Movatterモバイル変換

[0]ホーム
URL:

CN115858946A - Graph theory-based clue reasoning and intelligence prediction method - Google Patents

Graph theory-based clue reasoning and intelligence prediction methodDownload PDF

Info

Publication number
CN115858946A
CN115858946ACN202310151256.1ACN202310151256ACN115858946ACN 115858946 ACN115858946 ACN 115858946ACN 202310151256 ACN202310151256 ACN 202310151256ACN 115858946 ACN115858946 ACN 115858946A
Authority
CN
China
Prior art keywords
gang
graph
intelligence
criminal
predicted
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202310151256.1A
Other languages
Chinese (zh)
Inventor
王青旺
熊豪
沈韬
汪志锋
刘全君
宋健
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Kunming University of Science and Technology
Original Assignee
Kunming University of Science and Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Kunming University of Science and TechnologyfiledCriticalKunming University of Science and Technology
Priority to CN202310151256.1ApriorityCriticalpatent/CN115858946A/en
Publication of CN115858946ApublicationCriticalpatent/CN115858946A/en
Priority to CN202310544005.XAprioritypatent/CN116578719A/en
Pendinglegal-statusCriticalCurrent

Links

Images

Classifications

Landscapes

Abstract

The invention discloses a clue reasoning and intelligence predicting method based on graph theory, belonging to the technical field of social network analysis and comprising the following steps: acquiring social relations and known information of the gangs to be predicted, and constructing a ganged-information network diagram according to the known clues, cases, information and the related information of the member relations; predicting the relevance of the members in the group with other clues and case information, and updating the group-intelligence network diagram of the criminal group; and constructing a ganged-intelligence network diagram of the broken crime gangs, comparing the correlation between the crime trend gangs to be predicted and the broken gangs, and predicting the crime trend of the gangs. The method can fully mine the correlation between the gangues and information such as other clues and cases, predict the criminal tendency of other gangues and effectively improve the case detection rate.

Description

Translated fromChinese
一种基于图论的线索推理与情报预测方法A method of clue reasoning and intelligence prediction based on graph theory

技术领域Technical Field

本发明涉及一种基于图论的线索推理与情报预测方法，属于社交网络分析技术领域。The invention relates to a clue reasoning and intelligence prediction method based on graph theory, belonging to the technical field of social network analysis.

背景技术Background Art

随着社会的发展，犯罪手段和技术也在不断地演进，呈现出团伙作案的趋势。在团伙作案的过程中，会出现有组织、有分工地进行作案。犯罪团伙往往分工明确，在办案过程中会出现相关线索与个别成员形成密切联系的情况，但在办案过程中根据已知线索只能打掉其中一个环节或者几个环节，团伙核心作案团体不会受到根本上的打击，犯罪团伙仍然可能重新组织人手继续犯罪。因而，从产生原因、发生过程及危害等方面来看，犯罪首先是一种社会现象，社会性是犯罪行为的根本属性。一个犯罪网络首先也是一个社会网络，在网络中个体与他人通过亲属关系、朋友关系与同事关系等各种关系联系在一起。研究有组织犯罪预防与打击对策，不仅要探究犯罪人的特征与行为，还要分析团伙与整个网络的组织、结构与运作模式。社交关系网络分析为推进疑难案件的侦破带来了希望。With the development of society, criminal means and techniques are constantly evolving, showing a trend of gang crimes. In the process of gang crimes, there will be organized and divided areas for crime. Criminal gangs often have clear division of labor. In the process of handling cases, relevant clues will form close links with individual members. However, in the process of handling cases, only one or several links can be eliminated based on known clues. The core criminal group of the gang will not be fundamentally hit, and the criminal gang may still reorganize its manpower to continue committing crimes. Therefore, from the perspective of causes, occurrence process and harm, crime is first of all a social phenomenon, and sociality is the fundamental attribute of criminal behavior. A criminal network is first of all a social network, in which individuals and others are connected through various relationships such as kinship, friendship and colleague relationships. To study the prevention and combat measures of organized crime, it is necessary not only to explore the characteristics and behaviors of criminals, but also to analyze the organization, structure and operation mode of the gang and the entire network. Social relationship network analysis has brought hope for the detection of difficult cases.

