A method and system for causal discovery of wireless network performance metrics

Abstract

This invention relates to the technical field of artificial intelligence and causal relationship discovery, and proposes a causal discovery method and system for wireless network performance indicators, comprising the following steps: collecting performance indicator data of several observed variables of a wireless network to form wireless network performance indicator data; calculating the causal difference between the observed variables of the wireless network performance indicator data using higher-order cumulants; searching for the root node of the causal structure corresponding to the wireless network performance indicator data level by level based on the causal difference; considering the order in which the root nodes are searched as the causal order, and recovering the causal structure of the wireless network performance indicator data according to the causal order; the causal discovery method of this application can accurately identify causal relationships and has low computational complexity.

Description

Technical Field

[0001] This invention relates to the technical field of artificial intelligence and causal relationship discovery, and more specifically, to a causal discovery method and system for wireless network performance indicators. Background Technology

[0002] Generally, analyzing wireless network performance is a complex and crucial task, involving the evaluation of various network metrics, such as signaling channel availability, signaling channel congestion rate, channel call drop rate, voice channel availability, voice channel congestion rate, voice call drop rate, handover success rate, dual-frequency handover success rate, wireless channel utilization, and uplink / downlink voice quality. These metrics, from overall to specific, permeate all stages of wireless network operation and are the primary means of evaluating network performance. However, identifying and improving the causal relationships between these performance metrics is a key challenge in enhancing wireless network performance. In specific situations, inferring these causal relationships can help us find methods to optimize wireless network performance.

[0003] Several methods have been proposed in the prior art to identify causal relationships between two observed variables by discovering asymmetries. Independence tests are a typical tool used to determine the (non)independence between two observed variables. To determine the causal direction between two adjacent variables, conditional independence tests (CITs) and V-structures can be used to identify conditions. The Hilbert-Schmidt independence criterion (HSIC) test and the kernel-based conditional independence (KCI) test are two typical CIT methods in the linear non-Gaussian or nonlinear cases. They use kernel-based techniques to measure the dependencies between variables in a reproducing kernel Hilbert space (RKHS). This approach can introduce a large amount of uncertainty, leading to redundant information in characterizing the data. In this case, their performance depends on a large sample size, and the computational complexity is proportional to the cube of the sample size.

[0004] Existing techniques also propose several methods for estimating linear non-Gaussian acyclic models. For example, one approach is based on Independent Component Analysis (ICA), which estimates a mixture matrix showing how independent components influence the observed variables. The mixture matrix is ​​then converted into a causal strength matrix. However, ICA can get stuck in local optima and yield unreasonable solutions. Existing techniques also propose methods based on independence noise conditions, utilizing the independence between the residuals obtained after regression and the assumed dependent variable. For example, the DirectLiNGAM algorithm identifies the variable with the highest independence as an exogenous variable, then eliminates the influence of the identified exogenous variable and continues searching for the variable with the highest independence to reconstruct the causal graph.

[0005] These methods primarily employ least squares regression. However, least squares regression has limitations when dealing with non-Gaussian data, often leading to inaccurate residuals and affecting subsequent independence tests. This results in problems such as inaccurate identification of causal relationships for noisy heteroscedastic variables, high computational complexity, or poor performance when handling noisy non-Gaussian heteroscedastic variable data when processing wireless network performance metrics. Summary of the Invention

[0006] To overcome the shortcomings of inaccurate causal relationship identification and computation in the prior art, this invention provides a causal discovery method for wireless network performance indicators that is accurate in identifying causal relationships and has low computational complexity.

[0007] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0008] Collect performance index data of several observed variables of the wireless network to form wireless network performance index data;

[0009] Using higher-order cumulants, the causal difference between the observed variables of the wireless network performance index data is calculated;

[0010] Based on the causal difference, the root node of the causal structure corresponding to the wireless network performance index data is found step by step;

[0011] The order in which the root nodes are found is a causal order, and the causal structure of the wireless network performance index data is recovered based on the causal order.

[0012] This invention also proposes a causal discovery system for wireless network performance metrics to implement the aforementioned causal discovery method for wireless network performance metrics, the system comprising:

[0013] The data acquisition module is used to collect performance index data of several observed variables of the wireless network and form wireless network performance index data.

[0014] The causal difference calculation module is used to calculate the causal difference between the observed variables of the wireless network performance index data using higher-order cumulants.

[0015] The root node search module is used to search for the root node of the causal structure corresponding to the wireless network performance index data step by step based on the causal difference.

