Risk account determination method and apparatus, device, storage medium and program product

Abstract

The present application discloses a risk account determination method and apparatus, a device, a storage medium and a program product. The risk account determination method comprises: constructing a transfer network graph on the basis of transfer information among a plurality of accounts, wherein the plurality of accounts comprise at least one known risk account; converting the transfer network graph into a corresponding bipartite graph, and constructing a quantum walk space on the basis of the bipartite graph; using a plurality of quantum bits to perform quantum state preparation on each node among a plurality of nodes, so as to obtain an initial quantum state corresponding to each node; on the basis of the initial quantum state, taking each node as a starting point respectively to perform a quantum walk in the quantum walk space, so as to obtain a final quantum state corresponding to each node; on the basis of the final quantum state, determining embedded information corresponding to each node; and on the basis of the embedded information, determining, from among the plurality of accounts, an account similar to the known risk account as a risk account.

Description

Risk account identification methods, devices, equipment, storage media and program products

[0001] Cross-reference of related applications

[0002] This application claims priority to Chinese Patent Application No. 202410676804.7, filed on May 28, 2024, entitled “Risk Account Determination Method, Apparatus, Device, Storage Medium and Program Product”, the entire contents of which are incorporated herein by reference. Technical Field

[0003] This application pertains to financial risk control technology, and in particular relates to a method, apparatus, equipment, storage medium, and program product for determining risk accounts. Background Technology

[0004] With increasingly active financial activities, irregular fund transfers between accounts are also becoming more rampant. Financial institutions are the primary channel for these account transfers, making it essential to establish an efficient and accurate risk control system within them.

[0005] Currently, the traditional method for identifying risky accounts with non-compliant fund transfer behavior mainly relies on traditional graph neural network algorithms to analyze the characteristics of account transfer behavior in order to determine whether an account is a risky account.

[0006] Thus, existing graph neural network algorithms are computationally inefficient and have low accuracy in identifying risky accounts because they require a lot of computation. Summary of the Invention

[0007] This application provides a method, apparatus, device, storage medium, and program product for determining risk accounts, which can improve computational efficiency and increase the accuracy of risk account identification.

[0008] In a first aspect, embodiments of this application provide a method for determining risk accounts, the method comprising:

[0009] Based on the transfer information between multiple accounts, a transfer network graph is constructed with accounts as nodes and the transfer relationships between accounts as connecting edges, wherein at least one of the multiple accounts is a known risky account;

[0010] The transfer network graph is transformed into a corresponding bipartite graph, and a quantum walk space is constructed based on the bipartite graph;

[0011] Quantum states are prepared for each node corresponding to each of the multiple accounts using multiple qubits, to obtain the initial quantum state corresponding to each node;

[0012] Based on the initial quantum state corresponding to each node, a quantum walk is performed in the quantum walk space starting from each node to obtain the final quantum state corresponding to each node;

[0013] Based on the final quantum state corresponding to each node, determine the embedded information corresponding to each node;

[0014] Based on the embedding information corresponding to each node, accounts similar to the known risky accounts are identified from the multiple accounts as risky accounts.

[0015] Secondly, embodiments of this application provide a risk account determination device, the device comprising:

[0016] The network construction module is used to construct a transfer network graph with accounts as nodes and transfer relationships between accounts as connecting edges based on transfer information between multiple accounts, wherein at least one of the multiple accounts is a known risk account;

[0017] A spatial construction module is used to transform the transfer network graph into a corresponding bipartite graph, and construct a quantum walk space based on the bipartite graph;

[0018] The quantum state preparation module is used to prepare the quantum state of the node corresponding to each account in the multiple accounts using multiple qubits, so as to obtain the initial quantum state corresponding to each node;

[0019] The quantum walk module is used to perform quantum walks in the quantum walk space, starting from each node, based on the initial quantum state corresponding to each node, to obtain the final quantum state corresponding to each node;

[0020] The information determination module is used to determine the embedded information corresponding to each node based on the final quantum state corresponding to each node;

[0021] The account determination module is used to determine, based on the embedded information corresponding to each node, accounts similar to the known risky accounts from the multiple accounts as risky accounts.

[0022] Thirdly, embodiments of this application provide an electronic device, which includes: a processor and a memory storing computer program instructions;

[0023] When the processor executes the computer program instructions, it implements the steps of the risk account determination method as described in any embodiment of the first aspect.

[0024] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer program instructions that, when executed by a processor, implement the steps of the risk account determination method as described in any embodiment of the first aspect.

[0025] Fifthly, embodiments of this application provide a computer program product in which instructions, when executed by a processor of an electronic device, cause the electronic device to perform the steps of the risk account determination method as described in any embodiment of the first aspect.

