; and performing feature extraction on the fusion matrix to obtain the spatial features.

[0011] Optionally, determining the time characteristics corresponding to the network node based on the network data includes: determining the historical time characteristics and future time characteristics corresponding to the network node based on the network data, and combining the historical time characteristics and the future time characteristics to obtain the time characteristics.

[0012] Optionally, determining the historical time feature and the future time feature corresponding to the network node based on the network data includes: determining the historical time feature by the following formula: in, is the historical time feature, is the information in the t-1 historical time features retained at the current time step t, ⊙ represents the multiplication of data one by one, is the t-1th historical time feature, is the first update matrix of the gated recurrent unit, x t is the input network data, is the second update matrix of the gated recurrent unit, is the information in the t-1 historical time features reset at the current time step t, f represents the update of the historical direction, t is the current time step, and tanh is the hyperbolic tangent function; the future time features are determined by the following formula: in, is the future time feature, is the information of t+1 future time features retained at the current time step t, ⊙ represents the multiplication of data one by one, is the t+1th future time feature, is the third update matrix of the gated recurrent unit, x t is the input network data, is the fourth update matrix of the gated recurrent unit, is the information in the t+1 future time features reset at the current time step t, b represents the update in the future direction, t is the current time step, and tanh is the hyperbolic tangent function.

[0013] Optionally, the spatial features and temporal features of all network nodes are fused to obtain a fused feature value, including: calculating a first product value of the spatial feature of each network node and the attention weight, and calculating a first sum value of all first product values; calculating a second product value of the temporal feature of each network node and the attention weight, and calculating a second sum value of all second product values; and determining the sum of the first sum value and the second sum value as the fused feature value.

[0014] Based on the same inventive concept, the present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable by the processor, wherein the processor implements the method as described above when executing the computer program.

[0015] As can be seen from the above, the satellite-based network traffic prediction method and electronic device provided by the present application include obtaining network data corresponding to each network node in a historical time period. The network data corresponding to all network nodes are input into a trained prediction model, and the function similarity matrix and data similarity matrix corresponding to each network node are constructed through the prediction model, so that the relationship between the network data at the spatial level between the network nodes can be more comprehensively reflected. Based on the function similarity matrix and the data similarity matrix, the spatial characteristics corresponding to the network node are determined. The integration of diversified network data about space enables the prediction model to capture the complex spatial dependencies inherent in the network data, making the spatial characteristics of the network nodes more accurate. Based on the network data, the time characteristics corresponding to the network node are determined, which can accurately reflect the changing trend of the network data over time. The spatial characteristics and time characteristics of all network nodes are fused to obtain a fused feature value, and the network traffic of the time period to be predicted is calculated based on the fused feature value, and the network traffic is output through the prediction model, so as to achieve the purpose of accurately predicting the network traffic. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present application or related technologies, the drawings required for use in the embodiments or related technical descriptions are briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 A schematic diagram of a flow chart of a satellite-based network traffic prediction method according to an embodiment of the present application;

[0018] Figure 2 This is a schematic diagram of the structure of a satellite-based network traffic prediction device according to an embodiment of the present application;

[0019] Figure 3 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0020] In order to make the objectives, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below in combination with specific embodiments and with reference to the accompanying drawings.

[0021] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the embodiments of the present application should be the usual meanings understood by people with ordinary skills in the field to which the present application belongs. The "first", "second" and similar words used in the embodiments of the present application do not represent any order, quantity or importance, but are only used to distinguish different components. "Including" or "comprising" and similar words mean that the elements or objects appearing in front of the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0022] As described in the background technology section, with the development of communication technology, the integrated space-air network came into being. The integrated space-air network is a communication network that comprehensively utilizes satellites, aircraft (such as airships, high-altitude drones) and ground infrastructure. This communication network allows information and data to be efficiently transmitted and shared between different levels and environments, promoting applications in multiple fields, including military, civil, scientific research and emergency rescue. Satellite network is one of the important components of the integrated space-air network, and global data transmission and communication are achieved through satellite systems. Satellite networks can cover a wide geographical area, especially suitable for remote areas, oceans, mountainous areas, etc., providing a seamless connection solution for global communications. Satellite networks are complex and diverse, and accurate prediction of their network traffic can help optimize network performance, resource allocation and security management. Traffic prediction can be used to dynamically adjust network bandwidth to ensure the efficiency and quality of data transmission. Secondly, it helps to identify potential network bottlenecks and risks, thereby improving the robustness and security of the network. In addition, traffic prediction also provides a basis for network planning, which can better meet user needs, improve service quality, and reduce operating costs. Therefore, it is very necessary to predict the network traffic of satellites.