[0016] The causal structure recovery module is used to recover the causal structure of the wireless network performance index data based on the causal order in which the root nodes are found.

[0017] The present invention also proposes a computer device including a memory and a processor, wherein the memory stores computer-readable instructions, which, when executed by the processor, cause the processor to perform the steps of the above-described causal discovery method for wireless network performance metrics.

[0018] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

[0019] This invention utilizes higher-order cumulants to calculate the causal difference between observed variables of the wireless network performance index data, and searches for the root node of the causal structure corresponding to the wireless network performance index data level by level based on the causal difference. The order in which the root nodes are found is regarded as the causal order, and the causal structure of the wireless network performance index data is restored according to the causal order. It takes advantage of the causal order consistency of the causal difference between the root node and the leaf node, effectively handles the noise and heteroscedasticity problems that may exist in the data, can accurately identify causal relationships and has low computational complexity. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the causal discovery method for wireless network performance metrics proposed in Example 1.

[0021] Figure 2 This is a schematic diagram of the step-by-step search for the root node proposed in Example 1;

[0022] Figure 3 This is an overall framework diagram of a causal discovery system for wireless network performance metrics proposed in Example 3. Detailed Implementation

[0023] The accompanying drawings are for illustrative purposes only and should not be construed as limiting the scope of this embodiment.

[0024] To better illustrate this embodiment, some parts in the accompanying drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions;

[0025] It will be understood by those skilled in the art that certain well-known structures and their descriptions may be omitted in the accompanying drawings.

[0026] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0027] Example 1

[0028] This embodiment proposes a causal discovery method for wireless network performance metrics. Figure 1 This is a flowchart illustrating the causal discovery method for wireless network performance metrics in this embodiment.

[0029] The causal discovery method for wireless network performance metrics proposed in this embodiment includes the following steps:

[0030] S1: Collect performance index data of several observed variables of the wireless network to form wireless network performance index data;

[0031] S2: Calculate the causal difference between the observed variables of the wireless network performance index data using higher-order cumulants;

[0032] S3: Based on the causal difference, find the root node of the causal structure corresponding to the wireless network performance index data step by step;

[0033] S4: The order in which the root nodes are found is a causal order, and the causal structure of the wireless network performance index data is recovered according to the causal order.

[0034] In the specific implementation process, the causal difference between the observed variables of the wireless network performance index data is calculated by using higher-order cumulants. Based on the causal difference, the root node of the causal structure corresponding to the wireless network performance index data is searched level by level. The order in which the root nodes are found is regarded as the causal order. The causal structure of the wireless network performance index data is restored according to the causal order. By utilizing the characteristic that the causal difference between the root node and the leaf node has the consistency of causal order, the problem of noise heteroscedasticity that may exist in the data can be effectively handled. It can accurately identify causal relationships and has low computational complexity.

[0035] In an optional embodiment, the expression for the wireless network performance metric data includes:

[0036] X = [x1, x2, ..., x i ,…,x N ]

[0037] In the formula, x i This represents the performance metric data of the i-th observed variable of the wireless network, and N represents the total number of observed variables;

[0038] Before calculating the causal difference between the observed variables of the wireless network performance index data using higher-order cumulants, the wireless network performance index data is standardized. The expression for the standardization process includes:

[0039]

[0040] In the formula, Represents x after standardization. i mean(x) i ) represents x i The mean.

[0041] In an optional embodiment, the expression for calculating the causal difference between the observed variables of the wireless network performance index data using higher-order cumulants includes:

[0042] τ ij =|C4(x) i C 1,3 (x i ,x j )-C 2,2 (x i ,x j C 3,1 (x i ,x j )|

[0043] C4(x i ) = Cum(x i ,x i ,x i ,x i )

[0044] C 1,3 (x i ,x j ) = Cum(x i ,x j ,x j ,x j )

[0045] C 2,2 (x i ,x j ) = Cum(x i ,x i ,x j ,x j )

[0046] C 3,1 (x i ,x j ) = Cum(x i ,x i ,x i ,x j )

[0047] In the formula, τ io The i-th observation variable x represents the wireless network performance index data. i and the j-th observed variable x j The causal difference; Cum(·) denotes a higher-order cumulant function.