[0026] The risk account determination method, apparatus, device, storage medium, and program product in this application embodiment construct a transfer network graph with accounts as nodes and transfer relationships between accounts as connecting edges. This graph is then transformed into a corresponding bipartite graph, and a quantum walk space is constructed based on this bipartite graph. Multiple qubits are used to prepare quantum states for each node corresponding to each of the multiple accounts. Based on the initial quantum states corresponding to each node, a quantum walk is performed in the quantum walk space starting from each node to obtain the final quantum state corresponding to each node. Then, based on the final quantum state corresponding to each node, the embedding information corresponding to each node is determined. Using this embedding information, accounts similar to known risk accounts are identified as risk accounts from among the multiple accounts. Thus, because this application embodiment utilizes the advantages of quantum computing, performing quantum walks in the constructed quantum walk space starting from each node, compared to the random walk method in traditional graph neural network algorithms, quantum computing has lower complexity and higher computational efficiency, thereby effectively improving computational efficiency. Furthermore, since the walk in this embodiment is performed on a bipartite graph of the transfer network graph, the state space of the walk is expanded, providing more degrees of freedom and possibilities. This helps the quantum walk to explore the topology of the transfer network graph more comprehensively, thus improving the accuracy of risk account identification. Attached Figure Description

[0027] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 is a flowchart illustrating a risk account determination method provided in one embodiment of this application;

[0029] Figure 2 is a schematic diagram of a transfer network diagram provided in this application;

[0030] Figure 3 is a schematic diagram of a bipartite graph provided in this application;

[0031] Figure 4 is a schematic diagram of the structure of a risk account determination device provided in an embodiment of this application;

[0032] Figure 5 is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0033] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0034] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0035] The acquisition, storage, use, and processing of data (including but not limited to features and information mentioned in this document) in the technical solution of this application all comply with the relevant provisions of national laws and regulations.

[0036] When analyzing the characteristics of account transfer behavior, one can observe from the perspective of graph theory and network topology. Non-compliant fund transfer behavior will exhibit specific abnormal structural features in the financial network, such as ring, community, butterfly and dumbbell structures. These abnormal structural features can be used to identify risky accounts with non-compliant transfer behavior.

[0037] Random walk algorithms are currently the mainstream method for graph topology feature extraction. A classic random walk starts from a node and randomly selects connecting edges to move through. After numerous steps, the resulting path reflects the overall structure and characteristics of the transfer network graph. This classic random walk algorithm is computationally inefficient due to the large amount of computation required, and it also has a low accuracy rate in identifying risky accounts.

[0038] To address the problems of existing technologies, this application provides a method, apparatus, device, storage medium, and program product for identifying risky accounts. This risky account identification method is based on quantum theory and employs a quantum walk approach to mine the characteristics of a transfer network. Quantum walk utilizes the properties of quantum superposition to simultaneously explore all possible paths, traversing them with a certain probability. This allows for faster and more efficient network mining than traditional random walk algorithms, improving computational efficiency. During the quantum walk, the negative amplitude causes the positive and negative amplitudes of particles to cancel each other out on certain paths, producing a unique interference effect. This effect makes quantum walk more inclined to locate anomalous regions that are closely connected (such as communities composed of accounts with non-compliant fund transfer behavior).

[0039] Furthermore, quantum walks can capture local details while preserving the global structural information of a graph. Classical random walks are "short-sighted" Markov processes, with each step relying only on the local information of the current node. Quantum walks, through quantum superposition states, can establish long-range connections between different nodes, reflecting the overall properties of the graph. By analyzing the states of quantum walks, node embedding information containing graph structural information can be obtained. This quantum embedding can simultaneously consider both local and global characteristics of the graph, thereby improving the accuracy of risk account identification.

[0040] The risk account determination method provided in the embodiments of this application will be introduced first below.

[0041] Figure 1 is a flowchart illustrating a risk account determination method according to an embodiment of this application. This risk account determination method can be applied to scenarios involving the identification of risk accounts with non-compliant fund transfer activities, and can be executed by electronic devices with computing capabilities, such as computers or servers.

[0042] As shown in Figure 1, the method for determining risky accounts may specifically include the following steps:

[0043] S110. Based on the transfer information between multiple accounts, construct a transfer network graph with accounts as nodes and transfer relationships between accounts as connecting edges, including at least one known risk account among the multiple accounts;

[0044] S120. Transform the transfer network graph into a corresponding bipartite graph, and construct a quantum walk space based on the bipartite graph;

[0045] S130. Using multiple qubits, prepare the quantum state of the node corresponding to each account in multiple accounts to obtain the initial quantum state corresponding to each node;

[0046] S140. Based on the initial quantum state corresponding to each node, perform a quantum walk in the quantum walk space starting from each node to obtain the final quantum state corresponding to each node.

[0047] S150. Based on the final quantum state corresponding to each node, determine the embedded information corresponding to each node;

[0048] S160. Based on the embedded information corresponding to each node, identify accounts that are similar to known risky accounts from multiple accounts as risky accounts.

[0049] Thus, by leveraging the advantages of quantum computing, this embodiment performs quantum walks in a constructed quantum walk space, starting from each node. Compared to the random walk method in traditional graph neural network algorithms, quantum computing has lower complexity and higher computational efficiency. Furthermore, quantum walks, through quantum superposition states, can establish long-range associations between different nodes, reflecting the overall properties of the transfer network graph and making the identification results more accurate. Therefore, it can effectively improve computational efficiency and the accuracy of identifying risky accounts. In addition, since this embodiment performs walks on a bipartite graph of the transfer network graph, it expands the state space of the walk, providing more degrees of freedom and possibilities. This helps quantum walks to more comprehensively explore the topological structure of the transfer network graph, thus further improving the accuracy of identifying risky accounts.