[0023] In the prior art, traditional methods for network traffic prediction mainly rely on the analysis of satellite network data, such as ARIMA and Holt-Winters. Since one of the characteristics of satellite network traffic is its strong spatiotemporal correlation, it may be affected by various factors, such as seasonality, geographical location, weather conditions, etc., so traffic prediction needs to consider the impact of different times and places, which increases the complexity of the prediction. Traditional methods are generally difficult to effectively capture the complex spatiotemporal dependencies in network data. Most existing network traffic prediction methods mainly consider spatial similarity, that is, it is assumed that network nodes with closer distances have higher similarity and are more likely to establish connections in space. However, this spatial relationship may ignore the hidden relationship between network nodes, resulting in reduced prediction accuracy.

[0024] In view of this, the present application embodiment proposes a satellite-based network traffic prediction method, referring to Figure 1 , including the following steps:

[0025] Step 101: Acquire network data corresponding to each network node in a historical time period.

[0026] In this step, with the development of satellite communication technology, satellite network traffic has a wide range of applications. However, the volatility and potential abnormalities of network traffic may have a negative impact on network performance, including congestion, delay, packet loss, etc., thereby affecting the user's network experience. Accurate prediction of network traffic is a key factor in ensuring the reliability and high quality of network services. In order to achieve the prediction of satellite network traffic, first, it is necessary to obtain network data corresponding to each network node in the historical time period, wherein the network node corresponds to the satellite one-to-one, and each network node is in a predetermined area. The historical time period is the time period corresponding to the network data obtained according to the user's prediction needs. If the user needs to predict the traffic of the satellite in the future time period, it is necessary to obtain the network data of the historical time period belonging to the same time period as the future time period. The dates corresponding to the future time period and the historical time period are correlated, and can be close dates. For example, when the future time period is 8:00-9:00 tomorrow, the historical time period can be 8:00-9:00 today; when the future time period is 20:00-23:00 the day after tomorrow, the historical time period can be 20:00-23:00 today. The dates corresponding to the future time period and the historical time period can also be the same dates in different years. For example, when the future time period is 8:00-9:00 on January 1, 2023, the historical time period can be 8:00-9:00 on January 1, 2022.

[0027] Step 102: input the network data corresponding to all network nodes into the trained prediction model, and construct the function similarity matrix and data similarity matrix corresponding to each network node through the prediction model.

[0028] In this step, the rapid development of models based on deep learning technology has brought new possibilities for network traffic prediction. The network data corresponding to all network nodes are input into the trained prediction model. It should be noted that the network data needs to be preprocessed before all network data are input into the prediction model to speed up the processing speed of the prediction model for all network data. Through the prediction model, the functional similarity matrix and data similarity matrix corresponding to each network are constructed. Through the functional similarity matrix, other network nodes with similar functions to the network node can be mined from all network nodes except the network node to obtain the hidden relationship between the network node and other network nodes at the functional level, and establish the connection between the network node and other network nodes. The data similarity matrix can mine other network nodes with similar data to the network node from all network nodes except the network node to obtain the hidden relationship between the network node and other network nodes at the data level, and establish the connection between the network node and other network nodes. By constructing the functional similarity matrix and data similarity matrix for each network node, the hidden relationship of the network data between other network nodes that are connected to each network node can be fully mined. In this way, even the network nodes with relatively large distances can establish corresponding connections, rather than just mining the relationship between the network data of the network node and its adjacent nodes. This allows the relationship between the network data between network nodes at the spatial level to be more comprehensively reflected.

[0029] Step 103: Determine the spatial features corresponding to the network nodes based on the function similarity matrix and the data similarity matrix.

[0030] In this step, the functional similarity matrix and data similarity matrix of the network node comprehensively reflect the relationship between the network data of the network node and other network nodes at the spatial level. The introduction of different types of matrices, the adjacency matrix represents different spatial relationships in the network data, can more comprehensively capture multiple spatial aspects of the network data, and provide a more comprehensive representation of the network state, which means that the relationship between different parts of the complex satellite communication network can be more accurately captured, providing more accurate input for satellite communication network traffic prediction. The relationship between the functional similarity matrix and the data similarity matrix of the network nodes at the spatial level jointly determines the spatial characteristics. The integration of diverse network data about space enables the prediction model to capture the complex spatial dependencies inherent in the network data, making the spatial characteristics of the network nodes more accurate.

[0031] Step 104: Determine the time characteristics corresponding to the network nodes based on the network data.

[0032] In this step, network data has a strong spatial and temporal correlation. Therefore, after determining the spatial characteristics of network data, it is also necessary to determine the temporal characteristics of network data. Network data in network nodes changes over time. In order to better understand the dynamics and time series relationship in network data, the time characteristics corresponding to network nodes are determined through network data. Through the time characteristics, the changing trend of network data over time can be accurately reflected.

[0033] Step 105, the spatial features and temporal features of all network nodes are fused to obtain fused feature values, the network traffic in the time period to be predicted is calculated based on the fused feature values, and the network traffic is output via the prediction model.

