Data classification methods, devices, equipment, media and products

Abstract

This disclosure provides a data classification method, apparatus, device, medium, and product, relating to the field of data processing technology. The method includes: converting data feature vectors used for QSVM model training into corresponding quantum states using quantum circuits; calculating a quantum kernel function based on the inner product of the quantum states and constructing a corresponding dual optimization objective function; determining the parameters of the optimal classification hyperplane in the QSVM model based on the optimal solution of the dual optimization objective function to complete the construction of the QSVM model; the constructed QSVM model uses a quantum processor to perform quantum state mapping on the data feature vectors to be tested, and calculates the classification decision boundary using the quantum kernel function to output the corresponding classification result. Through the embodiments of this disclosure, the nonlinear correlation between data feature vectors is captured more accurately, not only improving the computational efficiency of the QSVM model based on the quantum kernel function but also enhancing the reliability and accuracy of classifying data feature vectors.

Description

Technical Field

[0001] This disclosure relates to the field of data processing technology, and more specifically, to a data classification method, apparatus, device, medium, and product. Background Technology

[0002] Currently, with the rapid development of wireless communication technology, modern communication networks are becoming increasingly complex. Especially in 5G and future 6G networks, the number of communication nodes such as base stations and cells has increased significantly, and their topology and dynamic characteristics present numerous challenges to network operation and maintenance. The stable operation of communication networks is crucial for ensuring critical business operations; however, frequent cell failures can lead to widespread network service outages, decreased user experience, and even economic losses.

[0003] In related technologies, classic machine learning algorithms are used to classify communication network data in the context of community fault detection. However, the computational complexity of high-dimensional feature data is high, making it difficult to converge quickly in multi-dimensional and complex scenarios. In addition, existing root cause fault localization methods are still at the static analysis stage, making it difficult to dynamically capture the nonlinear correlations between features.

[0004] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this disclosure is to provide a data classification method, apparatus, device, medium, and product that at least partially overcomes problems caused by limitations and defects in related technologies.

[0006] According to a first aspect of the present disclosure, a data classification method is provided, comprising: converting a data feature vector used for training a QSVM model into a quantum state corresponding to the data feature vector using a quantum circuit; calculating a quantum kernel function based on the inner product of the quantum states and constructing a corresponding dual optimization objective function; determining the parameters of the optimal classification hyperplane in the QSVM model based on the optimal solution of the dual optimization objective function, thereby completing the construction of the QSVM model; wherein the constructed QSVM model performs quantum state mapping on the data feature vector to be tested using a quantum processor and calculates the classification decision boundary using the quantum kernel function, thereby outputting the corresponding classification result.

[0007] In one exemplary embodiment of this disclosure, calculating the quantum kernel function based on the inner product of the quantum states and constructing the corresponding dual optimization objective function includes:

[0008] The quantum kernel function is calculated based on the inner product of the quantum state, and the optimization objective function of the QSVM model is constructed.

[0009] The optimization objective function is transformed into a dual optimization objective function represented by a quantum kernel function by introducing Lagrange multipliers.

[0010] In one exemplary embodiment of this disclosure, the expression of the optimization objective function includes N represents the number of data feature vectors, the Characterizing the regularization term, the The b represents the classification error term, and the b represents the classification bias. The slack variable represents the i-th data feature vector, and C represents the regularization parameter.

[0011] In one exemplary embodiment of this disclosure, the constraint condition corresponding to the optimization objective function is:

[0012] Among them, the The label representing the i-th data feature vector, the The quantum feature map characterizing the i-th data feature vector, the quantum feature map Through parameterized quantum circuits Implementation, wherein the quantum circuit Composed of a single-qubit rotation gate and a controlled entanglement gate, the aforementioned Encoded as quantum state .

[0013] In one exemplary embodiment of this disclosure, the expression of the dual optimization objective function includes , , wherein Representing the j-th Lagrange multiplier, the... Representing the i-th Lagrange multiplier, the The label representing the j-th data feature vector, the Characterization pairs and The quantum kernel function used for computation, wherein the quantum kernel function is used to calculate the squared modulus of the inner product of two quantum states, the The quantum state characterizing the i-th data eigenvector, the The quantum state characterizing the j-th data eigenvector.

[0014] In one exemplary embodiment of this disclosure, the quantum kernel function is implemented by executing a quantum circuit. And obtained by measuring the ground state probability estimate, the Characterize the Perform Hermitian conjugate operations.

[0015] In one exemplary embodiment of this disclosure, the quantum kernel function is verified by sampling statistics, which include quantum state tomography and / or SWAP test estimation. The quantum kernel function is configured to satisfy the Mercer condition, and the Gram matrix of the quantum kernel function is corrected for semi-positive definiteness after measurement by quantum hardware.

[0016] In one exemplary embodiment of this disclosure, a quantum-classical hybrid algorithm is used to solve the dual optimization objective function. The quantum-classical hybrid algorithm runs on a variable quantum eigenvalue solver. The quantum hardware in the variable quantum eigenvalue solver is used to calculate the kernel matrix terms of the dual optimization objective function, and the classical device in the variable quantum eigenvalue solver is used to calculate the gradient and update the Lagrange multipliers until the convergence condition is met.

[0017] In one exemplary embodiment of this disclosure, the constraints of the dual optimization objective function include: .

[0018] In one exemplary embodiment of this disclosure, determining the parameters of the optimal classification hyperplane in the QSVM model based on the optimal solution of the dual optimization objective function to complete the construction of the QSVM model includes:

[0019] The Lagrange multipliers and support vectors are determined based on the optimal solution of the dual optimization objective function.

[0020] The weight vector is calculated based on the Lagrange multipliers, the labels of the data feature vectors, and the quantum feature maps corresponding to the data feature vectors;

[0021] The classification bias is calculated based on the support vectors, the labels of the data feature vectors, and the quantum kernel function.

[0022] In one exemplary embodiment of this disclosure, the expression for calculating the weight vector based on the Lagrange multipliers, the labels of the data feature vectors, and the quantum feature maps corresponding to the data feature vectors includes: .

[0023] In one exemplary embodiment of this disclosure, the expression for calculating the classification bias based on the support vector, the labels of the data feature vectors, and the quantum kernel function includes: , wherein The support vector is represented by k, which represents the index of the support vector. S The set of indices that characterize the support vectors.

[0024] In one exemplary embodiment of this disclosure, the quantum circuit uses either angle encoding or amplitude encoding. The process of converting the data feature vectors used for QSVM model training into corresponding quantum states via the quantum circuit includes:

[0025] The data feature vector is mapped to a quantum state of the data feature vector using either angle encoding or amplitude encoding.

[0026] In one exemplary embodiment of this disclosure, the expression for mapping the data feature vector to the rotation angle of a qubit using an angle encoding method includes: Wherein, R represents a rotation gate about the coordinate axes of a Cartesian coordinate system, the coordinate axes being the x-axis, y-axis, or z-axis, and d represents the dimension of the data feature vector. The quantum state characterizing the data feature vector, the Characterizing the tensor product operator, the The standardized processing result characterizes the feature vector of the data.