[0050] The specific implementation methods for each of the above steps are described below.

[0051] In some implementations, in S110, the multiple accounts can be accounts that have made transfers within a preset time period. These multiple accounts may include at least one known risk account, which can be a known account flagged for non-compliant fund transfer behavior. The transfer information includes the transferring account information and the received account information.

[0052] A transfer network graph can be a topological graph constructed based on transfer information, with accounts as nodes and transfer relationships between accounts as connecting edges. When a transfer relationship exists between any two accounts, a connecting edge can be established between the nodes corresponding to those two accounts.

[0053] For example, a transfer network graph G can be constructed as shown in the following formula (1): G=(V,E) (1)

[0054] Where V = {V1,...,V} n Let $\mathbf{A}$ be the node set, $E$ be the edge set, and $N$ be the total number of nodes in graph G. Each node corresponding to an account can be its account number, such as a bank card number.

[0055] In some implementations, in S120, the original graph of the transfer network graph can be transformed into an equivalent bipartite graph, and quantum walks can be performed on the edges of the bipartite graph. Here, the way to transform the transfer network graph into a bipartite graph is to copy all nodes in the transfer network graph into two disjoint node sets. When there is a connecting edge between two nodes in the original graph of the transfer network graph, there is a connecting edge between the corresponding nodes in the two node sets of the bipartite graph.

[0056] In some implementations, the process of converting the transfer network graph into a corresponding bipartite graph in S120 above may specifically include:

[0057] The nodes in the transfer network graph are replicated to obtain two node sets, each of which includes nodes corresponding to multiple accounts.

[0058] Based on the two node sets and the connecting edges between nodes in the transfer network graph, a bipartite graph corresponding to the transfer network graph is generated. The connection relationship between any two nodes in the node sets of the bipartite graph is determined by the connecting edges between nodes in the transfer network graph.

[0059] Here, all nodes in the transfer network graph can be copied into node sets X and Y. When the transfer network graph contains N nodes, node sets X and Y also contain N nodes respectively.

[0060] For example, as shown in Figure 2, if the transfer network graph contains 4 nodes and there are connecting edges between the nodes corresponding to accounts with transfer relationships, then the transfer network graph can be transformed into a bipartite graph as shown in Figure 3, which contains a node set X and a node set Y. Two nodes that have connecting edges in the transfer network graph also have connecting edges between their corresponding nodes in node sets X and Y.

[0061] After converting the transfer network graph into a corresponding bipartite graph, a quantum walk space can be constructed based on the bipartite graph.

[0062] In this embodiment of the application, for any node v among N nodes i The operational basis can be represented in Dirac notation as |i>. |i> is an N×1 dimensional column vector, where the elements in the i-th row are 1s and the other elements are 0s. |i,j> represents the tensor product of column vectors |i> and |j>. Therefore, |i,j> is an N×N matrix, where the element in the i-th row and j-th column is 1, and the other elements are 0.

[0063] For example, the Hilbert spaces of node set X and node set Y are defined as follows: and in, The Hilbert space for quantum walks on a bipartite graph is then: The calculated ground state is Where {|i,j>:v i ∈X,v j ∈Y}. That is, a quantum walk space can be constructed based on the bipartite graph.

[0064] In this way, by walking along the connecting edges of the bipartite graph corresponding to the transfer network graph, instead of walking along the connecting edges of the transfer network graph, walking on the bipartite graph can conveniently utilize some properties of classical Markov chains. Furthermore, walking on the bipartite graph expands the state space of the walk, providing more degrees of freedom and possibilities. The connecting edges encode the multi-step connection information between nodes in the transfer network graph, which helps quantum walking to explore the topological structure of the graph more comprehensively.

[0065] In some implementations, in S130, n qubits can be used to represent the state of each node in the transfer network graph.

[0066] In quantum computing, since n qubits can represent 2^n qubits... n There are two states, and these states can be represented by 2 n The graph is represented by a 1×1 column vector. Therefore, if the total number of nodes in the transfer network graph is N, then a total of n≥log2N qubits are needed to represent the states of all nodes |i> and |j>.

[0067] In this embodiment of the application, an initial quantum state can be prepared for each node based on a Markov chain before the node performs a quantum walk.

[0068] In some implementations, the above-mentioned S130 may specifically include:

[0069] Generate an adjacency matrix based on the transfer network graph;

[0070] The probability transition matrix is ​​obtained by determining the transition probabilities between nodes in the transfer network graph based on the adjacency matrix.

[0071] Based on the probability transition matrix, the projection state of the node corresponding to each account in the Markov chain is determined, where a node is represented by multiple qubits;

[0072] The projected state corresponding to each node is taken as the initial quantum state.

[0073] For example, for the transfer network graph G shown in formula (1) above, the adjacency matrix A shown in formula (2) can be generated:

[0074] Where, the element in the i-th row and j-th element of A is A i，j A i,j =1 represents node vi and v j There is a transfer relationship between them.

