A suspicious organization query method for public security heterogeneous information networks

Abstract

The present invention relates to a suspicious organization query method for a public security heterogeneous information network. The present invention first obtains a personnel relationship dataset in the public security field, constructs a heterogeneous information network in the public security field based on the personnel relationship dataset, then obtains clue suspicious persons, traverses the heterogeneous information network with a meta-path based on the clue suspicious persons, obtains a set of candidate suspicious persons related to the clue suspicious persons, then obtains a set of close suspicious persons based on a k-core constraint, obtains all possible k-core subgraphs based on the close suspicious person set, and finally calculates the attribute distance of each k-core subgraph, ultimately obtaining suspicious organizations with similar attributes and a close structure. In the process of suspicious organization query, the present invention combines heterogeneous network structure information and vertex attribute information, and simultaneously considers the textual attributes and numerical attributes of the suspicious persons to find suspicious organizations with a close structure and similar attributes that contain the clue suspicious persons.

Description

Technical Field

[0001] The present invention belongs to the field of public security, and in particular relates to a suspicious organization query method for a public security heterogeneous information network. Background Art

[0002] In social networks, suspicious organizations inevitably manifest as communities with a specific structure. With the advancement of informatization, more and more data such as personal relationship networks and personal information has been accumulated.

[0003] Existing suspicious organization search methods mostly focus on homogeneous information networks, failing to consider heterogeneous information networks, which are more common in real-world applications. These networks contain nodes and edges of varying types, making them incapable of direct deployment in heterogeneous networks. For example, in real-world scenarios, node types often extend beyond just objects, encompassing events, cases, addresses, and organizations. Edge types encompass relationships between objects and events, between objects and cases, and between objects. Homogeneous information networks simply cannot display this rich information. Furthermore, existing suspicious organization search algorithms often consider only a single attribute type, such as simple text attributes or numerical attributes, failing to consider both types simultaneously, impacting query quality and accuracy. Summary of the Invention

[0004] Based on the above shortcomings and deficiencies in the existing technology, the present invention introduces the concept of meta-path in the process of suspicious organization query, which can directly query suspicious organizations in heterogeneous information networks. In addition, the present invention considers both textual attributes and numerical attributes of suspicious persons, and can find suspicious organizations with close structures and similar attributes that contain clues to suspicious persons.

[0005] In order to achieve the above-mentioned object of the invention, the present invention adopts the following technical solutions:

[0006] A suspicious organization query method for a public security heterogeneous information network includes the following steps:

[0007] S1, obtain the personnel relationship dataset in the public security field,

[0008] A heterogeneous information network in the public security field is constructed based on the personnel relationship dataset, wherein the heterogeneous information network includes vertices, edges connecting vertices, and vertex attributes; the vertices include vertices representing objects, cases, and events; the edges are used to represent the relationship between the vertices at both ends of the edge; and each vertex includes text attributes and numerical attributes;

[0009] S2, obtaining a clue of a suspicious person, and traversing the heterogeneous information network with a meta-path based on the clue of the suspicious person to obtain a set of candidate suspicious persons, where the meta-path refers to a specific path connecting two types of vertices;

[0010] S3, generating an induced subgraph based on the candidate suspicious person set, and removing the candidate suspicious persons that do not meet the k-core constraint from the induced subgraph according to the k-core constraint, to obtain a k-core subgraph containing the suspicious person, wherein the vertex set of the k-core subgraph is a close suspicious person set containing the clue suspicious person;

[0011] S4, constructing a search tree based on the set of closely related suspicious persons, wherein the search tree records all possible k-core subgraphs containing the clue suspicious persons;

[0012] S5, traverse the search tree, calculate the attribute distance of each k-core subgraph, and select the subgraph with the smallest attribute distance as the result to be returned. The set of all vertices in the returned subgraph is the suspicious organization containing the suspicious person.