[0048] In an optional embodiment, the step of finding the root node of the causal structure corresponding to the wireless network performance index data step by step based on the causal difference includes:

[0049] S1: For each observed variable in the wireless network performance index data, calculate the sum of the causal differences between it and other observed variables to obtain the sum of causal differences corresponding to each observed variable;

[0050] S2: The observed variable with the smallest causal difference is regarded as the root node x of the causal structure corresponding to the wireless network performance index data. m ;

[0051] S3: Remove the root node x from the wireless network performance index data. m The effects on other observed variables are used to obtain the residual matrix R;

[0052] S4: Using each observed variable in the residual matrix R as the operation object, iteratively execute steps S1 to S3, find the root node level by level, and record the order in which the root nodes are found, until all observed variables are identified as root nodes, then stop the iteration and obtain the causal order of the wireless network performance index data.

[0053] In this optional embodiment, Figure 2 This is a schematic diagram of the step-by-step search for the root node proposed in this embodiment; Figure 2 The steps are shown to find the root node of the causal structure corresponding to the wireless network performance index data step by step based on the causal difference.

[0054] In an optional embodiment, the expression for the causal difference sum includes:

[0055]

[0056] In the formula, T τ (x i ;U) represents x i The causal difference sum, where U represents the index set of variables in the feature set X, and U\i represents removing the i-th variable x from the index set U. i .

[0057] In an alternative embodiment, the root node x m The expressions include:

[0058]

[0059] In the formula, T τ (x i ;U) represents x i The difference between cause and effect.

[0060] In an optional embodiment, the expression for the residual matrix R includes:

[0061] R = [∈1, ∈2, ..., ∈ m-1 ,∈ m+1 ,…,∈N ]

[0062] x pi =b mi x m -∈ i

[0063] In the formula, x m Represents the root node; x pi b represents the observed variable corresponding to the non-root node; mi Represents the regression coefficient; ∈ m-1 Represents the observed variable x m-1 The corresponding residuals also represent the values ​​after removing the observed variable x. m-1 After the influence, the observed variable x m-1 The remaining part; ∈ i x represents pi The corresponding residuals; N represents the total number of observed variables.

[0064] In an optional embodiment, the step of recovering the causal structure of the wireless network performance metric data according to the causal order includes:

[0065] Construct a lower triangular matrix B using causal order, and let the element in the i-th row and j-th column of the lower triangular matrix B be B. ij If i ≤ j, then B ij =0 means x i For x j There is no direct causal relationship;

[0066] Otherwise, B ij In causal order, the observed variable x i For the observed variable x j The direct causal effect, in this case, let the observed variable x j The causal parent node is {x1, x2, ..., x j-1}, then B ij The expressions include:

[0067] B ij =b ji

[0068]

[0069] In the formula, ∈ j x represents j The corresponding residuals.

[0070] Example 2

[0071] This embodiment, based on the causal discovery method for wireless network performance metrics proposed in Embodiment 1, presents the following specific implementation examples:

[0072] This invention provides a method for discovering causal relationships based on higher-order cumulants, comprising:

[0073] S1: Set the observation data matrix X and standardize the data;

[0074] S2: For the standardized data, calculate the causal difference between each variable and other variables, determine the causal difference between each variable, and then use the summation method to sum the causal differences between each variable and other variables respectively to obtain the sum of the causal differences of each variable;

[0075] S3: Based on the sum of causal differences for each variable, determine the variable with the smallest sum of causal differences as the root node;

[0076] S4: Perform regression on the root node for the other variables and calculate the residual matrix R.

[0077] S5: Replace the data matrix X with the residual matrix R, and repeat steps 2 to 4 for R until all root nodes are found in sequence. The order in which the root nodes are found is the causal order. Restore the causal structure based on the causal order.

[0078] S6: Based on the causal relationship discovery algorithm, a directed acyclic graph model is found to obtain the causal relationship results between wireless network performance index data and feed them back to wireless network maintenance personnel, thus providing assistance for scientific discovery in wireless network performance optimization problems.

[0079] The technical solution for causal discovery of wireless network performance index data is a method for discovering linear non-Gaussian causal structures. It identifies two-variable causal structures by establishing statistics (causal difference) based on higher-order cumulants. The method is extended to multivariate cases, and its advantages over other methods in noisy and heteroscedastic situations are demonstrated. Its key feature is the consideration of higher-order information of the observed variables. Combining the idea of ​​the multivariate Directingam algorithm, it transforms a difficult problem into an easier one by identifying the root node of the causal order. The method includes the following steps:

[0080] Set the wireless network performance metrics data X = [x1, x2, ... x N ], where x i Let represent the i-th observed variable. To ensure the index satisfies the algorithm's assumptions, the data needs to be standardized. The specific steps are as follows:

[0081] Each variable x i Transform to zero mean, that is, for each x i Subtract its mean:

[0082]

[0083] Ensure that the mean of the data is 0.