[0075] Based on this, the probability transition matrix P of the transfer network graph G can be generated according to the following formula (3):

[0076] Among them, P i，j Represents node v i Transfer to node v j The probability, d j For node v j The node degree. The node degree refers to the number of edges associated with a node in the graph, which can be defined by the following formula (4):

[0077] The node degree d j That is, the sum of all elements in the j-th column of the adjacency matrix A.

[0078] Classical Markov chains consist of random variables Composition, each random variable has N possible states (X) t ∈{s1,...,s N The probability of each random variable depends only on the previous random variable, i.e.: P(X). t |X t-1 ,X t-2 ,...,X1)=P(X t |X t-1 If the Markov chain is time-independent, it satisfies the expression shown in formula (5) below:

[0079] The Markov transition process can be described by an N×N probability transition matrix P, where P i,j =P(X) t =s i |X t-1 =s j ) represents the state transition s j →s i The probability of and satisfying the column normalization constraint. The random walk process on the graph can be represented by a Markov chain, where the states of the Markov chain correspond to the nodes on the graph.

[0080] Based on this, for two adjacent nodes v in the transfer network graph G i and v j ,|i,j> indicates that the wanderer is currently located at node v i The next roaming target is v j |j,i> indicates that the wanderer is currently located at node v. j The next roaming target is vi During the random walk, starting from node v i Reaching node v j All possible paths have the same probability, therefore for node v i v i All possible destinations v j The states are superimposed to obtain the expression shown in formula (6) below:

[0081] Where, |ρ i >This can be viewed as the square root of the i-th column of the probability transition matrix P. Then, the node v can be transformed using the following formula (7). i The state |i> and the corresponding |ρ i By performing a tensor product operation, the node v can be obtained. i Projected state |ψ on a Markov chain i >:

[0082] The projection state |ψ i >Can be used as node v i The initial quantum state encoding, also known as the initial quantum state. Where 0 represents node v i A quantum walk has not yet been performed. The state after the quantum walk. It represents the superposition state of the nodes in the diagram. The probability of each node being observed after measurement reflects the tendency of quantum wandering near that node.

[0083] In some implementations, in S140, the walker can be controlled to traverse the quantum walking space constructed based on the bipartite graph, starting from each node. A t-step quantum walk is performed, and the final quantum state can be obtained for each node. Here, the walker can be controlled in the quantum walk space based on a preset diffusion operator. Random quantum walks are performed.

[0084] In some implementations, prior to S140 above, the method provided in this application embodiment may further include:

[0085] Obtain the target diffusion operator;

[0086] Based on the target diffusion operator and the preset exchange operator, determine the quantum walk expression corresponding to a single-step quantum walk.

[0087] Based on this, the aforementioned S140 may specifically include:

[0088] Based on the initial quantum state corresponding to each node, starting from each node, a quantum walk is performed in the quantum walk space according to the quantum walk expression to obtain the final quantum state corresponding to each node.

[0089] Here, the target diffusion operator R can be an operator used to control the walker's movement within the quantum walk space. In some examples, the target diffusion operator R can include the Grover operator. The Grover operator is a core component of the Grover search algorithm; it achieves the "rotation" of the quantum state through a reflection operation, enabling the algorithm to find the target node at a quadratic acceleration. Therefore, the Grover operator is also known as the reflection operator.

[0090] Thus, by using the Grover operator as the target diffusion operator, the algorithm achieves an exponential speedup. The time complexity of quantum walk based on the Grover operator is... Traditional random walks are O(N).

[0091] Of course, the target diffusion operator can also be other operators besides the Grover operator, which will not be elaborated here.

[0092] In addition, the preset swap operator can be an operator used to swap two quantum registers, such as the swap operator S shown in the following formula (8):

[0093] By using the commutation operator S in quantum walks, the walker can turn around during the process, increasing the flexibility and exploratory capabilities of the walk and enabling quantum interference between walk paths in different directions. Furthermore, by adjusting the parameters of the commutation operator, the strength and mode of interference can be controlled, thereby generating quantum behaviors such as delocalization and improving the ability to uncover potential risk accounts.

[0094] Based on this, the quantum walk expression corresponding to a single-step quantum walk, that is, the expression for a one-step quantum walk, can be determined according to the following formula (9): U QW =SR (9)

[0095] Among them, U QW Let S be the quantum walk expression corresponding to a single-step quantum walk, where S is the preset exchange operator and R is the target diffusion operator. Based on this, the state obtained by |i,j> after one quantum walk is as shown in the following formula (10):

[0096] According to the above formula (10), after one quantum walk, the walker moves from v i We arrived at v j The next step for the wanderers is from vj Departure, there The probability returns to v i ,have The probability of reaching another node v i Other than, and v j Adjacent node v k Among them, the node degree d j Represents and node v i The number of connected edges.

[0097] For example, with node v i Starting from, Performing t quantum walks It is possible to obtain the relationship with node v i The corresponding final quantum state For example, the initial quantum state can be determined according to the following formula (11). Perform the first quantum walk to obtain the quantum state.

[0098] By following this logic, the final quantum state can be obtained. Based on this, after each node is used as the starting point for a quantum walk and each node performs t quantum walks, the final quantum states corresponding to N nodes can be obtained.