[0027] In one exemplary embodiment of this disclosure, it further includes:

[0028] The constructed QSVM model is validated using the validation dataset corresponding to the data feature vectors.

[0029] Based on the validation results, the QSVM model is either tuned or tested on a test dataset.

[0030] The results include at least one of accuracy, precision, recall, and F1 score, and the parameter tuning process includes at least one of adjusting the feature mapping method, increasing the number of qubits, optimizing the parameters in the feature mapping circuit, and adding a regularization factor to the quantum kernel function.

[0031] In one exemplary embodiment of this disclosure, before converting the data feature vector used for QSVM model training into the quantum state corresponding to the data feature vector through a quantum circuit, the method further includes:

[0032] The collected data features are standardized, and the standardization process is either normalization or conversion to a normal distribution.

[0033] The standardized data features are vectorized to obtain the data feature vector.

[0034] In one exemplary embodiment of this disclosure, the expression for standardizing the collected data features includes: , wherein Characterizing the feature value of the nth column, the and stated These respectively represent the maximum and minimum values ​​among the feature values ​​in the nth column. The nth feature value representing the mth sample.

[0035] In one exemplary embodiment of this disclosure, before converting the data feature vector used for QSVM model training into the quantum state corresponding to the data feature vector through a quantum circuit, the method further includes:

[0036] The data feature vectors are divided into training dataset, validation dataset, and test dataset according to a preset ratio. The training dataset is used to train the QSVM model, the validation dataset is used to validate the QSVM model, and the test dataset is used to test the training of the QSVM model.

[0037] In one exemplary embodiment of this disclosure, it further includes:

[0038] The feature vector of the data to be tested is input into the decision function of the constructed QSVM model for classification, so as to generate the classification result;

[0039] In response to determining the category of the data feature vector to be tested as a fault warning category based on the classification result, the contribution of each fault root cause is calculated based on the support vector and the quantum kernel function.

[0040] The root cause of the fault corresponding to the data feature vector to be tested is determined based on the ranking result of the contribution.

[0041] In one exemplary embodiment of this disclosure, the expression of the decision function includes The The feature vector representing the data to be tested, the Characterization Support vectors greater than 0, the sign Characterizing the symbolic function.

[0042] In one exemplary embodiment of this disclosure, the expression for calculating the contribution includes .

[0043] According to a second aspect of the present disclosure, a data classification method for a communication network is provided, comprising:

[0044] Collect communication metrics from the communication network;

[0045] The communication index is determined as a data feature, and the data feature is calculated using the data classification method described in any of the above technical solutions to determine the data classification result of the communication index.

[0046] According to a third aspect of the present disclosure, a data classification apparatus is provided, comprising:

[0047] The conversion module is configured to convert the data feature vectors used for QSVM model training into the corresponding quantum states of the data feature vectors through quantum circuits.

[0048] The module is configured to calculate the quantum kernel function based on the inner product of the quantum state and construct the corresponding dual optimization objective function;

[0049] The classification module is configured to determine the parameters of the optimal classification hyperplane in the QSVM model based on the optimal solution of the dual optimization objective function, so as to complete the construction of the QSVM model. The constructed QSVM model performs quantum state mapping on the feature vector of the data to be tested through a quantum processor, and calculates the classification decision boundary using a quantum kernel function to output the corresponding classification result.

[0050] According to a fourth aspect of the present disclosure, a data classification apparatus for a communication network is provided, comprising:

[0051] The acquisition module is configured to acquire communication metrics in the communication network.

[0052] The calculation module is configured to determine the communication indicator as a data feature, and calculate the data feature using the data classification method described in any of the above technical solutions, so as to determine the data classification result of the communication indicator.

[0053] According to a fifth aspect of this disclosure, an electronic device is provided, comprising: a memory; and a processor coupled to the memory, the processor being configured to perform the method as described in any one of the preceding methods based on instructions stored in the memory.

[0054] According to a sixth aspect of this disclosure, a computer-readable storage medium is provided having a program stored thereon that, when executed by a processor, implements the data classification method or the data classification method of a communication network as described in any of the preceding claims.

[0055] According to a seventh aspect of this disclosure, a computer program product is provided, comprising a computer program, characterized in that, when executed by a processor, the computer program implements the data classification method or the data classification method of a communication network as described in any of the preceding claims.

[0056] In this embodiment, a quantum circuit is used to transform the data feature vectors used for QSVM model training into corresponding quantum states. A quantum kernel function is calculated based on the inner product of these quantum states, and a corresponding dual optimization objective function is constructed. The parameters of the optimal classification hyperplane in the QSVM model are determined based on the optimal solution of the dual optimization objective function, thus completing the construction of the QSVM model. The constructed QSVM model uses a quantum processor to map the data feature vectors to be tested into quantum states and utilizes the quantum kernel function to calculate the classification decision boundary, outputting the corresponding classification result. This achieves the mapping of data feature vectors to high-dimensional quantum states, which is beneficial for more accurately capturing the nonlinear correlations between data feature vectors. It not only improves the computational efficiency of the QSVM model based on the computational characteristics of the quantum kernel function but also enhances the reliability and accuracy of the QSVM model in classifying data feature vectors.

[0057] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0058] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure. It is obvious that the drawings described below are merely some embodiments of this disclosure, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0059] Figure 1 A schematic diagram of an exemplary system architecture for a data classification scheme to which embodiments of the present invention can be applied is shown;

[0060] Figure 2 This is a flowchart of a data classification method in an exemplary embodiment of this disclosure;

[0061] Figure 3 This is a flowchart of another data classification method in an exemplary embodiment of this disclosure;

[0062] Figure 4 This is a block diagram of a data classification apparatus according to an exemplary embodiment of the present disclosure;

[0063] Figure 5 This is a block diagram of another data classification apparatus in an exemplary embodiment of this disclosure;

[0064] Figure 6 This is a block diagram of an electronic device according to an exemplary embodiment of the present disclosure. Detailed Implementation

[0065] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided to make this disclosure more comprehensive and complete, and to fully convey the concept of the example embodiments to those skilled in the art. The described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a full understanding of embodiments of this disclosure. However, those skilled in the art will recognize that the technical solutions of this disclosure can be practiced with one or more of the specific details omitted, or other methods, components, apparatus, steps, etc., can be employed. In other instances, well-known technical solutions are not shown or described in detail to avoid obscuring various aspects of this disclosure.

[0066] Furthermore, the accompanying drawings are merely illustrative of this disclosure, and the same reference numerals in the drawings denote the same or similar parts, thus repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0067] Figure 1 A schematic diagram of an exemplary system architecture for which an embodiment of the present invention can be applied is shown.

[0068] like Figure 1 As shown, system architecture 100 may include one or more of terminal devices 101, 102, and 103, a network 104, and a server 105. Network 104 serves as the medium for providing communication links between terminal devices 101, 102, and 103 and server 105. Network 104 may include various connection types, such as wired or wireless communication links or fiber optic cables, etc.

[0069] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, there can be any number of terminal devices, networks, and servers. For example, server 105 could be a server cluster composed of multiple servers.