[0034] In this step, the spatial and temporal features of all network nodes are fused to obtain fused feature values, which comprehensively reflect the spatial and temporal features of the traffic data of all network nodes. By fusing the feature values, the network traffic of the time period to be predicted is calculated. The characteristics of the satellite's network data are fully considered, which helps to better cope with various complex situations and ensure the reliability and high-quality service of the satellite communication network. The network traffic is output through the prediction model, which can capture the rich spatial dependencies in the data, thereby improving the accuracy and robustness of the network traffic prediction and achieving the purpose of accurately predicting the network traffic.

[0035] Through the above scheme, the network data corresponding to each network node in the historical time period is obtained. The network data corresponding to all network nodes are input into the trained prediction model. Through the prediction model, the function similarity matrix and data similarity matrix corresponding to each network node are constructed, so that the relationship between the network data at the spatial level between the network nodes can be more comprehensively reflected. Based on the function similarity matrix and the data similarity matrix, the spatial characteristics corresponding to the network nodes are determined. The integration of diversified network data about space enables the prediction model to capture the complex spatial dependencies inherent in the network data, making the spatial characteristics of the network nodes more accurate. Based on the network data, the time characteristics corresponding to the network nodes are determined, which can accurately reflect the changing trend of network data over time. The spatial characteristics and time characteristics of all network nodes are fused to obtain fused feature values. Based on the fused feature values, the network traffic of the time period to be predicted is calculated, and the network traffic is output through the prediction model, so as to achieve the purpose of accurately predicting the network traffic.

[0036] In some embodiments, constructing a functional similarity matrix corresponding to each network node includes: performing binary conversion on the network data corresponding to each network node to obtain a character string; based on all the character strings, respectively calculating the Hamming distance between the network node and each other network node in all the network nodes except the network node; based on the Hamming distance and a predetermined distance, determining the functional similarity between the network node and each other network node; and constructing the functional similarity matrix based on all the functional similarities.

[0037] In this embodiment, the network data corresponding to each node is converted into a binary system, and the network data is converted into a string. Exemplarily, the binary conversion can be performed using the SAX (Symbolic Aggregate Approximation) algorithm, which can convert a segment of network data X with a length of n into a string of arbitrary length. The network data of each node is processed as follows using the SAX algorithm: (1) The network data X of the network node is converted into (x1, x2, ..., x n ) is divided into w equal-length subnetwork data segments, where w is a positive integer. (2) Calculate the average value of each subnetwork data segment and map it to a symbol, where letters in the alphabet are usually used to represent different average values. (3) Based on the number of alphabets, divide the average value into a equal-width intervals. (4) Based on the symbolization result, construct a SAX symbol sequence S = (s1, s2, ..., s w ), where s i is the SAX symbol of the ith subnetwork data segment in X. (5) The SAX symbol sequence S is used to calculate the approximate distance between time series, and the Hamming distance is introduced to compare the difference between two symbol sequences of equal length, where the Hamming distance represents the number of different characters in the corresponding positions of two strings of the same length. For the traffic sequence X = (x1, x2, ..., x n ) and Y = (y1, y2, ..., y n ) corresponds to the SAX symbol sequences S and S′, the Hamming distance H(S, S′) is calculated by the following formula:

[0038]

[0039] Among them, δ(s i ,s i ′) is used to represent the symbol s i and i Equality indicator function, if the symbols si and s i ′ are equal, then δ(s i ,s i ′)=0, if the symbol si and i ′ are not equal, then δ(s i ,s i ′)=1.

[0040] Exemplarily, assume that there are two network nodes x and y, and their network data are as follows: the network data of node x is: (10.3, 12.7, 15.2, ..., 14.4, 13.1, 16.3), and the network data of node y is: (9.8, 13.7, 20.5, ..., 9.2, 28.1, 10.9). Binary conversion and Hamming distance calculation are performed according to the steps of the SAX algorithm. First, the network data is divided into several sub-network data segments of equal length. The sub-network data segments of network node x are: (10.3, 12.7, 15.2), ..., (14.4, 13.1, 16.3), and the sub-network data segments of network node y are: (9.8, 13.7, 20.5), ..., (9.2, 28.1, 10.9). Then calculate the average value of each subnetwork data segment, the average value of network node x is: (12.7,...,9.6), and the average value of network node y is: (14.6,...,16.1). Map the average value to SAX symbols, assuming that 26 equal-width intervals (0, 5, 10, 15,...,125) are used, and use letters in the alphabet to represent different average values. The string of network node x is: ('C',...,'B'), and the string of network node y is: ('C',...,'D').