[0084] Next, the causal difference τ between each pair of variables is calculated using higher-order cumulants. ij Higher-order cumulants are important tools for measuring linear correlations between random variables. Here, we use fourth-order cumulants and mixed cumulants to capture causal relationships in the data:

[0085] τ ij =|C4(x) i C 1,3 (x i ,x j )-C 2,2 (x i ,x j C 3,1 (x i ,x j )|

[0086] Among them, the fourth-order accumulator C4(x) i ) represents a measure of a single variable x i The fourth-order statistical information reflects its higher-order structural characteristics. Mixed accumulation C 1,3 (x i ,x j ) and C (2,2) (x i ,x j ) Measure the two variables x respectively i and x j The joint higher-order features; It is the accumulated amount, denoted as

[0087] By employing a summation method, the causal impact of each performance metric on the overall network is quantified through the calculation of the causal difference, especially in the case of noise and heteroscedasticity. For each performance metric x... i The causal difference for each indicator is obtained by summing the causal differences with other indicators:

[0088] T τ (x i ;U)=∑ {j∈U\i} τ ij .

[0089] Causal difference and T τ (x i ;U) can help identify the causal effect of each indicator on other indicators under the influence of noise heteroscedasticity, and the T of the variables τ (x i The smaller the value of U, the less the indicator is affected by other indicators.

[0090] The root node in a causal relationship is determined by identifying the index with the smallest sum of causal differences. Based on the sum of causal differences for each variable, the variable with the smallest sum of causal differences is identified as the root node, calculated as follows:

[0091]

[0092] Where, x m It is the root node.

[0093] This application can effectively identify the root node that is least affected by other variables in the case of heteroscedastic noise. For example, if the causal difference in bandwidth utilization is minimized, it may indicate that it is the root cause of other performance metrics such as latency or packet loss rate.

[0094] By calculating and iterating the residual matrix, the influence of exogenous variables is gradually removed, each root node is found step by step, and the causal order of wireless network performance indicators is finally determined.

[0095] In the first root node x m Once identified, this variable is known to be the least affected by other variables in the causal structure, and is therefore considered the starting point of the causal relationship. At this point, it is necessary to extract the influence of this root node from the data to obtain the residual matrix of the remaining variables.

[0096] Determine the root node x m Next, we need to remove the influence of the root node on other variables from the data matrix X. Assume the current data matrix X contains N variables, x m If it is the first identified root node, then the steps for calculating the residual matrix R are as follows:

[0097] For each non-root node x i (i≠m) Perform regression, assuming it is the root node x m Linear functions:

[0098] x i =b mi x m -∈ i

[0099] Here b mi It is the regression coefficient, ∈ i It is the residual, that is, after removing the root node x. m After the influence, variable x i The remaining part.

[0100] By using the regression equation, the residuals ∈ i These residuals form a new residual matrix R:

[0101] R = [∈1, ∈2, ..., ∈ m-1 ,∈O+1 ,…,∈ N ]

[0102] In this new matrix R, each column represents a variable from the original data matrix after removing the root node x. m The residuals after the influence of the effect.

[0103] To maintain data consistency, the residual matrix R is standardized to zero mean.

[0104] After obtaining the residual matrix R, repeat steps 2 to 4 to identify the next root node. The specific process is as follows: recalculate the causal difference τ for each variable using the residual matrix R. ij and its causal difference and T τ (x i ;U);Find the variable x that minimizes the sum of causal differences m′ Treat it as the new root node; then remove the root node x again using linear regression. m′ The influence of each root node on other variables is gradually removed until all variables are identified as root nodes in sequence.

[0105] Through multiple iterations, the causal order of all wireless network performance metrics can eventually be determined. This method is unaffected by noise heteroscedasticity on the causal structure, making the identified causal relationships more accurate and reliable.

[0106] In methods based on higher-order cumulants, the problem of noise and heteroscedasticity that may exist in the data can be effectively handled by calculating the causal difference between each variable and other variables. Specifically, the causal difference between exogenous variables (root nodes) and endogenous variables (leaf nodes) has causal order consistency, that is, the smaller the causal difference, the more likely the variable is to be the root node in the causal structure.