[0070] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Terminal devices 101, 102, and 103 can be various electronic devices with displays, including but not limited to smartphones, tablets, laptops, and desktop computers, etc.

[0071] In some embodiments, the data classification method provided in this invention is generally executed by server 105, and correspondingly, the data classification device is generally located in terminal device 103 (or terminal device 101 or 102). In other embodiments, some terminals may have functions similar to those of the server device to execute this method.

[0072] The exemplary embodiments of this disclosure will now be described in detail with reference to the accompanying drawings.

[0073] Figure 2 This is a flowchart of a data classification method in an exemplary embodiment of this disclosure.

[0074] refer to Figure 2 Data classification methods can include:

[0075] Step S202: The data feature vectors used for QSVM model training are transformed into quantum states corresponding to the data feature vectors through quantum circuits;

[0076] Step S204: Calculate the quantum kernel function based on the inner product of the quantum state, and construct the corresponding dual optimization objective function;

[0077] Step S206: Based on the optimal solution of the dual optimization objective function, determine the parameters of the optimal classification hyperplane in the QSVM model to complete the construction of the QSVM model. The constructed QSVM model uses a quantum processor to perform quantum state mapping on the feature vector of the data to be tested, and uses a quantum kernel function to calculate the classification decision boundary to output the corresponding classification result.

[0078] This embodiment of the disclosure uses quantum circuits to transform the data feature vectors used for QSVM model training into corresponding quantum states. A quantum kernel function is calculated based on the inner product of these quantum states, and a corresponding dual optimization objective function is constructed. The parameters of the optimal classification hyperplane in the QSVM model are determined based on the optimal solution of the dual optimization objective function, thus completing the construction of the QSVM model. The constructed QSVM model uses a quantum processor to map the data feature vectors to be tested into quantum states and utilizes the quantum kernel function to calculate the classification decision boundary to output the corresponding classification result. This achieves the mapping of data feature vectors to high-dimensional quantum states, which is beneficial for more accurately capturing the nonlinear correlations between data feature vectors. It not only improves the computational efficiency of the QSVM model based on the computational characteristics of the quantum kernel function but also enhances the reliability and accuracy of the QSVM model in classifying data feature vectors.

[0079] The following section provides a detailed explanation of each step in the data classification method.

[0080] In one exemplary embodiment of this disclosure, calculating the quantum kernel function based on the inner product of the quantum states and constructing the corresponding dual optimization objective function includes:

[0081] The quantum kernel function is calculated based on the inner product of the quantum state, and the optimization objective function of the QSVM (Quantum Support Vector Machine) model is constructed.

[0082] The optimization objective function is transformed into a dual optimization objective function represented by a quantum kernel function by introducing Lagrange multipliers.

[0083] In the above embodiments, by introducing Lagrange multipliers, the optimization objective function is transformed into a dual optimization objective function represented by a quantum kernel function, thus eliminating the complex weights in the original optimization objective function. w and slack variables Eliminated, transformed into optimization only To further address this problem, a quantum kernel function is used to replace the traditional inner product, which not only preserves the consistency of the problem form but also achieves an efficient solution for handling high-dimensional feature maps.

[0084] In one exemplary embodiment of this disclosure, the expression of the optimization objective function includes N represents the number of data feature vectors, the Characterizing the regularization term, the The b represents the classification error term, and the b represents the classification bias. The slack variable represents the i-th data feature vector, and C represents the regularization parameter.

[0085] In one exemplary embodiment of this disclosure, the constraint condition corresponding to the optimization objective function is:

[0086] Among them, the The label representing the i-th data feature vector, the The quantum feature map characterizing the i-th data feature vector, the quantum feature map Through parameterized quantum circuits Implementation, wherein the quantum circuit Composed of a single-qubit rotation gate and a controlled entanglement gate, the aforementioned Encoded as quantum state .

[0087] In one exemplary embodiment of this disclosure, the expression of the dual optimization objective function includes , , wherein Representing the j-th Lagrange multiplier, the... Representing the i-th Lagrange multiplier, the The label representing the j-th data feature vector, the Characterization pairs and The quantum kernel function used for computation, the The quantum state characterizing the i-th data eigenvector, the The quantum state characterizing the j-th data eigenvector.

[0088] In one exemplary embodiment of this disclosure, the quantum kernel function is implemented by executing a quantum circuit. And obtained by measuring the ground state probability estimate, U( ) is a parameter xj The generated unitary matrix (quantum gate) acts on the quantum state, the Characterize the Perform Hermitian conjugate operation, i.e., with parameters xi The conjugate transpose (inverse operation) of the generated unitary matrix.

[0089] In one exemplary embodiment of this disclosure, the quantum kernel function is verified by sampling statistics, which include quantum state tomography and / or SWAP test estimation. The quantum kernel function is configured to satisfy the Mercer condition, and the Gram matrix of the quantum kernel function is corrected for semi-positive definiteness after measurement by quantum hardware.

[0090] In the above embodiments, by verifying the quantum kernel function through sampling statistics, the noise impact on the quantum computing device during operation in the current NISQ era is reduced.

[0091] In one exemplary embodiment of this disclosure, a quantum-classical hybrid algorithm is employed to solve the dual optimization objective function. This algorithm runs on a variable quantum eigenvalue solver, where quantum hardware is used to compute the kernel matrix terms of the dual optimization objective function. The classical apparatus in the variable quantum eigenvalue solver is used to calculate the gradient and update the Lagrange multipliers until the convergence condition is met.

[0092] In one exemplary embodiment of this disclosure, the constraints of the dual optimization objective function include: .

[0093] In one exemplary embodiment of this disclosure, determining the parameters of the optimal classification hyperplane in the QSVM model based on the optimal solution of the dual optimization objective function to complete the construction of the QSVM model includes:

[0094] The Lagrange multipliers and support vectors are determined based on the optimal solution of the dual optimization objective function.

[0095] The weight vector is calculated based on the Lagrange multipliers, the labels of the data feature vectors, and the quantum feature maps corresponding to the data feature vectors;

[0096] The classification bias is calculated based on the support vectors, the labels of the data feature vectors, and the quantum kernel function.

[0097] In one exemplary embodiment of this disclosure, the expression for calculating the weight vector based on the Lagrange multipliers, the labels of the data feature vectors, and the quantum feature maps corresponding to the data feature vectors includes: .

[0098] In one exemplary embodiment of this disclosure, the expression for calculating the classification bias based on the support vector, the labels of the data feature vectors, and the quantum kernel function includes: , wherein The support vector is represented by k, which represents the index of the support vector. S The set of indices that characterize the support vectors.

[0099] In one exemplary embodiment of this disclosure, the quantum circuit uses either angle encoding or amplitude encoding. The process of converting the data feature vectors used for QSVM model training into corresponding quantum states via the quantum circuit includes:

[0100] The data feature vector is mapped to a quantum state of the data feature vector using either angle encoding or amplitude encoding.