[0099] Furthermore, the aforementioned target diffusion operator can also be determined based on a preset projection operator. In some embodiments, the aforementioned target diffusion operator may specifically include:

[0100] A target diffusion operator is constructed based on a preset projection operator, wherein the preset projection operator is used to project vectors in the quantum walk space onto the target Hilbert space.

[0101] Here, the preset projection operator corresponds to the target Hilbert space.

[0102] In quantum mechanics, a set of orthogonal and uniform vectors {|ψ i The sequence {i = 1, ..., N} spans an N-dimensional Hilbert space Ψ. Based on this, a predefined projection operator can be defined as the projection operator Ώ, through which any quantum walk space can be projected. The vector in the vector is projected onto the space Ψ, which is the target Hilbert space, as shown in the following formula (12):

[0103] Based on this, the target diffusion operator R can be constructed using the projection operator ∏ according to the following formula (13): R=2П-I (13)

[0104] Where I is the identity matrix.

[0105] In some implementations, in S150, the embedding information can be an embedding vector that can represent the characteristics of a node. After obtaining the final quantum state corresponding to each node, each final quantum state can be measured to obtain the probability amplitude on each node, and then the probability amplitude of each node can be used as the embedding information corresponding to that node, which can be an embedding vector.

[0106] In some implementations, S150 may specifically include: for each node corresponding to each of the plurality of accounts, performing the following steps to obtain embedding information corresponding to each node. This step may specifically include:

[0107] Apply the Hadamard gate to the multiple qubits corresponding to the target node to convert the final quantum state of the target node into a measurement basis. The target node is the node corresponding to any one of the multiple accounts.

[0108] Based on the measurement basis, multiple qubits corresponding to the target node are measured to obtain the measurement results;

[0109] The embedding information corresponding to the target node is determined based on the measurement results.

[0110] For example, with the target node as node v i For example, we can first prepare an n-bit quantum register and initialize the final quantum state. Apply a Hadamard gate H to each of the n qubits to determine the final quantum state. The measurement basis is then converted. When applying the Hadamard mandrel, the measurement basis can be obtained by processing according to the following formula (14).

[0111] Where x is the binary representation of the measurement result, x i Let α be the value of the i-th position. i For node v i The corresponding probability amplitude.

[0112] Based on this measurement basis, each of the n qubits can be measured separately to obtain the measurement result. Using this measurement result, the embedding information corresponding to the target node can be determined.

[0113] To improve the accuracy of measurement results, multiple measurements can be performed on the qubit, and the probability of a certain result can be statistically analyzed. This probability can then be used as the node's embedding information. Based on this, in some implementations, determining the embedding information corresponding to the target node based on the measurement results can specifically include:

[0114] Return to the execution process and apply the Hadamard gate to multiple qubits corresponding to the target node, converting the final quantum state of the target node into a measurement basis, until the preset number of cycles is obtained, resulting in multiple measurement results;

[0115] The probability of the target quantum state appearing in the statistical measurement results;

[0116] The embedding information corresponding to the target node is determined based on probability.

[0117] Here, the preset number of cycles can be set as needed. After repeating the measurement multiple times, the probability that the measurement result at each node is the target quantum state can be calculated, and this probability can then be used as an estimate of the node's probability amplitude.

[0118] For example, by measuring the quantum state of n qubits, the probability P(x) of the measurement result being x can be obtained as shown in the following formula (15):

[0119] Where n is the number of qubits.

[0120] Taking the target quantum state as |1> as an example, the node v can be statistically analyzed. i The probability that the state of each dimension in the corresponding measurement results (N dimensions in total) is |1> is used as the node v. i The probability magnitude estimate can then be obtained for node v. i The embedding vector.

[0121] In some implementations, in S160, after obtaining the embedding information corresponding to each node, the embedding information of each node can be used to calculate the similarity between accounts with unknown risks and accounts with known risks among multiple accounts. Then, if the similarity of an account with unknown risks meets a preset condition, it is identified as a risk account. The preset condition may include that the similarity with a known risk account is greater than a preset similarity threshold, or that the account belongs to the top preset number of accounts with relatively high similarity to known risk accounts.

[0122] In some implementations, S160 may specifically include:

[0123] Based on the embedding information corresponding to each node, calculate the Euclidean distance between the node corresponding to the target account and the node corresponding to the known risk account among multiple accounts. The target account is the other accounts among the multiple accounts besides the known risk account.

[0124] Risky accounts are identified from multiple accounts based on Euclidean distance.

[0125] Here, the similarity between embedding vectors can be represented using Euclidean distance.

[0126] For example, if there are K nodes corresponding to known risky accounts among N nodes, and the embedding vectors of these K nodes are {μ1, μ2, ..., μ...} k Then, the Euclidean distance between the other nodes in the N nodes (excluding the K nodes) and the K nodes can be calculated according to the following formula (16):

[0127] Among them, D ij Let represent the Euclidean distance between the embedding vector corresponding to the i-th node and the embedding vector corresponding to the j-th known risk account node, where d represents the total number of dimensions of the embedding vectors, and here d = N, and l represents the l-th dimension of the embedding vectors.

[0128] In some examples, after obtaining the Euclidean distance, nodes whose Euclidean distance is greater than a preset distance threshold can be identified as risk nodes.