[0041] Compare the Hamming distance with the predetermined distance. The Hamming distance represents the number of different characters at corresponding positions of two strings of the same length. If the Hamming distance is greater than or equal to the predetermined value, it means that the number of different characters at corresponding positions in the two strings is relatively large, and the functional similarity between the two network nodes corresponding to the two strings is relatively low. If the Hamming distance is less than the predetermined value, it means that the number of different characters at corresponding positions in the two strings is relatively small, and the functional similarity between the two network nodes corresponding to the two strings is relatively high. By comparing the Hamming distance with the predetermined distance, the functional similarity between the network node and each other network node can be accurately determined. Based on all functional similarities, the functional similarity associated with each other network node is placed at a predetermined position in the pre-constructed first matrix framework corresponding to the network node, so as to achieve the purpose of accurately constructing a functional similarity matrix.

[0042] In some embodiments, determining the functional similarity between the network node and each other network node based on the Hamming distance and the predetermined distance includes: in response to determining that the Hamming distance is greater than or equal to the predetermined distance, determining the functional similarity between the network node and the other network nodes corresponding to the Hamming distance as 0; or, in response to determining that the Hamming distance is less than the predetermined distance, determining the functional similarity between the network node and the other network nodes corresponding to the Hamming distance as 1.

[0043] In this embodiment, when the Hamming distance is greater than or equal to the predetermined distance, it indicates that the functional similarity between the two network nodes is relatively low, and the functional similarity between the network node and other network nodes corresponding to the Hamming distance is determined to be 0. When the Hamming distance is less than the predetermined distance, it indicates that the functional similarity between the two network nodes is relatively high, and the functional similarity between the network node and other network nodes corresponding to the Hamming distance is determined to be 1. The functional similarity between the network node and each other network node is determined by the following formula:

[0044]

[0045] in, is the functional similarity, SAX(x,y) is the Hamming distance between network nodes x and y, and λ is the predetermined distance. By quantifying the functional similarity between a network node and each other network node, the purpose of accurately determining the functional similarity is achieved.

[0046] Exemplarily, the Hamming distance of the string corresponding to the network data of the network node is calculated, and then the Hamming distance is compared with the predetermined distance. If the Hamming distance is greater than or equal to the predetermined distance, the functional similarity between the network nodes is considered to be high enough, and the corresponding value in the functional similarity matrix can be set to 1, otherwise it is set to 0. The functional similarity matrix of the network nodes can be:

[0047] In some embodiments, constructing a data similarity matrix corresponding to each network node includes: based on all network data, determining the data similarity between the network node and each other network node in all network nodes except the network node; and constructing the data similarity matrix based on all data similarities.

[0048] In this embodiment, in order to construct a data similarity matrix corresponding to each network node, it is necessary to determine the data similarity between the network node and each other network node in all network nodes except the network node, and place the data similarity associated with each other network node at a predetermined position in the pre-constructed second matrix framework corresponding to the network node to obtain a constructed data similarity matrix, thereby achieving the purpose of accurately constructing a data similarity matrix.

[0049] Exemplarily, the data similarity matrix can be

[0050] In some embodiments, the determining of the data similarity between the network node and each other network node among all network nodes except the network node based on all network data includes: determining the data similarity coefficient between the network node and each other network node among all network nodes except the network node based on all network data; in response to determining that the data similarity coefficient is greater than or equal to a predetermined coefficient, determining the data similarity between the network node and the other network nodes corresponding to the data similarity coefficient as 1; or, in response to determining that the data similarity coefficient is less than a predetermined coefficient, determining the data similarity between the network node and the other network nodes corresponding to the data similarity coefficient as 0.

[0051] In this embodiment, the data similarity coefficient between the network node and each other network node among all network nodes except the network node is determined by the following formula:

[0052]

[0053] Among them, P(x,y) is the data similarity coefficient between network node x and other network nodes y, X i is the i-th sub-network data in network node x, is the average value of the network data of network node x, n is the total number of sub-network data of network node x and the total number of sub-network data of other network nodes y, the total number of sub-network data of network node x is the same as the total number of sub-network data of other network nodes y, Y i is the i-th sub-network data in other network nodes y, is the average value of the network data in other network nodes y. When the data similarity is greater than or equal to the predetermined coefficient, it means that the data similarity between the network node and other network nodes corresponding to the data similarity coefficient is relatively large, and the data similarity is determined to be 1. When the data similarity is less than the predetermined coefficient, it means that the data similarity between the network node and other network nodes corresponding to the data similarity coefficient is relatively small, and the data similarity is determined to be 0. By quantifying the data similarity between the network node and each other network node, the purpose of accurately determining the data similarity is achieved.

[0054] In some embodiments, determining the spatial features corresponding to the network nodes based on the functional similarity matrix and the data similarity matrix includes: constructing a first diagonal matrix corresponding to the functional similarity matrix based on the functional similarity matrix; constructing a second diagonal matrix corresponding to the data similarity matrix based on the data similarity matrix; determining the product of the functional similarity matrix, the first diagonal matrix, the data similarity matrix and the second diagonal matrix to obtain a fusion matrix corresponding to the network nodes; and performing feature extraction on the fusion matrix to obtain the spatial features.