[0013] As a preferred solution, step S2 specifically includes the following steps:

[0014] S2-1 The obtained clue suspicious person is set as q, the meta path is set as p, and the queue to be processed Q is initialized. c For the clue suspicious person {q}, initialize the candidate suspicious person set C as the clue suspicious person {q}, the waiting queue Q c Suspicious persons to be traversed are stored in it;

[0015] S2-2 For any suspicious person v∈Q c , find all suspicious persons connected to the meta-path p starting from the suspicious person v, and save them into the set Ψ[v]. For any suspicious person u in Ψ[v], if the suspicious person u is not included in the candidate suspicious person set C, then the suspicious person u is added to the candidate suspicious person set C and the waiting queue Q at the same time. c Repeat the above steps until the queue to be processed Q c is empty;

[0016] S2-3 All the suspicious persons in the candidate suspicious person set C constitute the candidate suspicious person set related to the clue suspicious person q.

[0017] As a preferred solution, step S4 specifically includes the following steps:

[0018] S4-1 initializes a search tree T based on the set of closely related suspicious persons. Each tree node in the search tree T records a k-core subgraph and its child nodes. The root node root of the initialized search tree records the induced subgraph H generated by the set of candidate suspicious persons in step 3. Initializes a queue to be constructed Q as the root node {root} of the search tree. The queue to be constructed Q stores the search tree nodes to be constructed.

[0019] S4-2 For any tree node n∈Q, take out the subgraph H in the tree node n. The set of suspicious persons in the subgraph H is recorded as S. For any suspicious person s in the suspicious person set S, try to remove s from S, and then find the suspicious person set that still meets the k-core constraint from the induced subgraph composed of the remaining suspicious persons. if If it is not empty, a new tree node n1 is created and recorded as a child node of tree node n. Tree node n1 records the set of suspicious persons. The induced subgraph constructed adds the tree node n1 to the queue to be constructed Q;

[0020] S4-3 repeats step S4-2 until the queue Q to be constructed is empty, completing the construction of the search tree.

[0021] As a preferred solution, in step S5, the calculation formula of the attribute distance of the k-core subgraph is: v1, v2 are any two points in the subgraph H, N is the number of object type vertices in the subgraph H; where f(v1, v2) = α*f t (v1,v2)+(1-α)*f n (v1,v2); where f t (v1,v2) represents the text attribute distance between suspicious persons v1 and v2, f n (v1, v2) represents the numerical attribute distance between suspicious persons v1 and v2, and α is the weight ratio of the text attribute distance and the numerical attribute example.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] The present invention combines heterogeneous network structure information and suspicious person attribute information during the suspicious organization query process, and considers the text attributes and numerical attributes of the suspicious persons at the same time to find suspicious organizations with a close structure and similar attributes that contain clue suspicious persons. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a flow chart of a suspicious organization query method for a public security heterogeneous information network according to an embodiment of the present invention;

[0025] Figure 2 is a schematic diagram of a public security heterogeneous information network according to an embodiment of the present invention;

[0026] Figure 3 It is a homogeneous graph composed of APA meta-paths in an embodiment of the present invention. DETAILED DESCRIPTION

[0027] To more clearly illustrate the embodiments of the present invention, specific embodiments of the present invention will be described below with reference to the accompanying drawings. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings and other embodiments can be obtained based on these drawings without inventive efforts.

[0028] This embodiment provides a suspicious organization query method for a public security heterogeneous information network, the flow chart of which is as follows: Figure 1 As shown, this embodiment takes part of the personnel relationship dataset in the public security field as an example.

[0029] (1) First, perform step S1 to obtain a personnel relationship dataset in the public security field.

[0030] A heterogeneous information network in the public security field is constructed based on the personnel relationship data set, and the heterogeneous information network includes vertices, edges connecting vertices, and vertex attributes; the vertices include three types of vertices: objects, cases, and events; the edges are used to represent the relationship between the vertices at both ends of the edge; each of the vertices includes text attributes and numerical attributes; in this embodiment, the vertices include two types of vertices: object nodes representing objects (A) and event nodes (P) representing events. The edges represent that the object is related to a certain event, and the object type vertices have text attributes (occupational category, marital status, criminal record, etc.) and numerical attributes (control level, age, etc.).