犯罪网络除具有社会网络的一些特点外，还具有以下特征：其一，网络中的节点代表犯罪分子，节点间的边表示犯罪分子之间的联系，如消息传递，共同参与某项活动。其二，各节点在网络中的作用或者“位置”基本不同，核心成员往往掌握犯罪团伙重要信息，属于网络的少数。其三，大的犯罪网络往往由几个子团伙组成，在整个团伙中各子团伙起着不同的功能作用。其四，犯罪团伙并不是孤立存在的，团伙间存在一些联系和交互，因此，对犯罪团伙进行信息挖掘、分析团伙内的交互信息并挖掘团伙间的隐性关系显得尤为重要。In addition to some characteristics of social networks, criminal networks also have the following features: First, the nodes in the network represent criminals, and the edges between nodes represent the connections between criminals, such as message transmission and joint participation in an activity. Second, the role or "position" of each node in the network is basically different. The core members often have important information about the criminal gang and belong to the minority of the network. Third, large criminal networks are often composed of several sub-gangs, and each sub-gang plays a different functional role in the entire gang. Fourth, criminal gangs do not exist in isolation. There are some connections and interactions between gangs. Therefore, it is particularly important to mine information about criminal gangs, analyze the interactive information within the gangs, and mine the implicit relationships between gangs.

目前，多数方法在犯罪团伙社交网络中未能实现从海量信息中发现团伙犯罪线索和推理的情报预警，对于团伙挖掘的效率相对较低，未能有效地推断团伙成员与海量信息间的隐性联系。同时，目前方法未能有效利用已侦破案件辅助侦破疑难案件，因此，如何对海量信息进行建模，辅助侦破疑难案件，同时根据已知信息推理案件中犯罪成员与相关线索之间的隐性关系，是目前亟待解决的技术问题。At present, most methods fail to achieve intelligence warnings of gang crime clues and inferences from massive information in criminal gang social networks. The efficiency of gang mining is relatively low, and the implicit connection between gang members and massive information cannot be effectively inferred. At the same time, current methods fail to effectively use solved cases to assist in solving difficult cases. Therefore, how to model massive information, assist in solving difficult cases, and at the same time infer the implicit relationship between criminal members and related clues in the case based on known information is a technical problem that needs to be solved urgently.

发明内容Summary of the invention

本发明要解决的技术问题是提出了一种基于图论的线索推理与情报预测方法，用以解决传统方法未能充分挖掘团伙与其他线索、案件等信息的相关性；未能有效预测其他团伙的犯罪趋势的问题。The technical problem to be solved by the present invention is to propose a clue reasoning and intelligence prediction method based on graph theory to solve the problem that traditional methods fail to fully explore the correlation between gangs and other clues, cases and other information; and fail to effectively predict the crime trends of other gangs.

本发明的技术方案是：一种基于图论的线索推理与情报预测方法，其特征在于：获取团伙的社交关系，根据已知线索、案件、情报与成员相关信息构建社交关系网络图；预测团伙中的成员与其他线索、案件等信息的相关性，更新社交关系网络图；构建已知团伙集合，生成待预测犯罪趋势团伙的已知信息社交关系网络图；比较待预测犯罪趋势团伙与已知团伙的相关性，预测团伙的犯罪趋势；The technical solution of the present invention is: a clue reasoning and intelligence prediction method based on graph theory, characterized by: obtaining the social relationship of the gang, building a social relationship network diagram based on known clues, cases, intelligence and member-related information; predicting the correlation between members in the gang and other clues, cases and other information, and updating the social relationship network diagram; building a known gang set, generating a known information social relationship network diagram of the gang with the crime trend to be predicted; comparing the correlation between the gang with the crime trend to be predicted and the known gangs, and predicting the crime trend of the gang;

具体步骤为：The specific steps are:

Step1：获取待预测团伙的社交关系及已知情报信息，根据已知线索、案件、情报与成员关系相关信息构建团伙-情报网络图。Step 1: Obtain the social relationships and known intelligence information of the gang to be predicted, and build a gang-intelligence network diagram based on known clues, cases, intelligence and member relationship related information.