[0055] In the present embodiment, the first sum of the functional similarity corresponding to each row of the functional similarity matrix is ​​calculated, and the number of rows of the first sum in the functional similarity matrix is ​​determined. The first sum in each row is placed in the same row as the number of rows of the functional similarity matrix in the first matrix framework, and the position in each row is the same as the number of rows, and 0 is placed in the remaining position in the first matrix framework, and the first diagonal matrix corresponding to the functional similarity matrix can be obtained. The second sum of the data similarity corresponding to each row of the data similarity matrix is ​​calculated and the number of rows of the second sum in the data similarity matrix is ​​determined. The second sum corresponding to each row is placed in the same row as the number of rows of the data similarity matrix in the second matrix framework, and the position in each row is the same as the row number, and 0 is placed in the remaining position in the second matrix framework, and the second diagonal matrix corresponding to the data similarity matrix can be obtained. The fusion matrix is ​​determined by the following formula:

[0056]

[0057] Among them, R is the fusion matrix, is the first diagonal matrix, A fun is the function similarity matrix, is the second diagonal matrix, A peais the data similarity matrix. Fusion of the functional similarity matrix and the data similarity matrix helps to fully represent the network status of network nodes. The functional similarity matrix and the data similarity matrix capture different aspects of the spatial relationship in the satellite network data and provide a multi-angle spatial feature extraction method. This integration of diverse spatial information can capture the complex spatial dependencies inherent in the satellite network data, thereby achieving more accurate and reliable predictions.

[0058] In some embodiments, determining the time characteristics corresponding to the network node based on the network data includes: determining the historical time characteristics and future time characteristics corresponding to the network node based on the network data, and combining the historical time characteristics and the future time characteristics to obtain the time characteristics.

[0059] In this embodiment, the historical time features and future time features corresponding to the network data enable the prediction model to capture the historical and future context of the network data, as well as the temporal dependencies of the network data, and provide a more comprehensive understanding of the temporal dynamics. The historical time features and future time features are combined to obtain time features. Through the time features, the dynamics and time series relationships in the network data can be better understood, thereby improving the performance and effect of traffic prediction, helping to better cope with various complex situations, ensuring the reliability of the satellite communication network and high-quality services, which is of great significance to the management of the satellite communication network.

[0060] In some embodiments, determining the historical time characteristics and future time characteristics corresponding to the network node based on the network data includes: determining the historical time characteristics by the following formula: in, is the historical time feature, is the information in the t-1 historical time features retained at the current time step t, ⊙ represents the multiplication of data one by one, is the t-1th historical time feature, is the first update matrix of the gated recurrent unit, x t is the input network data, is the second update matrix of the gated recurrent unit, is the information in the t-1 historical time features reset at the current time step t, f represents the update of the historical direction, t is the current time step, and tanh is the hyperbolic tangent function; the future time features are determined by the following formula: in, is the future time feature, is the information of t+1 future time features retained at the current time step t, ⊙ represents the multiplication of data one by one, is the t+1th future time feature, is the third update matrix of the gated recurrent unit, x t is the input network data, is the fourth update matrix of the gated recurrent unit, is the information in the t+1 future time features reset at the current time step t, b represents the update in the future direction, t is the current time step, and tanh is the hyperbolic tangent function.

[0061] In this embodiment, the information in the t-1 historical time features retained at the current time step t is determined by the following formula:

[0062]

[0063] Here, σ represents the Sigmoid function, which is used to map the input value to the range between 0 and 1. and is the weight matrix associated with the update, used to adjust the input network data x t and the t-1th historical time feature Impact on updates.

[0064] The information in the t+1 future time features reset at the current time step t is determined by the following formula:

[0065]

[0066] in, and is the weight matrix associated with the reset, used to adjust the input network data x t and the t-1th historical time feature Impact on reset.

[0067] The information in the t+1 future time features retained at the current time step t is determined by the following formula:

[0068]

[0069] Here, σ represents the Sigmoid function, which is used to map the input value to the range between 0 and 1. and is the weight matrix associated with the update, used to adjust the input network data x t and the t+1th future time feature Impact on updates.

[0070] The information in the t+1 future time features reset at the current time step t is determined by the following formula:

[0071]

[0072] in, and is the weight matrix associated with the reset, used to adjust the input network data x t and the t+1th future time feature Impact on reset.

[0073] Through specific formulas, the spatial and temporal characteristics are quantified, making them more intuitive.

[0074] In some embodiments, the spatial features and temporal features of all network nodes are fused to obtain a fused feature value, including: calculating a first product value of the spatial feature of each network node and the attention weight, and calculating a first sum value of all first product values; calculating a second product value of the temporal feature of each network node and the attention weight, and calculating a second sum value of all second product values; and determining the sum of the first sum value and the second sum value as the fused feature value.