[0107] This consistency provides an important criterion for identifying causal structures, enabling the method in this application to accurately identify causal relationships even when performance data contains noise and heteroscedasticity. Traditional causal discovery methods often lack this consistency, which may lead to biases or errors in causal structure identification when faced with noise and heteroscedasticity.

[0108] Therefore, the method based on high-order cumulants can still maintain high recognition performance under noisy heteroscedastic conditions, effectively revealing the true causal relationship between wireless network performance indicators and providing a more reliable basis for network performance optimization.

[0109] This iterative processing method for residual matrices can effectively identify key causal relationships in wireless network performance metrics, even in the presence of significant noise and heteroscedasticity in the data, thus providing a solid foundation for subsequent network performance optimization.

[0110] By using the identified root nodes and their causal order, the data causal structure of the entire wireless network performance indicators is gradually restored, thereby clarifying the causal relationships between the various performance indicators.

[0111] After multiple rounds of residual matrix calculation and root node identification, the causal order of wireless network performance indicators has been obtained. This order reflects the hierarchical relationship of each performance indicator in the causal structure, that is, the variable first identified as the root node is at the beginning of the causal chain, and the variables subsequently identified are influenced by the variables preceding them in turn.

[0112] To systematically recover the causal structure, a lower triangular matrix B is constructed using the causal order. Each element of this matrix B... ij Represents variable x i For x j The causal influence. Matrix B is an N×N matrix, where N is the number of performance indicators. The elements of matrix B... ij Defined as follows: If i > j, then B ij Indicating the causal order, x i For x j The direct causal effect; if i≤j, then B ij =0 means x i For x j There is no direct causal effect because, in the causal order, x i It cannot affect x in the order that follows it. j .

[0113] For example, if the causal order is x1→x2→x3, then the structure of the lower triangular matrix is:

[0114]

[0115] To determine the specific values ​​of the non-zero elements in matrix B (i.e., causal connection coefficients), regression analysis is required for each performance metric. For each performance metric x... j (j>1), perform regression analysis to estimate the influence of its causal parent node. Assume x j The causal parent node is {x1, x2, ..., x j-1 Then the regression model can be expressed as:

[0116]

[0117] b ji These are the regression coefficients that need to be estimated, representing x.i For x j The direct causal influence. After constructing the lower triangular matrix B and determining the connection coefficients, the causal structure between wireless network performance indicators can be recovered through this matrix.

[0118] Once the final causal structure is determined, the causal relationships in wireless network performance metrics data can be obtained from the causal graph. The application of causal relationships is of great significance in various aspects such as wireless network performance optimization, troubleshooting, prediction, and control. Through detailed causal structure analysis, the mutual influences between network performance metrics can be fundamentally understood, thus providing strong support for practical applications.

[0119] This application enables the discovery of causal relationships in wireless network performance data with heteroscedastic noise by establishing a two-variable causal relationship discovery method using high-order cumulants and a multi-variable causal relationship discovery algorithm to aggregate the two-variable causal relationships. This allows for the discovery of causal structures in noisy, heteroscedastic observation data. Furthermore, since causal differences are based on cumulants, and finite-order cumulants can be calculated from finite-order moments, the complexity is reduced while still maintaining good identification performance. This allows for more accurate and rapid discovery of causal mechanisms in noisy, heteroscedastic observation variables, providing significant assistance for subsequent scientific discovery of wireless network performance optimization problems and wireless network maintenance.

[0120] Example 3

[0121] This embodiment proposes a causal discovery system for wireless network performance indicators, which is used to implement the causal discovery method for wireless network performance indicators proposed in Embodiment 1.

[0122] Figure 3 This is an overall framework diagram of the causal discovery system for wireless network performance metrics in this embodiment.

[0123] The causal discovery system for wireless network performance metrics includes:

[0124] The data acquisition module is used to collect performance index data of several observed variables of the wireless network and form wireless network performance index data.

[0125] The causal difference calculation module is used to calculate the causal difference between the observed variables of the wireless network performance index data using higher-order cumulants.

[0126] The root node search module is used to search for the root node of the causal structure corresponding to the wireless network performance index data step by step based on the causal difference.

[0127] The causal structure recovery module is used to recover the causal structure of the wireless network performance index data based on the causal order in which the root nodes are found.

[0128] This embodiment proposes a computer device, including a memory and a processor. The memory stores computer-readable instructions, which, when executed by the processor, cause the processor to perform the steps of the causal discovery method for wireless network performance indicators described in Embodiment 1.