[0129] To obtain stable output results, in other examples, the above method determines risky accounts from multiple accounts based on Euclidean distance, which may specifically include:

[0130] The target accounts corresponding to the preset number of nodes with the smallest Euclidean distance are identified as risk accounts.

[0131] For example, the Euclidean distance D can be selected. ij The most recent Q nodes are output as risk nodes, and so on, for a total of KQ nodes are output, and the accounts corresponding to these KQ nodes are identified as risk accounts.

[0132] In summary, the quantum walk algorithm employed in this application exhibits a faster convergence speed than general quantum walk algorithms. The state transitions of the quantum walk in this application are described by block matrices, and these transition matrices are unitary. These properties allow walkers to propagate and mix within the network at a faster rate. Fast convergence means the algorithm can converge to a stationary distribution with fewer iterations, discovering key patterns and anomalous behaviors in the network, thus resulting in higher computational efficiency. This advantage is particularly important when dealing with large-scale complex networks.

[0133] Furthermore, the quantum walk state in this application is characterized by more parameters, and the Hilbert space in which the quantum state resides has a higher dimension, thus possessing a more powerful expressive ability. This high-dimensional state space can accommodate richer and more diverse network information and walk dynamics. Through the precise design and control of the quantum state, the quantum walk in this application can achieve more complex walk behaviors and patterns, potentially revealing hidden risk account characteristics that are difficult to characterize with conventional quantum walks.

[0134] It should be noted that the application scenarios described in the above embodiments of this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0135] Based on the same inventive concept, this application also provides a risk account determination device. This will be described in detail with reference to Figure 4.

[0136] Figure 4 is a schematic diagram of the structure of a risk account determination device provided in an embodiment of this application.

[0137] As shown in Figure 4, the risk account determination device 400 may include:

[0138] The network construction module 401 is used to construct a transfer network graph with accounts as nodes and transfer relationships between accounts as connecting edges based on the transfer information between multiple accounts, wherein the multiple accounts include at least one known risk account;

[0139] The space construction module 402 is used to convert the transfer network graph into a corresponding bipartite graph and construct a quantum walk space based on the bipartite graph;

[0140] The quantum state preparation module 403 is used to prepare the quantum state of the node corresponding to each account in the multiple accounts using multiple qubits, so as to obtain the initial quantum state corresponding to each node;

[0141] The quantum walk module 404 is used to perform quantum walks in the quantum walk space, starting from each node, based on the initial quantum state corresponding to each node, to obtain the final quantum state corresponding to each node;

[0142] The information determination module 405 is used to determine the embedded information corresponding to each node based on the final quantum state corresponding to each node;

[0143] The account determination module 406 is used to determine, based on the embedded information corresponding to each node, an account similar to the known risky account from the plurality of accounts as a risky account.

[0144] The risk account determination device 400 described above is explained in detail below:

[0145] In some embodiments, the space construction module 402 includes:

[0146] The node replication submodule is used to replicate the nodes in the transfer network graph to obtain two node sets, each of which includes nodes corresponding to the multiple accounts respectively.

[0147] The bipartite graph generation submodule is used to generate a bipartite graph corresponding to the transfer network graph based on the two node sets and the connecting edges between nodes in the transfer network graph. The connection relationship between any two nodes in the node sets in the bipartite graph is determined based on the connecting edges between nodes in the transfer network graph.

[0148] In some embodiments, the risk account determination device 400 further includes:

[0149] The operator acquisition module is used to acquire the target diffusion operator before performing quantum walks in the quantum walk space starting from each node according to the initial quantum state corresponding to each node, and obtaining the final quantum state corresponding to each node.

[0150] The expression determination module is used to determine the quantum walk expression corresponding to a single-step quantum walk based on the target diffusion operator and the preset exchange operator;

[0151] The quantum walking module 404 is specifically used for:

[0152] Based on the initial quantum state corresponding to each node, starting from each node, a quantum walk is performed in the quantum walk space according to the quantum walk expression to obtain the final quantum state corresponding to each node.

[0153] In some embodiments, the operator acquisition module is specifically used for:

[0154] A target diffusion operator is constructed based on a preset projection operator, wherein the preset projection operator is used to project vectors in the quantum walk space onto the target Hilbert space.

[0155] In some embodiments, the target diffusion operator includes the Grover operator.

[0156] In some embodiments, the quantum state preparation module 403 includes:

[0157] A matrix generation submodule is used to generate an adjacency matrix based on the transfer network graph;

[0158] The probability determination submodule is used to determine the transition probability between nodes in the transfer network graph based on the adjacency matrix, and obtain the probability transition matrix;

[0159] The projection state determination submodule is used to determine the projection state of the node corresponding to each of the multiple accounts in the Markov chain based on the probability transition matrix, wherein a node is represented by multiple qubits;

[0160] The initial state determination submodule is used to take the projected state corresponding to each node as the initial quantum state.

[0161] In some embodiments, the information determination module 405 includes:

[0162] The basis conversion submodule is used to apply a Hadamard gate to multiple qubits corresponding to the target node, and convert the final quantum state corresponding to the target node into a measurement basis. The target node is the node corresponding to any one of the multiple accounts.

[0163] The quantum measurement submodule is used to measure multiple qubits corresponding to the target node based on the measurement substrate, and obtain the measurement results;

[0164] The information determination submodule is used to determine the embedded information corresponding to the target node based on the measurement results.