Step2：预测团伙中的成员与其他线索、案件信息的相关性，更新犯罪团伙的团伙-情报网络图。Step 2: Predict the correlation between gang members and other clues and case information, and update the gang-intelligence network diagram of the criminal gang.

Step3：构建已破获的犯罪团伙的团伙-情报网络图。Step 3: Construct a gang-intelligence network diagram of the uncovered criminal gangs.

Step4：比较待预测犯罪趋势团伙与已破获团伙的相关性，预测团伙的犯罪趋势。Step 4: Compare the correlation between the gangs whose criminal trends are to be predicted and the gangs that have been cracked, and predict the criminal trends of the gangs.

所述Step1中，通过图论建模犯罪团伙的团伙-情报网络图，将其定义为G=(V, E)，其中V表示图中节点集合，具体包含团伙中已知线索、案件、情报与成员关系相关信息节点，E表示节点间关系。InStep 1, the gang-intelligence network graph of the criminal gang is modeled by graph theory and defined asG= (V, E ), whereV represents the set of nodes in the graph, specifically including nodes related to known clues, cases, intelligence and member relationships in the gang, andE represents the relationship between nodes.

特别地，基于图论的犯罪团伙团伙-情报网络图G中，节点V的特征以四元数形式表示，即

Figure SMS_1
，其中H表示特征空间，M为节点维度，N为特征维度，d为对偶四元空间。In particular, in the criminal gang-intelligence network graphG based on graph theory, the characteristics of the nodeV are expressed in the form of quaternions, that is,
Figure SMS_1
, where H represents the feature space,M is the node dimension,N is the feature dimension,and d is the dual quaternion space.

Step2.1：构建四元空间空域图卷积运算，将表示团伙中的成员、相关线索节点的V建模为四元空间中的节点特征向量集，在四元空间中进行图卷积，即采用空域图卷积算子构建四元空间空域图卷积：

Figure SMS_4
其中，上标DQ表示对偶四元空间d，k表示卷积迭代次数；
Figure SMS_7
表示非线性激活函数；v和u表示实体或者关系节点，
Figure SMS_10
表示所有节点的集合，
Figure SMS_2
是拉普拉斯算子归一化的邻接矩阵
Figure SMS_5
中的节点v和u之间的边常数，其中
Figure SMS_9
，
Figure SMS_12
是
Figure SMS_3
的对角线节点度矩阵，
Figure SMS_6
表示输入的加权邻接矩阵；W(k),DQ表示对偶四元数权重矩阵；
Figure SMS_8
表示对偶四元数乘法；
Figure SMS_11
表示在对偶四元空间中第k代实体或关系节点。Step 2.1: Construct a quaternary spatial domain graph convolution operation, modelV representing the members of the gang and related clue nodes as a set of node feature vectors in the quaternary space, and perform graph convolution in the quaternary space, that is, use the spatial domain graph convolution operator to construct the quaternary spatial domain graph convolution:
Figure SMS_4
Wherein, the superscriptDQ represents the dual quaternion spaced , and k represents the number of convolution iterations;
Figure SMS_7
represents a nonlinear activation function;v andu represent entity or relationship nodes,
Figure SMS_10
represents the set of all nodes,
Figure SMS_2
is the Laplacian-normalized adjacency matrix
Figure SMS_5
The edge constant between nodesv andu in
Figure SMS_9
,
Figure SMS_12
yes
Figure SMS_3
The diagonal node degree matrix of
Figure SMS_6
represents the input weighted adjacency matrix;W(k ),DQ represents the dual quaternion weight matrix;
Figure SMS_8
Represents dual quaternion multiplication;
Figure SMS_11
Represents the k-th generation entity or relationship node in the dual quaternion space.