[0075] In this embodiment, an attention mechanism is introduced to dynamically assign weights to each feature, which is achieved by calculating the attention weight A, where the length of A is equal to the number of features. The attention weight is calculated using the softmax function: A = softmax(Z)(9), where Z is a score vector, which is calculated as Z i =tanh(W z ·Concatenated Features[i] +b z )(10), where W z and b z is the learned parameter, Concatenated Features[i] is the comprehensive feature of the i-th spatial feature and temporal feature.

[0076] The fusion feature value is determined by the following formula:

[0077]

[0078] Among them, F final is the fusion eigenvalue, A k is the attention weight calculated by the i-th spatial feature and temporal feature, S features [k] is the spatial feature of the kth network node, T features [k] is the time feature of the kth network node, and N is the total number of multiple network nodes. Dynamically assign weights to spatial features and time features to make the determined fusion feature value more accurate.

[0079] It should be noted that the fused feature value F is passed through the fully connected layer of the prediction model. final To generate the final output, which is the predicted value of satellite communication network traffic. The fully connected layer can be expressed as: Output = W fc ·F final +b fc (12), where W fc and b fc are the weights and biases of the fully connected layer respectively.

[0080] It should be noted that the method of the embodiment of the present application can be performed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario and completed by multiple devices cooperating with each other. In the case of such a distributed scenario, one of the multiple devices can only perform one or more steps in the method of the embodiment of the present application, and the multiple devices will interact with each other to complete the described method.

[0081] It should be noted that the above describes some embodiments of the present application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the above embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0082] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, the present application also provides a satellite-based network traffic prediction device.

[0083] refer to Figure 2 , the satellite-based network traffic prediction device comprises:

[0084] The receiving module 10 is configured to obtain network data corresponding to each network node within a historical time period.

[0085] The construction module 20 is configured to input the network data corresponding to all network nodes into the trained prediction model, and construct the function similarity matrix and data similarity matrix corresponding to each network node through the prediction model.

[0086] The first determination module 30 is configured to determine the spatial features corresponding to the network nodes based on the function similarity matrix and the data similarity matrix.

[0087] The second determination module 40 is configured to determine the time feature corresponding to the network node based on the network data.

[0088] The calculation module 50 is configured to fuse the spatial features and temporal features of all network nodes to obtain a fused feature value, calculate the network traffic in the time period to be predicted based on the fused feature value, and output the network traffic via the prediction model.

[0089] Through the above-mentioned device, the network data corresponding to each network node in the historical time period is obtained. The network data corresponding to all network nodes are input into the trained prediction model. Through the prediction model, the function similarity matrix and data similarity matrix corresponding to each network node are constructed, so that the relationship between the network data at the spatial level between the network nodes can be more comprehensively reflected. Based on the function similarity matrix and the data similarity matrix, the spatial characteristics corresponding to the network node are determined. The integration of diversified network data about space enables the prediction model to capture the complex spatial dependencies inherent in the network data, making the spatial characteristics of the network nodes more accurate. Based on the network data, the time characteristics corresponding to the network node are determined, which can accurately reflect the changing trend of the network data over time. The spatial characteristics and time characteristics of all network nodes are fused to obtain fused feature values. Based on the fused feature values, the network traffic of the time period to be predicted is calculated, and the network traffic is output through the prediction model, so as to achieve the purpose of accurately predicting the network traffic.

[0090] In some embodiments, the construction module 20 is further configured to perform binary conversion on the network data corresponding to each network node to obtain a character string; based on all the character strings, respectively calculate the Hamming distance between the network node and each other network node in all network nodes except the network node; based on the Hamming distance and a predetermined distance, determine the functional similarity between the network node and each other network node; based on all functional similarities, construct the functional similarity matrix.

[0091] In some embodiments, the construction module 20 is further configured to, in response to determining that the Hamming distance is greater than or equal to the predetermined distance, determine the functional similarity between the network node and other network nodes corresponding to the Hamming distance as 0; or, in response to determining that the Hamming distance is less than the predetermined distance, determine the functional similarity between the network node and other network nodes corresponding to the Hamming distance as 1.

[0092] In some embodiments, the construction module 20 is further configured to determine the data similarity between the network node and each other network node in all network nodes except the network node based on all network data; and construct the data similarity matrix based on all data similarities.

[0093] In some embodiments, the construction module 20 is also configured to determine, based on all network data, a data similarity coefficient between the network node and each other network node in all network nodes except the network node; in response to determining that the data similarity coefficient is greater than or equal to a predetermined coefficient, determine the data similarity between the network node and the other network nodes corresponding to the data similarity coefficient as 1; or, in response to determining that the data similarity coefficient is less than a predetermined coefficient, determine the data similarity between the network node and the other network nodes corresponding to the data similarity coefficient as 0.