[0165] In some embodiments, the information determination submodule includes:

[0166] The loop execution unit is used to return to the execution of applying the Hadamard gate to multiple qubits corresponding to the target node, converting the final quantum state corresponding to the target node into a measurement basis, until a preset number of loops is obtained, and multiple measurement results are obtained.

[0167] A probability statistics unit is used to calculate the probability of the target quantum state appearing in the measurement results;

[0168] An information determination unit is used to determine the embedded information corresponding to the target node based on the probability.

[0169] In some embodiments, the account determination module 406 includes:

[0170] The distance calculation submodule is used to calculate the Euclidean distance between the node corresponding to the target account and the node corresponding to the known risk account among the plurality of accounts based on the embedding information corresponding to each node, wherein the target account is the other account among the plurality of accounts besides the known risk account;

[0171] The account determination submodule is used to determine risky accounts from the plurality of accounts based on the Euclidean distance.

[0172] In some embodiments, the account determination submodule is specifically used for:

[0173] The target accounts corresponding to the preset number of nodes with the smallest Euclidean distance are identified as risk accounts.

[0174] Therefore, by constructing a transfer network graph with accounts as nodes and transfer relationships between accounts as connecting edges, and then transforming the transfer network graph into a corresponding bipartite graph, a quantum walk space is constructed based on this bipartite graph. Furthermore, multiple qubits are used to prepare quantum states for each node corresponding to each account in multiple accounts. Based on the prepared initial quantum states corresponding to each node, a quantum walk is performed in the quantum walk space starting from each node, thereby obtaining the final quantum state corresponding to each node. Then, based on the final quantum state corresponding to each node, the embedding information corresponding to each node is determined. Using the embedding information corresponding to each node, accounts similar to known risky accounts are identified from multiple accounts as risky accounts. Thus, because this embodiment utilizes the advantages of quantum computing, performing quantum walks in the constructed quantum walk space starting from each node, compared to the random walk method in traditional graph neural network algorithms, quantum computing has lower complexity and higher computational efficiency, thus effectively improving computational efficiency. Furthermore, since the walk in this embodiment is performed on a bipartite graph of the transfer network graph, the state space of the walk is expanded, providing more degrees of freedom and possibilities. This helps the quantum walk to explore the topology of the transfer network graph more comprehensively, thus improving the accuracy of risk account identification.

[0175] Figure 5 is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0176] Electronic device 500 may include processor 501 and memory 502 storing computer program instructions.

[0177] Specifically, the processor 501 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0178] Memory 502 may include mass storage for data or instructions. For example, and not limitingly, memory 502 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 502 may include removable or non-removable (or fixed) media. Where appropriate, memory 502 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 502 is non-volatile solid-state memory.

[0179] In certain embodiments, the memory may include read-only memory (ROM), random access memory (RAM), disk storage media devices, optical storage media devices, flash memory devices, and electrical, optical, or other physical / tangible memory storage devices. Thus, typically, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to one aspect of this application.

[0180] The processor 501 implements any of the risk account determination methods in the above embodiments by reading and executing computer program instructions stored in the memory 502.

[0181] In some examples, the electronic device 500 may also include a communication interface 503 and a bus 510. As shown in Figure 5, the processor 501, memory 502, and communication interface 503 are connected via the bus 510 and communicate with each other.

[0182] The communication interface 503 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0183] Bus 510 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not as a limitation, bus 510 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 510 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, any suitable bus or interconnect is contemplated herein.

[0184] For example, the electronic device 500 can be a server, computer, mobile phone, tablet computer, laptop computer, handheld computer, in-vehicle electronic device, ultra-mobile personal computer (UMPC), netbook, or personal digital assistant (PDA), etc.

[0185] The electronic device 500 can execute the risk account determination method in the embodiments of this application, thereby realizing the risk account determination method and apparatus described in conjunction with Figures 1 and 4.

[0186] Furthermore, in conjunction with the risk account determination methods in the above embodiments, this application embodiment can provide a computer-readable storage medium for implementation. This computer-readable storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the risk account determination methods in the above embodiments. Examples of computer-readable storage media include non-transitory computer-readable storage media, such as portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, etc.

[0187] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0188] The functional blocks shown in the above-described structural diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0189] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0190] The aspects of this application have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by dedicated hardware performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0191] The above description is merely a specific implementation of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A method for determining risky accounts, comprising: Based on the transfer information between multiple accounts, a transfer network graph is constructed with accounts as nodes and the transfer relationships between accounts as connecting edges, wherein at least one of the multiple accounts is a known risky account; The transfer network graph is transformed into a corresponding bipartite graph, and a quantum walk space is constructed based on the bipartite graph; Quantum states are prepared for each node corresponding to each of the multiple accounts using multiple qubits, to obtain the initial quantum state corresponding to each node; Based on the initial quantum state corresponding to each node, a quantum walk is performed in the quantum walk space starting from each node to obtain the final quantum state corresponding to each node; Based on the final quantum state corresponding to each node, determine the embedded information corresponding to each node; Based on the embedding information corresponding to each node, accounts similar to the known risky accounts are identified from the multiple accounts as risky accounts.