[0101] In one exemplary embodiment of this disclosure, the expression for mapping the data feature vector to the rotation angle of a qubit using an angle encoding method includes: Wherein, R represents a rotation gate about the coordinate axes of a Cartesian coordinate system, the coordinate axes being the x-axis, y-axis, or z-axis, and d represents the dimension of the data feature vector. The quantum state characterizing the data feature vector, the Characterizing the tensor product operator, the The standardized processing result characterizes the feature vector of the data.

[0102] In one exemplary embodiment of this disclosure, it further includes:

[0103] The constructed QSVM model is validated using the validation dataset corresponding to the data feature vectors.

[0104] Based on the validation results, the QSVM model is either tuned or tested on a test dataset.

[0105] The results include at least one of accuracy, precision, recall, and F1 score, and the parameter tuning process includes at least one of adjusting the feature mapping method, increasing the number of qubits, optimizing the parameters in the feature mapping circuit, and adding a regularization factor to the quantum kernel function.

[0106] In one exemplary embodiment of this disclosure, before converting the data feature vector used for QSVM model training into the quantum state corresponding to the data feature vector through a quantum circuit, the method further includes:

[0107] The collected data features are standardized, and the standardization process is either normalization or conversion to a normal distribution.

[0108] The standardized data features are vectorized to obtain the data feature vector.

[0109] In one exemplary embodiment of this disclosure, the expression for standardizing the collected data features includes: , wherein Characterizing the feature value of the nth column, the and stated These respectively represent the maximum and minimum values ​​among the feature values ​​in the nth column. The nth feature value representing the mth sample.

[0110] In one exemplary embodiment of this disclosure, before converting the data feature vector used for QSVM model training into the quantum state corresponding to the data feature vector through a quantum circuit, the method further includes:

[0111] The data feature vectors are divided into training dataset, validation dataset, and test dataset according to a preset ratio. The training dataset is used to train the QSVM model, the validation dataset is used to validate the QSVM model, and the test dataset is used to test the training of the QSVM model.

[0112] In one exemplary embodiment of this disclosure, it further includes:

[0113] The feature vector of the data to be tested is input into the decision function of the constructed QSVM model for classification, so as to generate the classification result;

[0114] In response to determining the category of the data feature vector to be tested as a fault warning category based on the classification result, the contribution of each fault root cause is calculated based on the support vector and the quantum kernel function.

[0115] The root cause of the fault corresponding to the data feature vector to be tested is determined based on the ranking result of the contribution.

[0116] In one exemplary embodiment of this disclosure, the expression of the decision function includes The The feature vector representing the data to be tested, the Characterization Support vectors greater than 0, the sign Characterizing the symbolic function.

[0117] In one exemplary embodiment of this disclosure, the expression for calculating the contribution includes .

[0118] Figure 3 This is a flowchart of a data classification method for a communication network in an exemplary embodiment of this disclosure.

[0119] refer to Figure 3 Data classification methods for communication networks may include:

[0120] Step S302: Collect communication indicators in the communication network;

[0121] Step S304: The communication index is determined as a data feature, and the data feature is calculated using the data classification method described in any of the above technical solutions to determine the data classification result of the communication index.

[0122] Corresponding to the above method embodiments, this disclosure also provides a data classification apparatus that can be used to execute the above method embodiments.

[0123] Figure 4 This is a block diagram of a data classification apparatus according to an exemplary embodiment of the present disclosure.

[0124] refer to Figure 4 The data classification device 400 may include:

[0125] The conversion module 402 is configured to convert the data feature vectors used for QSVM model training into quantum states corresponding to the data feature vectors through quantum circuits.

[0126] The construction module 404 is configured to calculate the quantum kernel function based on the inner product of the quantum state and construct the corresponding dual optimization objective function;

[0127] The classification module 406 is configured to determine the parameters of the optimal classification hyperplane in the QSVM model based on the optimal solution of the dual optimization objective function, so as to complete the construction of the QSVM model. The constructed QSVM model performs quantum state mapping on the feature vector of the data to be tested through a quantum processor, and calculates the classification decision boundary using a quantum kernel function to output the corresponding classification result.

[0128] In one exemplary embodiment of this disclosure, the construction module 404 is further configured as follows:

[0129] The quantum kernel function is calculated based on the inner product of the quantum state, and the optimization objective function of the QSVM model is constructed.

[0130] The optimization objective function is transformed into a dual optimization objective function represented by a quantum kernel function by introducing Lagrange multipliers.

[0131] In one exemplary embodiment of this disclosure, the expression of the optimization objective function includes N represents the number of data feature vectors, the Characterizing the regularization term, the The b represents the classification error term, and the b represents the classification bias. The slack variable represents the i-th data feature vector, and C represents the regularization parameter.

[0132] In one exemplary embodiment of this disclosure, the constraint condition corresponding to the optimization objective function is: .

[0133] Wherein, the label representing the i-th data feature vector, the The quantum feature map characterizing the i-th data feature vector, the quantum feature map Through parameterized quantum circuits Implementation, wherein the quantum circuit Composed of a single-qubit rotation gate and a controlled entanglement gate, the aforementioned Encoded as quantum state .

[0134] In one exemplary embodiment of this disclosure, the expression of the dual optimization objective function includes , , wherein Representing the j-th Lagrange multiplier, the... Representing the i-th Lagrange multiplier, the The label representing the j-th data feature vector, the Characterization pairs and The quantum kernel function used for computation, the The quantum state characterizing the i-th data eigenvector, the The quantum state characterizing the j-th data eigenvector.

[0135] In one exemplary embodiment of this disclosure, the quantum kernel function is implemented by executing a quantum circuit. And obtained by measuring the ground state probability estimate, the Characterize the Perform Hermitian conjugate operations.

[0136] In one exemplary embodiment of this disclosure, the quantum kernel function is verified by sampling statistics, which include quantum state tomography and / or SWAP test estimation. The quantum kernel function is configured to satisfy the Mercer condition, and the Gram matrix of the quantum kernel function is corrected for semi-positive definiteness after measurement by quantum hardware.

[0137] In one exemplary embodiment of this disclosure, a quantum-classical hybrid algorithm is used to solve the dual optimization objective function. This quantum-classical hybrid algorithm runs on a variable quantum eigenvalue solver, where quantum hardware is used for computation. The classical apparatus in the variable quantum eigenvalue solver is used to calculate the gradient and update the Lagrange multipliers until the convergence condition is met.

[0138] In one exemplary embodiment of this disclosure, the constraints of the dual optimization objective function include: .

[0139] In one exemplary embodiment of this disclosure, the classification module 406 is further configured as follows:

[0140] The Lagrange multipliers and support vectors are determined based on the optimal solution of the dual optimization objective function.

[0141] The weight vector is calculated based on the Lagrange multipliers, the labels of the data feature vectors, and the quantum feature maps corresponding to the data feature vectors;

[0142] The classification bias is calculated based on the support vectors, the labels of the data feature vectors, and the quantum kernel function.

[0143] In one exemplary embodiment of this disclosure, the expression for calculating the weight vector based on the Lagrange multipliers, the labels of the data feature vectors, and the quantum feature maps corresponding to the data feature vectors includes: .