[0094] In some embodiments, the first determination module 30 is further configured to construct a first diagonal matrix corresponding to the functional similarity matrix based on the functional similarity matrix; construct a second diagonal matrix corresponding to the data similarity matrix based on the data similarity matrix; determine the product of the functional similarity matrix, the first diagonal matrix, the data similarity matrix and the second diagonal matrix to obtain a fusion matrix corresponding to the network node; and perform feature extraction on the fusion matrix to obtain the spatial feature.

[0095] In some embodiments, the second determination module 40 is further configured to determine the historical time characteristics and future time characteristics corresponding to the network node based on the network data, and combine the historical time characteristics and the future time characteristics to obtain the time characteristics.

[0096] In some embodiments, the second determination module 40 is further configured to determine the historical time feature by the following formula: in, is the historical time feature, is the information in the t-1 historical time features retained at the current time step t, ⊙ represents the multiplication of data one by one, is the t-1th historical time feature, is the first update matrix of the gated recurrent unit, x t is the input network data, is the second update matrix of the gated recurrent unit, is the information in the t-1 historical time features reset at the current time step t, f represents the update of the historical direction, t is the current time step, and tanh is the hyperbolic tangent function; the future time features are determined by the following formula: in, is the future time feature, is the information of t+1 future time features retained at the current time step t, ⊙ represents the multiplication of data one by one, is the t+1th future time feature, is the third update matrix of the gated recurrent unit, x t is the input network data, is the fourth update matrix of the gated recurrent unit, is the information in the t+1 future time features reset at the current time step t, b represents the update in the future direction, t is the current time step, and tanh is the hyperbolic tangent function.

[0097] In some embodiments, the calculation module 50 is also configured to calculate a first product value of the spatial feature of each network node and the attention weight, and calculate a first sum value of all first product values; calculate a second product value of the temporal feature of each network node and the attention weight, and calculate a second sum value of all second product values; and determine the sum of the first sum value and the second sum value as the fusion feature value.

[0098] For the convenience of description, the above device is described in terms of functions divided into various modules. Of course, when implementing the present application, the functions of each module can be implemented in the same or multiple software and / or hardware.

[0099] The device of the above embodiment is used to implement the corresponding satellite-based network traffic prediction method in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.

[0100] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments and methods, the present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the satellite-based network traffic prediction method as described in any of the above embodiments is implemented.

[0101] Figure 3 A more specific schematic diagram of the hardware structure of an electronic device provided in this embodiment is shown, and the device may include: a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040 are connected to each other through the bus 1050 in the device.

[0102] The processor 1010 can be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0103] The memory 1020 may be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 may store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program codes are stored in the memory 1020 and are called and executed by the processor 1010.

[0104] The input / output interface 1030 is used to connect the input / output module to realize information input and output. The input / output module can be configured in the device as a component (not shown in the figure), or it can be externally connected to the device to provide corresponding functions. The input device may include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device may include a display, a speaker, a vibrator, an indicator light, etc.

[0105] The communication interface 1040 is used to connect a communication module (not shown) to realize communication interaction between the device and other devices. The communication module can realize communication through a wired mode (such as USB, network cable, etc.) or a wireless mode (such as mobile network, WIFI, Bluetooth, etc.).

[0106] The bus 1050 includes a path that transmits information between the various components of the device (eg, the processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040).

[0107] It should be noted that, although the above device only shows the processor 1010, the memory 1020, the input / output interface 1030, the communication interface 1040 and the bus 1050, in the specific implementation process, the device may also include other components necessary for normal operation. In addition, it can be understood by those skilled in the art that the above device may also only include the components necessary for implementing the embodiments of the present specification, and does not necessarily include all the components shown in the figure.

[0108] The electronic device of the above embodiment is used to implement the corresponding satellite-based network traffic prediction method in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.

[0109] Based on the same inventive concept, corresponding to any of the above-mentioned embodiment methods, the present application also provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the satellite-based network traffic prediction method as described in any of the above embodiments.

[0110] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.

[0111] The computer instructions stored in the storage medium of the above embodiment are used to enable the computer to execute the satellite-based network traffic prediction method as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0112] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present application (including the claims) is limited to these examples. In line with the concept of the present application, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the embodiments of the present application as described above, which are not provided in detail for the sake of simplicity.

[0113] In addition, to simplify the description and discussion, and in order not to make the embodiments of the present application difficult to understand, the known power supply / ground connection with the integrated circuit (IC) chip and other components may or may not be shown in the provided drawings. In addition, the device can be shown in the form of a block diagram to avoid making the embodiments of the present application difficult to understand, and this also takes into account the fact that the details of the implementation of these block diagram devices are highly dependent on the platform to be implemented in the embodiments of the present application (that is, these details should be fully within the scope of understanding of those skilled in the art). In the case of elaborating specific details (e.g., circuits) to describe exemplary embodiments of the present application, it is obvious to those skilled in the art that the embodiments of the present application can be implemented without these specific details or when these specific details are changed. Therefore, these descriptions should be considered to be illustrative rather than restrictive.