[0031] (2) Obtain clues of suspicious persons. Based on the clues of suspicious persons, traverse the heterogeneous information network with a meta-path to obtain a set of candidate suspicious persons, wherein the meta-path refers to a specific path connecting two types of vertices. The meta-path can be 'object to event to object' or 'object to event to case to event to object'. Taking 'object to case to object' as an example, the gang we get is a set of objects under the same event. Taking 'object to event to case to event to object' as an example, we get a set of objects under all events under the same case. Path characteristics: The initial point type and the end point type of the meta-path must be objects, and the meta-path is symmetrical. The meta-path can be object to case to object or object to event to object. Different types of intermediate vertices are different meta-paths. The process is explained as follows with examples:

[0032] (2-1) The obtained clue suspicious person is set as a2, and the meta path is set as Initialize the pending queue Q c For the clue suspect object {a2}, initialize the candidate suspect object set C as the clue suspect object {a2};

[0033] (2-2) Q from the pending queue c Take out a suspect object from the set. At this time, there is only suspect object a2, so take out suspect object a2, find all suspect objects connected to suspect object a2 through the meta-path APA, and record them in Ψ[a2] = {a1, a3, a6}. Suspect objects a1, a3, a6 in Ψ[a2] = {a1, a3, a6} are not in the candidate suspect object set C, so a1, a3, a6 are added to the candidate suspect object set C and the waiting queue Q at the same time. c Repeat the above steps until the queue Q c is empty;

[0034] (2-3) All suspect objects C = {a1, a2, a3, a4, a5, a6, a7, a8} in the candidate suspect object set C constitute the candidate suspect object set related to the clue suspect object a2.

[0035] (3) Then proceed to step S3 to generate an induced subgraph based on the candidate suspect object set, such as Figure 3 As shown in the figure, according to the k-core constraint, taking k=3 as an example, the suspect a5 that does not meet the conditions is removed from the induced subgraph to obtain the k-core subgraph containing the suspect. At this time, the vertex set {a1, a2, a3, a4, a6, a7, a8} of the k-core subgraph is the close suspect set containing the clue suspect.

[0036] (4) Then proceed to step S4, constructing a search tree based on the closely related set of suspicious persons. This process is described below with an example:

[0037] (4-1) Initialize a search tree T based on the set of close candidate suspect objects, where each tree node in the search tree T records a k-core subgraph and the child nodes of the node. Initialize the root node root of the search tree to record the induced subgraph H generated by the set of close candidate suspect objects {a1, a2, a3, a4, a6, a7, a8} in step 3, and initialize the queue to be constructed Q as the root node {root} of the search tree;

[0038] (4-2) Take out a tree node root from the queue to be constructed Q. At this time, there is only the tree node root, so take out the tree node root, take out the suspect object set S = {a1, a2, a3, a4, a6, a7, a8} in the subgraph H in the tree node root, take out the suspect object a4 from the set S, consider deleting the suspect object a4 from the subgraph H, and then find the suspect object set that still meets the 3-core from the remaining suspect objects Create a new tree node n1 and record tree node n1 as the child node of tree node root, add tree node n1 to the queue to be built Q, continue to take out other suspect objects in set S, and repeat the above operation until every suspect object in set S has been selected.

[0039] (4-3) Repeat the steps in (4-2) until the queue Q to be constructed is empty, and the construction of the search tree is completed.

[0040] (5) Finally, step S5 is performed to traverse the search tree, calculate the attribute distance of each k-core subgraph, and select the subgraph with the smallest attribute distance as the result. The set of all vertices in the returned subgraph is the suspicious organization containing the suspicious person. The attribute distance calculation formula for the subgraph is: v1, v2 are any two points in the subgraph H, and N is the number of object type vertices in the subgraph H. Where f(v1, v2) = α*f t (v1,v2)+(1-α)* f n (v1,v2). Where f t (v1, v2) represents the text attribute distance between suspect objects v1 and v2, f n (v1, v2) represents the numerical attribute distance of suspect objects v1 and v2, and α is the weight ratio of the text attribute distance and the numerical attribute example. When α is 0.5, the effect is best, that is, the text attribute and the numerical attribute have the same proportion.