[0144] In one exemplary embodiment of this disclosure, the expression for calculating the classification bias based on the support vector, the labels of the data feature vectors, and the quantum kernel function includes: Among them, the The support vector is represented by k, which represents the index of the support vector. S The set of indices that characterize the support vectors.

[0145] In one exemplary embodiment of this disclosure, the construction module 404 is configured as follows:

[0146] The data feature vector is mapped to a quantum state of the data feature vector using either angle encoding or amplitude encoding.

[0147] In one exemplary embodiment of this disclosure, the expression for mapping the data feature vector to the rotation angle of a qubit using an angle encoding method includes: Wherein, R represents a rotation gate about the coordinate axes of a Cartesian coordinate system, the coordinate axes being the x-axis, y-axis, or z-axis, and d represents the dimension of the data feature vector. The quantum state characterizing the data feature vector, the Characterizing the tensor product operator, the The standardized processing result characterizes the feature vector of the data.

[0148] In one exemplary embodiment of this disclosure, the data classification device 400 is further configured to:

[0149] The constructed QSVM model is validated using the validation dataset corresponding to the data feature vectors.

[0150] Based on the validation results, the QSVM model is either tuned or tested on a test dataset.

[0151] The results include at least one of accuracy, precision, recall, and F1 score, and the parameter tuning process includes at least one of adjusting the feature mapping method, increasing the number of qubits, optimizing the parameters in the feature mapping circuit, and adding a regularization factor to the quantum kernel function.

[0152] In one exemplary embodiment of this disclosure, the data classification device 400 is further configured to:

[0153] The collected data features are standardized, and the standardization process is either normalization or conversion to a normal distribution.

[0154] The standardized data features are vectorized to obtain the data feature vector.

[0155] In one exemplary embodiment of this disclosure, the expression for standardizing the collected data features includes: , wherein Characterizing the feature value of the nth column, the and stated These respectively represent the maximum and minimum values ​​among the feature values ​​in the nth column. The nth feature value representing the mth sample.

[0156] In one exemplary embodiment of this disclosure, the data classification device 400 is further configured to:

[0157] The data feature vectors are divided into training dataset, validation dataset, and test dataset according to a preset ratio. The training dataset is used to train the QSVM model, the validation dataset is used to validate the QSVM model, and the test dataset is used to test the training of the QSVM model.

[0158] In one exemplary embodiment of this disclosure, the data classification device 400 is further configured to:

[0159] The feature vector of the data to be tested is input into the decision function of the constructed QSVM model for classification, so as to generate the classification result;

[0160] In response to determining the category of the data feature vector to be tested as a fault warning category based on the classification result, the contribution of each fault root cause is calculated based on the support vector and the quantum kernel function.

[0161] The root cause of the fault corresponding to the data feature vector to be tested is determined based on the ranking result of the contribution.

[0162] In one exemplary embodiment of this disclosure, the expression of the decision function includes The The feature vector representing the data to be tested, the Characterization Support vectors greater than 0, the sign Characterizing the symbolic function.

[0163] In one exemplary embodiment of this disclosure, the expression for calculating the contribution includes .

[0164] Another data classification device in an exemplary embodiment of this disclosure includes a data acquisition module, a data standardization module, a data segmentation module, a quantum feature mapping module, a QSVM model training module, a QSVM model evaluation module, a fault prediction module, and a root cause localization module. The specific process is as follows:

[0165] Step 1: The data acquisition module collects historical indicator data and marks whether a fault has occurred. The dataset is defined as D={( xi,yi )|i=1,2,…,N}, where, xi = [xi1, xi2, ..., xid] ] is the feature vector of the i-th sample. yi It is the label of the i-th sample, where -1 indicates no fault and +1 indicates a fault.

[0166] Step 2: The data standardization module scales the feature values ​​to the [0, 1] interval or a standard normal distribution to improve the convergence speed of the model.

[0167] Step 3: The data segmentation module divides the data into training, validation, and test datasets according to a certain ratio. The training dataset is used to train the QSVM model, the validation dataset is used to evaluate the QSVM model, and the test dataset is used for fault prediction and root cause localization. The order in which the three datasets are used is training dataset, validation dataset, and test dataset.

[0168] Step 4: The quantum feature mapping module maps each sample xi to a quantum state. , ,in It is a quantum circuit that constructs quantum states using classical features. Commonly used feature mapping methods include angle encoding and amplitude encoding.

[0169] Step 5: The training objective of the QVSM model training module is to find the optimal classification hyperplane. First, define the optimization objective; the original optimization problem is: The constraints are: Next, it is transformed into a dual problem, and the dual optimization problem is solved using a quantum kernel function. Represented as: The constraints are: Next, we will use quantum computing resources to calculate the quantum kernel function: ,in and It is a quantum state. Finally, by optimizing the solution of N Lagrange multipliers... Determine the optimal classification hyperplane parameters and weights. Classification bias , where k is the index of the support vector.

[0170] Step 6: The QSVM model evaluation module uses a validation dataset to check the performance of the QSVM model, ensuring its generalization ability to unknown data. Evaluation typically includes accuracy, precision, recall, and F1 score. If these evaluation criteria are not met, the process returns to the quantum feature mapping module for parameter tuning and optimization. This includes adjusting the feature mapping method, increasing the number of qubits, optimizing parameters in the feature mapping circuit, and adding a regularization factor to the quantum kernel function to reduce overfitting. The parameter tuning and optimization process continues until the requirements of the QSVM model evaluation module are met, specifically including the following:

[0171] (1) The validation dataset is preprocessed by the data standardization module to ensure that the feature values ​​are scaled to the [0, 1] interval or the standard normal distribution. The feature vectors of the validation dataset are converted into quantum states by the quantum feature mapping module.

[0172] (2) Input the data after quantum feature mapping into the QSVM model for evaluation, calculate the support vector through the quantum kernel function, find the optimal classification hyperplane, optimize the Lagrange multiplier according to the dual optimization problem, determine the support vector through the Lagrange multiplier, and finally determine the classification bias.

[0173] (3) The QSVM model evaluation module calculates the model performance on the validation dataset and evaluates it using accuracy (the proportion of correctly classified samples), precision (the proportion of samples predicted as positive that are actually positive), recall (the proportion of samples actually positive that are predicted as positive), and F1 score (the harmonic mean of precision and recall, which takes into account both precision and recall).

[0174] If the target is not met, it is necessary to return to the quantum feature mapping module for parameter tuning and optimization. This includes modifying the mapping method (such as changing from angle encoding to amplitude encoding), increasing the number of qubits (the number of qubits used can be increased by multiples of 2 from a smaller number), optimizing the parameters in the quantum circuit (optimizing the control gate parameters in the quantum circuit, such as rotation angle and gate operation sequence to improve mapping efficiency), and adding a regularization factor to the quantum kernel function. Through appropriate regularization, the model's ability to generalize to new data can be improved, avoiding overfitting.

[0175] (4) After adjusting the parameters, repeat steps 5 and 6 until the requirements of step 6 are met.