[0114] Although the present application has been described in conjunction with specific embodiments of the present application, many replacements, modifications and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may use the embodiments discussed.

[0115] The embodiments of the present application are intended to cover all such substitutions, modifications and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present application should be included in the scope of protection of the present application.

Claims

1. A satellite-based network traffic prediction method, characterized in that: include: Obtain the network data corresponding to each network node in the historical time period; Inputting the network data corresponding to all network nodes into the trained prediction model, and constructing the function similarity matrix and data similarity matrix corresponding to each network node through the prediction model; Determining spatial features corresponding to the network nodes based on the function similarity matrix and the data similarity matrix; Based on the network data, determining a time characteristic corresponding to the network node; The spatial features and temporal features of all network nodes are integrated to obtain integrated feature values, the network traffic in the time period to be predicted is calculated based on the integrated feature values, and the network traffic is output through the prediction model; The constructing of the function similarity matrix corresponding to each network node includes: performing binary conversion on the network data corresponding to each network node to obtain a character string; based on all the character strings, respectively calculating the Hamming distance between the network node and each other network node in all the network nodes except the network node; based on the Hamming distance and a predetermined distance, determining the function similarity between the network node and each other network node; based on all the function similarities, constructing the function similarity matrix; The determining, based on the Hamming distance and the predetermined distance, the functional similarity between the network node and each other network node includes: in response to determining that the Hamming distance is greater than or equal to the predetermined distance, determining the functional similarity between the network node and the other network nodes corresponding to the Hamming distance as 0; or, in response to determining that the Hamming distance is less than the predetermined distance, determining the functional similarity between the network node and the other network nodes corresponding to the Hamming distance as 1; The constructing of the data similarity matrix corresponding to each network node comprises: based on all network data, respectively determining the data similarity between the network node and each other network node among all network nodes except the network node; and constructing the data similarity matrix based on all data similarities; The determining, based on all network data, respectively the data similarity between the network node and each other network node among all network nodes except the network node, comprises: respectively determining, based on all network data, the data similarity coefficient between the network node and each other network node among all network nodes except the network node; in response to determining that the data similarity coefficient is greater than or equal to a predetermined coefficient, determining the data similarity between the network node and the other network nodes corresponding to the data similarity coefficient as 1; or, in response to determining that the data similarity coefficient is less than a predetermined coefficient, determining the data similarity between the network node and the other network nodes corresponding to the data similarity coefficient as 0; The determining of the spatial features corresponding to the network nodes based on the function similarity matrix and the data similarity matrix includes: constructing a first diagonal matrix corresponding to the function similarity matrix based on the function similarity matrix; constructing a second diagonal matrix corresponding to the data similarity matrix based on the data similarity matrix; determining the product of the function similarity matrix, the first diagonal matrix, the data similarity matrix and the second diagonal matrix to obtain a fusion matrix corresponding to the network node; Performing feature extraction on the fusion matrix to obtain the spatial feature; The determining, based on the network data, the time feature corresponding to the network node comprises: determining, based on the network data, a historical time feature and a future time feature corresponding to the network node, and combining the historical time feature and the future time feature to obtain the time feature; The determining, based on the network data, the historical time features and the future time features corresponding to the network nodes includes: determining the historical time features by the following formula: in, is the historical time feature, is the information in the t-1 historical time features retained at the current time step t, ⊙ represents the multiplication of data one by one, is the t-1th historical time feature, is the first update matrix of the gated recurrent unit, x t is the input network data, is the second update matrix of the gated recurrent unit, is the information in the t-1 historical time features reset at the current time step t, f represents the update of the historical direction, t is the current time step, and tanh is the hyperbolic tangent function; The future time characteristic is determined by the following formula: in, is the future time feature, is the information of t+1 future time features retained at the current time step t, ⊙ represents the multiplication of data one by one, is the t+1th future time feature, is the third update matrix of the gated recurrent unit, x t is the input network data, is the fourth update matrix of the gated recurrent unit, is the information in the t+1 future time features reset at the current time step t, b represents the update in the future direction, t is the current time step, and tanh is the hyperbolic tangent function.

2. The method according to claim 1, characterized in that: The fusion of the spatial features and temporal features of all network nodes to obtain fusion feature values ​​includes: Calculate the first product value of the spatial feature and the attention weight of each network node, and calculate the first sum of all the first product values; Calculating a second product value of the time feature of each network node and the attention weight, and calculating a second sum value of all second product values; The sum of the first sum value and the second sum value is determined as the fused feature value.

3. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method according to any one of claims 1 to 2 is implemented.

Images