2. The method according to claim 1, wherein, The step of converting the transfer network graph into a corresponding bipartite graph includes: The nodes in the transfer network graph are copied to obtain two node sets, each of which includes nodes corresponding to the multiple accounts respectively; Based on the two node sets and the connecting edges between nodes in the transfer network graph, a bipartite graph corresponding to the transfer network graph is generated. The connection relationship between any two nodes in the node sets in the bipartite graph is determined based on the connecting edges between nodes in the transfer network graph.

3. The method according to claim 1, wherein, Before performing quantum walks in the quantum walk space starting from each node based on the initial quantum state corresponding to each node to obtain the final quantum state corresponding to each node, the method further includes: Obtain the target diffusion operator; Based on the target diffusion operator and the preset exchange operator, determine the quantum walk expression corresponding to a single-step quantum walk; The process of performing quantum walks in the quantum walk space, starting from each node and based on the initial quantum state corresponding to each node, to obtain the final quantum state corresponding to each node includes: Based on the initial quantum state corresponding to each node, starting from each node, a quantum walk is performed in the quantum walk space according to the quantum walk expression to obtain the final quantum state corresponding to each node.

4. The method according to claim 3, wherein, The acquisition of the target diffusion operator includes: A target diffusion operator is constructed based on a preset projection operator, wherein the preset projection operator is used to project vectors in the quantum walk space onto the target Hilbert space.

5. The method according to claim 3 or 4, wherein, The target diffusion operator includes the Grover operator.

6. The method according to claim 1, wherein, The step of preparing a quantum state for each node corresponding to each of the multiple accounts using multiple qubits to obtain an initial quantum state corresponding to each node includes: Generate an adjacency matrix based on the aforementioned transfer network graph; The transition probabilities between nodes in the transfer network graph are determined based on the adjacency matrix, resulting in a probability transition matrix; Based on the probability transition matrix, the projection state of the node corresponding to each of the multiple accounts in the Markov chain is determined, wherein a node is represented by multiple qubits; The projected state corresponding to each node is taken as the initial quantum state.

7. The method according to claim 1, wherein, The process of determining the embedding information corresponding to each node based on the final quantum state of each node includes: For each node corresponding to one of the multiple accounts, the following steps are performed to obtain the embedding information corresponding to each node: Apply a Hadamard gate to multiple qubits corresponding to the target node to convert the final quantum state corresponding to the target node into a measurement basis. The target node is the node corresponding to any one of the multiple accounts. Based on the measurement substrate, multiple qubits corresponding to the target node are measured to obtain the measurement results; The embedding information corresponding to the target node is determined based on the measurement results.

8. The method according to claim 7, wherein, Determining the embedding information corresponding to the target node based on the measurement results includes: Return to the execution of applying the Hadamard gate to the multiple qubits corresponding to the target node, converting the final quantum state corresponding to the target node into a measurement basis, until a preset number of cycles is obtained, and multiple measurement results are obtained; The probability of the target quantum state appearing in the measurement results is statistically analyzed; The embedding information corresponding to the target node is determined based on the probability.

9. The method according to claim 1, wherein, The step of determining accounts similar to the known risky accounts from the plurality of accounts based on the embedding information corresponding to each node includes: Based on the embedding information corresponding to each node, calculate the Euclidean distance between the node corresponding to the target account and the node corresponding to the known risk account among the plurality of accounts, wherein the target account is the other account among the plurality of accounts besides the known risk account; Risky accounts are identified from the plurality of accounts based on the Euclidean distance.

10. The method according to claim 9, wherein, The step of determining risky accounts from the plurality of accounts based on the Euclidean distance includes: The target accounts corresponding to the preset number of nodes with the smallest Euclidean distance are identified as risk accounts.

11. A risk account determination device, comprising: The network construction module is used to construct a transfer network graph with accounts as nodes and transfer relationships between accounts as connecting edges based on transfer information between multiple accounts, wherein at least one of the multiple accounts is a known risk account; A spatial construction module is used to transform the transfer network graph into a corresponding bipartite graph, and construct a quantum walk space based on the bipartite graph; The quantum state preparation module is used to prepare the quantum state of the node corresponding to each account in the multiple accounts using multiple qubits, so as to obtain the initial quantum state corresponding to each node; The quantum walk module is used to perform quantum walks in the quantum walk space, starting from each node, based on the initial quantum state corresponding to each node, to obtain the final quantum state corresponding to each node; The information determination module is used to determine the embedded information corresponding to each node based on the final quantum state corresponding to each node; The account determination module is used to determine, based on the embedded information corresponding to each node, accounts similar to the known risky accounts from the multiple accounts as risky accounts.

12. An electronic device, the electronic device comprising: Processor and memory storing computer program instructions; When the processor executes the computer program instructions, it implements the steps of the risk account determination method as described in any one of claims 1-10.

13. A computer-readable storage medium storing computer program instructions that, when executed by a processor, implement the steps of the risk account determination method as described in any one of claims 1-10.

14. A computer program product, wherein instructions in the computer program product, when executed by a processor of an electronic device, cause the electronic device to perform the steps of the risk account determination method as described in any one of claims 1-10.

Images