[0176] Step 7: The fault prediction module uses the test dataset to perform fault prediction. For new data points... Classification is performed using a decision function: The judgment rule is: if f( If f > 0, it is predicted as a possible fault category (+1), and a fault warning is output. This is achieved by adjusting f( The normalization process is performed to convert f(f) into probability values, which are used to characterize the likelihood of a fault; if f(f) If the probability of failure is less than or equal to 0, the prediction is classified as low probability of failure (-1), the process ends, and the process restarts from step 1.

[0177] Step 8: If Step 7 predicts a possible fault category (+1), after outputting a fault warning, the root cause localization module uses support vectors... and quantum kernel function Analyze which features contribute most to the classification decision and calculate the contribution degree. The system outputs the root cause contribution rankings, identifies the feature with the highest contribution as the root cause of the failure, performs time embedding analysis on historical indicator data to locate the time node when the root cause of the failure occurred, ends the process, and restarts from step 1.

[0178] Figure 5 This is a block diagram of a data classification device for a communication network according to an exemplary embodiment of the present disclosure.

[0179] refer to Figure 5 The data classification device 500 of the communication network may include:

[0180] The acquisition module 502 is configured to acquire communication metrics in the communication network;

[0181] The calculation module 504 is configured to determine the communication index as a data feature, and calculate the data feature using the data classification method described in any of the above technical solutions, so as to determine the data classification result of the communication index.

[0182] The following section takes the data indicators of 5G wireless cells as an example. Based on the data feature classification scheme in the aforementioned communication network, the intelligent method of QSVM is used to achieve wireless cell fault prediction and root cause localization. By combining the high-dimensional feature mapping capabilities of quantum computing, the bottleneck problem of traditional methods in handling nonlinear complex relationships and high-dimensional data is solved. The specific steps are as follows:

[0183] Step 1: The data acquisition module collects operator network performance data for 5,000 wireless cells in a certain city from January to July 2024, including the following key performance indicators: cell ID, PRRUId, average uplink PRB occupancy rate, PDCCH channel occupancy rate, average number of RRC connected users, maximum number of RRC connected users, uplink RLC layer user plane traffic, downlink RLC layer user plane traffic, average actual transmit power of the cell, maximum actual transmit power of the cell, actual average transmit power of RRU / AAU, actual maximum transmit power of RRU / AAU, maximum transmit rate of NG interface user plane, etc., but not limited to these.

[0184] Historical data is recorded hourly, with each hour also labeled to indicate whether a fault occurred in the community during that time period, forming a dataset D={( xi,yi )|i=1,2,…,N}, where, xi= [ xi1, xi2, ..., xid ] is the feature vector of the i-th sample. yi It is the label of the i-th sample, where -1 indicates no fault and +1 indicates a fault;

[0185] Step 2: The data standardization module standardizes the features in the dataset, using the min-max standardization method to map each feature value to the [0, 1] interval to eliminate differences in the units of measurement between features and improve model training efficiency. The standardization formula is: , where max( ) and min ( ) represent features respectively The maximum and minimum values.

[0186] Step 3: The data splitting module divides the data into training, validation, and test datasets in a 7:2:1 ratio. The training dataset contains 35,000 samples for training the QSVM model; the validation dataset contains 10,000 samples for evaluating and optimizing the QSVM model; and the test dataset contains 5,000 samples for fault prediction and root cause localization. The order in which the three datasets are used is training, validation, and then test.

[0187] Step 4: The quantum feature mapping module maps the classical eigenvector xi to a quantum state. , ,in This is a quantum circuit. In this embodiment, an angle encoding method is used to map the feature values ​​to the rotation angle of the qubits. ,in, It is a rotation gate around the y-axis, where d is the dimension of the feature vector. High-dimensional feature mapping is achieved through quantum circuits.

[0188] Step 5: The training objective of the QVSM model training module is to find the optimal classification hyperplane, transforming the primal optimization problem into a dual problem using a quantum kernel function. Calculate the similarity between samples: Calculate Lagrange multipliers using the training dataset Determine the optimal classification hyperplane parameters and , where k is the index of the support vector.

[0189] Step 6: The QSVM model evaluation module uses the validation dataset to calculate metrics such as accuracy, precision, recall, and F1 score. If the above evaluation criteria are not met, the module returns to the quantum feature mapping module to begin parameter tuning and optimization. The model's F1 score on the validation dataset stabilizes at 97.8%, meeting the requirements of the evaluation module.

[0190] Step 7: Add new data points to the test dataset Input the trained QSVM model and use the decision function to calculate the classification result: The results show that f If the value is greater than 0, it is predicted to be a possible fault category. A fault warning is output, and the fault probability is obtained by normalization.

[0191] After the above scheme was applied to the test dataset, the accuracy of fault prediction reached 96.3% and the F1 score was 97.6%, which is higher than the traditional neural network model used for comparison (accuracy 93.1%, F1 score 91.2%).

[0192] Step 8: For samples predicted as fault categories, analyze the contribution of features to the decision using support vectors and quantum kernel functions. In the test dataset, PDCCH channel occupancy rate and maximum number of RRC connected users were repeatedly identified as the main factors causing cell overload.

[0193] By analyzing the time series of the above two indicators, the time point of the fault can be further located. In cell ID 692047, the PDCCH channel occupancy rate continued to exceed 96% at 16:00 on June X, 2024, and the maximum number of RRC connected users reached its peak, resulting in cell signaling overload.

[0194] Since the functions of the data classification device 400 and the data classification device 500 of the communication network have been described in detail in their respective method embodiments, they will not be repeated here.

[0195] The innovations of the embodiments disclosed herein compared to existing optimization methods include, but are not limited to:

[0196] 1. Designed for wireless cell fault prediction and root cause localization scenarios. QSVM is applied to wireless cell fault prediction and root cause localization, fully leveraging the high dynamism and spatiotemporal correlation of multidimensional data in wireless communication networks. This makes the trained QSVM model more customized and better suited to the actual needs of communication network operation and maintenance. Both fault prediction and root cause localization objectives can be achieved with just one QSVM model, making it more intelligent, efficient, and faster than classic prediction methods.

[0197] 2. More targeted data classification and optimized model training based on wireless data evaluation metrics. Addressing the high-dimensionality, sparsity, and multi-scale characteristics of wireless cell data, the data is standardized and divided into three categories for model training, validation, and testing, respectively. This improves the accuracy of data feature representation and the model's generalization ability. A model evaluation module is added, incorporating wireless data metrics to comprehensively assess model effectiveness. Furthermore, the model's expressive power is enhanced through optimized quantum mapping and QSVM model parameter tuning.

[0198] 3. Existing technical frameworks are mostly general fault prediction designs, which are difficult to adapt to the multidimensional temporal fault data characteristics of wireless cells. This invention innovatively proposes a quantum support vector machine model that combines spatiotemporal characteristics and communication network features. By utilizing the parallelism of quantum computing and its better similarity analysis capabilities under high-dimensional data, fault prediction and root cause localization can be achieved more efficiently, intelligently, and quickly.