[0041] It should be noted that the above is only a detailed description of the preferred embodiments and principles of the present invention. For ordinary technicians in this field, based on the ideas provided by the present invention, there will be changes in the specific implementation methods, and these changes should also be regarded as the scope of protection of the present invention.

Claims

1. A suspicious organization query method for public security heterogeneous information networks, characterized by: The following steps are involved: S1, obtain the personnel relationship dataset in the public security field, A heterogeneous information network in the public security field is constructed based on the personnel relationship dataset, wherein the heterogeneous information network includes vertices, edges connecting vertices, and vertex attributes; the vertices include vertices representing three types: objects, cases, and events; the edges are used to represent the relationship between the vertices at both ends of the edge; and each vertex includes text attributes and numerical attributes; S2, obtaining a clue of a suspicious person, and traversing the heterogeneous information network with a meta-path based on the clue of the suspicious person to obtain a set of candidate suspicious persons, where the meta-path refers to a specific path connecting two types of vertices; S3, generating an induced subgraph based on the candidate suspicious person set, and removing the candidate suspicious persons that do not meet the k-core constraint from the induced subgraph according to the k-core constraint, to obtain a k-core subgraph containing the suspicious person, wherein the vertex set of the k-core subgraph is a close suspicious person set containing the clue suspicious person; S4, constructing a search tree based on the set of closely related suspicious persons, wherein the search tree records all possible k-core subgraphs containing the clue suspicious persons; S5, traverse the search tree, calculate the attribute distance of each k-core subgraph, and select the subgraph with the smallest attribute distance as the result and return it. The set of all vertices in the returned subgraph is the suspicious organization containing the suspicious person; the calculation formula of the attribute distance of the k-core subgraph is: v1, v2 are any two points in the subgraph H, N is the number of object type vertices in the subgraph H; where f(v1, v2) = α*f t (v1,v2)+(1-α)*f n (v1,v2); where f t (v1,v2) represents the text attribute distance between suspicious persons v1 and v2, f n (v1, v2) represents the numerical attribute distance between suspicious persons v1 and v2, and α is the weight ratio of the text attribute distance and the numerical attribute example.

2. A suspicious organization query method for a public security heterogeneous information network according to claim 1, characterized in that: Step S2 specifically includes the following steps: S2-1 The obtained clue suspicious person is set as q, the meta path is set as p, and the queue to be processed Q is initialized. c For the clue suspicious person {q}, initialize the candidate suspicious person set C as the clue suspicious person {q}, the waiting queue Q c Suspicious persons to be traversed are stored in it; S2-2 For any suspicious person v∈Q c , find all suspicious persons connected to the meta-path p starting from the suspicious person v, and save them into the set Ψ[v]. For any suspicious person u in Ψ[v], if the suspicious person u is not included in the candidate suspicious person set C, then the suspicious person u is added to the candidate suspicious person set C and the waiting queue Q at the same time. c Repeat the above steps until the queue to be processed Q c is empty; S2-3 All the suspicious persons in the candidate suspicious person set C constitute the candidate suspicious person set related to the clue suspicious person q.

3. A suspicious organization query method for a public security heterogeneous information network according to claim 2, characterized in that: Step S4 specifically includes the following steps: S4-1 initializes a search tree T based on the set of closely related suspicious persons. Each tree node in the search tree T records a k-core subgraph and its child nodes. The root node root of the initialized search tree records the induced subgraph H generated by the set of candidate suspicious persons in step 3. Initializes a queue to be constructed Q as the root node {root} of the search tree. The queue to be constructed Q stores the search tree nodes to be constructed. S4-2 For any tree node n∈Q, take out the subgraph H in the tree node n. The set of suspicious persons in the subgraph H is recorded as S. For any suspicious person s in the suspicious person set S, try to remove s from S, and then find the suspicious person set that still meets the k-core constraint from the induced subgraph composed of the remaining suspicious persons. if If it is not empty, a new tree node n1 is created and recorded as a child node of tree node n. Tree node n1 records the set of suspicious persons. The induced subgraph constructed adds the tree node n1 to the queue to be constructed Q; S4-3 repeats step S4-2 until the queue Q to be constructed is empty, completing the construction of the search tree.

Images