[0129] It is understood that the causal discovery system and computer device for wireless network performance indicators in this embodiment improve upon the method of embodiment 1. The options in embodiment 1 are also applicable to this embodiment, so they will not be described again here.

[0130] The same or similar labels correspond to the same or similar parts;

[0131] The terms used to describe positional relationships in the accompanying drawings are for illustrative purposes only and should not be construed as limiting the scope of this embodiment.

[0132] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A causal discovery method for wireless network performance metrics, characterized in that, Includes the following steps: Collect performance index data of several observed variables of the wireless network to form wireless network performance index data; Using higher-order cumulants, the causal difference between the observed variables of the wireless network performance index data is calculated; Based on the causal difference, the root node of the causal structure corresponding to the wireless network performance index data is searched level by level. The order in which the root nodes are searched is a causal order, and the causal structure of the wireless network performance index data is recovered based on the causal order. The expression for calculating the causal difference between observed variables of the wireless network performance index data using higher-order cumulants includes: In the formula, The first of the wireless network performance index data Observed variables and the Observed variables The difference in causality; Represents a higher-order cumulant function; The steps of searching for the root node of the causal structure corresponding to the wireless network performance index data step by step based on the causal difference include: S1: For each observed variable in the wireless network performance index data, calculate the sum of the causal differences between it and other observed variables to obtain the sum of causal differences corresponding to each observed variable; S2: The observed variable with the smallest causal difference is regarded as the root node of the causal structure corresponding to the wireless network performance index data. ; S3: Remove the root node from the wireless network performance index data. The effects on other observed variables are used to obtain the residual matrix. ; S4: Using the residual matrix Each observed variable is the operation object. Steps S1 to S3 are executed iteratively to search for the root node level by level and record the order in which the root nodes are searched. The iteration stops when all observed variables are identified as root nodes, and the causal order of the wireless network performance index data is obtained. The steps for restoring the causal structure of the wireless network performance metric data according to the causal order include: Construct a lower triangular matrix using causal order Let the lower triangular matrix be... The Line 1 The elements of the column are ,like ,but ,express right There is no direct causal relationship; otherwise, Indicating the observed variable in the causal order For observed variables The direct causal effect, in this case, let the observed variable be... The causal parent node is ,but The expressions include: In the formula, express The corresponding residuals.

2. The causal discovery method for wireless network performance metrics according to claim 1, characterized in that, The expressions for the wireless network performance metric data include: In the formula, The first part represents the wireless network. Performance index data for each observed variable, Indicates the total number of observed variables; Before calculating the causal difference between the observed variables of the wireless network performance index data using higher-order cumulants, the wireless network performance index data is standardized. The expression for the standardization process includes: In the formula, Indicates the standardized processing , express The mean.

3. The causal discovery method for wireless network performance metrics according to claim 1, characterized in that, The expression for the causal difference sum includes: In the formula, express The difference between cause and effect, Representation of feature set The set of subscripts of variables in the middle. Indicates in the subscript set Remove the first Variables .

4. The causal discovery method for wireless network performance metrics according to claim 1, characterized in that, The root node The expressions include: In the formula, express The difference between cause and effect.

5. The causal discovery method for wireless network performance metrics according to claim 1, characterized in that, The residual matrix The expressions include: In the formula, Indicates the root node; This represents the observed variable corresponding to the non-root node; Represents the regression coefficient; Represents observed variables The corresponding residuals also indicate the elimination of observed variables. After the influence, observed variables The remaining part; express The corresponding residuals; This represents the total number of observed variables.

6. A causal discovery system for wireless network performance metrics, used to implement the causal discovery method for wireless network performance metrics as described in any one of claims 1 to 5, characterized in that, include: The data acquisition module is used to collect performance index data of several observed variables of the wireless network and form wireless network performance index data. The causal difference calculation module is used to calculate the causal difference between the observed variables of the wireless network performance index data using higher-order cumulants. The root node search module is used to search for the root node of the causal structure corresponding to the wireless network performance index data step by step based on the causal difference. The causal structure recovery module is used to recover the causal structure of the wireless network performance index data based on the causal order in which the root nodes are found.

7. A computer device comprising a memory and a processor, wherein the memory stores computer-readable instructions, characterized in that, When the computer-readable instructions are executed by the processor, the processor performs the steps of the causal discovery method for wireless network performance metrics as described in any one of claims 1 to 5.

Images