[0199] 4. Existing technologies are limited to classification tasks. This invention innovatively extends classification to an integrated solution for prediction and root cause localization, realizing full-link innovation from data classification to scenario application, and enhancing the functional depth and value of the technical solution.

[0200] 5. It does not rely on powerful quantum computing resources; it can be achieved using NISQ-era quantum computing machines and quantum computing simulators.

[0201] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0202] In an exemplary embodiment of this disclosure, an electronic device capable of implementing the above-described method is also provided.

[0203] Those skilled in the art will understand that various aspects of the present invention can be implemented as systems, methods, or program products. Therefore, various aspects of the present invention can be specifically implemented in the following forms: entirely in hardware, entirely in software (including firmware, microcode, etc.), or in a combination of hardware and software, collectively referred to herein as “circuit,” “module,” or “system.”

[0204] The following reference Figure 6 To describe an electronic device 600 according to this embodiment of the present invention. Figure 6The electronic device 600 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0205] like Figure 6 As shown, the electronic device 600 is manifested in the form of a general-purpose computing device. The components of the electronic device 600 may include, but are not limited to: at least one processing unit 610, at least one storage unit 620, and a bus 630 connecting different system components (including storage unit 620 and processing unit 610).

[0206] The storage unit stores program code that can be executed by the processing unit 610, causing the processing unit 610 to perform the steps described in the "Exemplary Methods" section of this specification according to various exemplary embodiments of the present invention. For example, the processing unit 610 can perform the method shown in the embodiments of this disclosure.

[0207] Storage unit 620 may include readable media in the form of volatile storage units, such as random access memory (RAM) 6201 and / or cache memory 6202, and may further include read-only memory (ROM) 6203.

[0208] Storage unit 620 may also include a program / utility 6204 having a set (at least one) program module 6205, such program module 6205 including but not limited to: operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0209] Bus 630 can represent one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the various bus structures.

[0210] Electronic device 600 can also communicate with one or more external devices 640 (e.g., keyboard, pointing device, Bluetooth device, etc.), and with one or more devices that enable a user to interact with electronic device 600, and / or with any device that enables electronic device 600 to communicate with one or more other computing devices (e.g., router, modem, etc.). This communication can be performed via input / output (I / O) interface 650. Furthermore, electronic device 600 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 660. As shown, network adapter 660 communicates with other modules of electronic device 600 via bus 630. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0211] From the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, terminal device, or network device, etc.) to execute the methods according to the embodiments of this disclosure.

[0212] In exemplary embodiments of this disclosure, a computer-readable storage medium is also provided, on which a program product capable of implementing the methods described above is stored. In some possible embodiments, various aspects of the invention may also be implemented as a program product comprising program code that, when the program product is run on a terminal device, causes the terminal device to perform the steps of the various exemplary embodiments of the invention described in the "Exemplary Methods" section of this specification.

[0213] The program product for implementing the above-described method according to embodiments of the present invention may employ a portable compact disc read-only memory (CD-ROM) and include program code, and may run on a terminal device, such as a personal computer. However, the program product of the present invention is not limited thereto. In this document, the readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0214] The readable medium can be a readable signal medium or a readable storage medium. A readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0215] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting programs for use by or in conjunction with an instruction execution system, apparatus, or device.

[0216] The program code contained on the readable medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0217] In exemplary embodiments of this disclosure, a computer program product is also provided. This computer program product can be loaded or stored on any combination of one or more readable media. The program code for performing operations of the present invention can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on a user's computing device, partially on a user's computing device, as a standalone software package, partially on a user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0218] Furthermore, the above figures are merely illustrative of the processes included in the method according to exemplary embodiments of the present invention, and are not intended to be limiting. It is readily understood that the processes shown in the above figures do not indicate or limit the temporal order of these processes. Additionally, it is readily understood that these processes may be executed synchronously or asynchronously, for example, in multiple modules.

[0219] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and concept of this disclosure are indicated by the claims.

Claims

1. A data classification method, characterized in that, include: The data feature vectors used for QSVM model training are transformed into quantum states corresponding to the data feature vectors through quantum circuits. The quantum kernel function is calculated based on the inner product of the quantum states, and the corresponding dual optimization objective function is constructed, including: The quantum kernel function is calculated based on the inner product of the quantum state, and the optimization objective function of the QSVM model is constructed. The optimization objective function is transformed into the dual optimization objective function by introducing Lagrange multipliers. The parameters of the optimal classification hyperplane in the QSVM model are determined based on the optimal solution of the dual optimization objective function to complete the construction of the QSVM model, including: The Lagrange multipliers and support vectors are determined based on the optimal solution of the dual optimization objective function. The regularization term is calculated based on the Lagrange multipliers, the labels of the data feature vectors, and the quantum feature maps corresponding to the data feature vectors; The classification bias is calculated based on the support vectors, the labels of the data feature vectors, and the quantum kernel function; The expression for calculating the regularization term based on the Lagrange multipliers, the labels of the data feature vectors, and the quantum feature maps corresponding to the data feature vectors includes... The N represents the number of the data feature vectors, Representing the i-th Lagrange multiplier, the The label representing the i-th data feature vector, the The quantum feature map characterizing the i-th data feature vector, the quantum feature map Through parameterized quantum circuits Implementation, wherein the quantum circuit Composed of a single-qubit rotation gate and a controlled entanglement gate, the aforementioned Encoded as quantum state ; The expression for calculating the classification bias based on the support vectors, the labels of the data feature vectors, and the quantum kernel function includes... , , wherein x i Characterizing support vectors, the k The index that represents the support vector, the S The set of indices representing the support vectors, Characterization pairs and The quantum kernel function used for computation, which is executed by quantum circuits. And obtained by measuring the ground state probability estimate, the Characterize the Perform Hermitian conjugate operations; The constructed QSVM model uses a quantum processor to perform quantum state mapping on the feature vectors of the data to be tested, and uses a quantum kernel function to calculate the classification decision boundary in order to output the corresponding classification result; The feature vector of the data to be tested is input into the decision function of the constructed QSVM model for classification, so as to generate the classification result; In response to determining the category of the data feature vector to be tested as a fault warning category based on the classification result, the contribution of each fault root cause is calculated based on the support vector and the quantum kernel function. The root cause of the fault corresponding to the data feature vector to be tested is determined based on the ranking result of the contribution. The expression of the decision function includes: sign ( The The feature vector representing the data to be tested, the Characterization Support vectors greater than 0, the sign The characterization function, and the expression for calculating the contribution, include... .

2. The data classification method as described in claim 1, characterized in that, The expression of the optimization objective function includes N represents the number of data feature vectors, the Characterizing the regularization term, the Characterizing the classification error term, the Characterizing classification bias, the The slack variable representing the i-th data feature vector is... Characterize the regularization parameter.

3. The data classification method as described in claim 2, characterized in that, The constraints corresponding to the optimization objective function are: , , ., Among them, the The label representing the i-th data feature vector, the The quantum feature map that represents the i-th data feature vector.

4. The data classification method as described in claim 3, characterized in that, The expression of the dual optimization objective function includes , , Among them, the Representing the j-th Lagrange multiplier, the... Representing the i-th Lagrange multiplier, the The label representing the j-th data feature vector, the Characterization pairs and The quantum kernel function used for computation, wherein the quantum kernel function is used to calculate the squared modulus of the inner product of two quantum states, the The quantum state characterizing the i-th data eigenvector, the The quantum state characterizing the j-th data eigenvector.

5. The data classification method as described in claim 4, characterized in that, The quantum kernel function is verified by sampling statistics, which include quantum state tomography and / or SWAP test estimation. The quantum kernel function is configured to satisfy the Mercer condition, and the Gram matrix of the quantum kernel function is corrected for semi-positive definiteness after measurement by quantum hardware.

6. The data classification method as described in claim 4, characterized in that, The dual optimization objective function is solved using a quantum-classical hybrid algorithm, which runs on a variable quantum eigenvalue solver. The quantum hardware in the variable quantum eigenvalue solver is used to compute the kernel matrix terms of the dual optimization objective function, while the classical hardware in the variable quantum eigenvalue solver is used to compute the gradient and update the Lagrange multipliers until the convergence condition is met.

7. The data classification method as described in claim 4, characterized in that, The constraints of the dual optimization objective function include: .

8. The data classification method as described in claim 1, characterized in that, The quantum circuit uses either angle encoding or amplitude encoding. The process of converting the data feature vectors used for QSVM model training into corresponding quantum states via the quantum circuit includes: The data feature vector is mapped to a quantum state of the data feature vector using the angle encoding method or the amplitude encoding method.

9. The data classification method as described in claim 8, characterized in that, The expression for mapping the data feature vector to the rotation angle of a qubit using angle encoding includes: , wherein A rotation gate representing a coordinate axis in a Cartesian coordinate system, wherein the coordinate axis is the x-axis, y-axis, or z-axis, and d represents the dimension of the data feature vector. The quantum state characterizing the data feature vector, the Characterizing the tensor product operator, the The standardized processing result characterizes the feature vector of the data.

10. The data classification method according to any one of claims 1-9, characterized in that, Also includes: The constructed QSVM model is validated using the validation dataset corresponding to the data feature vectors. Based on the validation results, the QSVM model is either tuned or tested on a test dataset. The results include at least one of accuracy, precision, recall, and F1 score, and the parameter tuning process includes at least one of adjusting the feature mapping method, increasing the number of qubits, optimizing the parameters in the feature mapping circuit, and adding a regularization factor to the quantum kernel function.

11. The data classification method according to any one of claims 1-9, characterized in that, Before converting the data feature vectors used for QSVM model training into corresponding quantum states via quantum circuits, the process also includes: The collected data features are standardized, and the standardization process is either normalization or conversion to a normal distribution. The standardized data features are vectorized to obtain the data feature vector.

12. The data classification method as described in claim 11, characterized in that, The expression for standardizing the features of the collected data includes: , wherein Characterizing the feature value of the nth column, the and stated These respectively represent the maximum and minimum values ​​among the feature values ​​in the nth column. The nth feature value representing the mth sample.

13. The data classification method according to any one of claims 1-9, characterized in that, Before converting the data feature vectors used for QSVM model training into corresponding quantum states via quantum circuits, the process also includes: The data feature vectors are divided into training dataset, validation dataset, and test dataset according to a preset ratio. The training dataset is used to train the QSVM model, the validation dataset is used to validate the QSVM model, and the test dataset is used to test the training of the QSVM model.

14. A data classification method for a communication network, characterized in that, include: Collect communication metrics from the communication network; The communication index is identified as a data feature, and the data feature is calculated using the data classification method as described in any one of claims 1-13 to determine the data classification result of the communication index.

15. A data classification device, characterized in that, include: The conversion module is configured to convert the data feature vectors used for QSVM model training into the corresponding quantum states of the data feature vectors through quantum circuits. The construction module is configured to calculate the quantum kernel function based on the inner product of the quantum state and construct the corresponding dual optimization objective function, including: The quantum kernel function is calculated based on the inner product of the quantum state, and the optimization objective function of the QSVM model is constructed. The optimization objective function is transformed into the dual optimization objective function by introducing Lagrange multipliers. The classification module is configured to determine the parameters of the optimal classification hyperplane in the QSVM model based on the optimal solution of the dual optimization objective function, thereby completing the construction of the QSVM model, including: The Lagrange multipliers and support vectors are determined based on the optimal solution of the dual optimization objective function. The regularization term is calculated based on the Lagrange multipliers, the labels of the data feature vectors, and the quantum feature maps corresponding to the data feature vectors; The classification bias is calculated based on the support vectors, the labels of the data feature vectors, and the quantum kernel function; The expression for calculating the regularization term based on the Lagrange multipliers, the labels of the data feature vectors, and the quantum feature maps corresponding to the data feature vectors includes... N represents the number of data feature vectors, the Representing the i-th Lagrange multiplier, the The label representing the i-th data feature vector, the The quantum feature map characterizing the i-th data feature vector, the quantum feature map Through parameterized quantum circuits Implementation, wherein the quantum circuit Composed of a single-qubit rotation gate and a controlled entanglement gate, the aforementioned Encoded as quantum state ; The expression for calculating the classification bias based on the support vectors, the labels of the data feature vectors, and the quantum kernel function includes... , , wherein Characterizing support vectors, the The index that represents the support vector, the S The set of indices representing the support vectors, Characterization pairs and The quantum kernel function used for computation, which is executed by quantum circuits. And obtained by measuring the ground state probability estimate, the Characterize the Perform Hermitian conjugate operations; The constructed QSVM model uses a quantum processor to perform quantum state mapping on the feature vectors of the data to be tested, and uses a quantum kernel function to calculate the classification decision boundary in order to output the corresponding classification result; The feature vector of the data to be tested is input into the decision function of the constructed QSVM model for classification, so as to generate the classification result; In response to determining the category of the data feature vector to be tested as a fault warning category based on the classification result, the contribution of each fault root cause is calculated based on the support vector and the quantum kernel function. The root cause of the fault corresponding to the data feature vector to be tested is determined based on the ranking result of the contribution. The expression of the decision function includes: sign ( The The feature vector representing the data to be tested, the Characterization Support vectors greater than 0, the sign The characterization function, and the expression for calculating the contribution, include... .

16. A data classification device for a communication network, characterized in that, include: The acquisition module is configured to acquire communication metrics in the communication network. The calculation module is configured to determine the communication index as a data feature, and calculate the data feature using the data classification method as described in any one of claims 1-13, so as to determine the data classification result of the communication index.

17. An electronic device, characterized in that, include: Memory; as well as A processor coupled to the memory, the processor being configured to execute the data classification method as described in any one of claims 1-13 or the data classification method for a communication network as described in claim 14, based on instructions stored in the memory.

18. A computer-readable storage medium having a program stored thereon that, when executed by a processor, implements the data classification method as claimed in any one of claims 1-13 or the data classification method for a communication network as claimed in claim 14.

19. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the data classification method as described in any one of claims 1-13 or the data classification method for a communication network as described in claim 14.

Images