Step2.2：根据四元空间空域图卷积计算团伙节点成员四元空间特征和其相关关系节点四元空间特征，将其连接起来产生新的四元空间特征，将待预测的团伙-情报网络图建模为待预测犯罪团伙的情报-网络知识图谱，将知识图谱头节点表示为vhQ，将知识图谱关系节点表示为vrQ，将知识图谱尾节点表示为vtQ；其中，h表示知识图谱中的头，r表示知识图谱中的关系，t表示知识图谱中的尾，上标Q表示对偶四元空间d。Step 2.2: Calculate the quaternary space features of gang node members and the quaternary space features of their related relationship nodes according to the quaternary space spatial domain graph convolution, connect them to generate new quaternary space features, model the gang-intelligence network graph to be predicted as the intelligence-network knowledge graph of the criminal gang to be predicted, represent the head node of the knowledge graph asvhQ , represent the relationship node of the knowledge graph asvrQ , and represent the tail node of the knowledge graph asvtQ ; whereh represents the head in the knowledge graph,r represents the relationship in the knowledge graph,t represents the tail in the knowledge graph, and the superscriptQ represents the dual quaternary space d.

Step2.3：为待预测犯罪团伙的情报-网络知识图谱构建新的关系，根据四元分数评价公式计算新建关系的得分：

Figure SMS_13
Step 2.3: Build a new relationship for the intelligence-network knowledge graph of the criminal gang to be predicted, and calculate the score of the newly created relationship according to the four-element score evaluation formula:
Figure SMS_13

其中，h表示知识图谱中的头，r表示知识图谱中的关系，t表示知识图谱中的尾，上标Q表示对偶四元空间d，

Figure SMS_14
表示Hamilton乘法，
Figure SMS_15
表示归一化的四元数，·表示四元数内积；
Figure SMS_16
表示对偶四元空间头节点，
Figure SMS_17
表示归一化的对偶四元空间关系节点，
Figure SMS_18
表示对偶四元空间尾节点。Among them,h represents the head in the knowledge graph,r represents the relationship in the knowledge graph,t represents the tail in the knowledge graph, and the superscriptQ represents the dual quaternion space d.
Figure SMS_14
represents Hamilton multiplication,
Figure SMS_15
represents the normalized quaternion, · represents the inner product of the quaternion;
Figure SMS_16
represents the dual quaternion space head node,
Figure SMS_17
represents the normalized dual quaternion space relationship node,
Figure SMS_18
Represents the tail node of the dual quaternion space.

Step2.4：选取评分前2名的新建关系作为挖掘出的有效关系，保留有效关系，对应犯罪团伙的团伙-情报网络图中的边，更新犯罪团伙的团伙-情报网络图。Step 2.4: Select the top two newly created relationships as the mined valid relationships, retain the valid relationships, correspond to the edges in the gang-intelligence network graph of the criminal gang, and update the gang-intelligence network graph of the criminal gang.

将Step3得到的已知团伙的团伙-情报网络图记为

Figure SMS_19
；将待预测团伙的团伙-情报网络图记为
Figure SMS_20
；其中，
Figure SMS_21
解释为已知团伙n犯罪类别i对应倾向的特征值，
Figure SMS_22
解释为待预测团伙犯罪类别i对应倾向的特征值，上标
Figure SMS_23
表示待预测；根据犯罪倾向预测公式计算待预测团伙的团伙-情报网络图
Figure SMS_24
与已知犯罪团伙犯罪类别倾向度：
Figure SMS_25
The gang-intelligence network graph of the known gang obtained inStep 3 is denoted as
Figure SMS_19
; The gang-intelligence network graph of the gang to be predicted is recorded as
Figure SMS_20
;in,
Figure SMS_21
It is interpreted as the characteristic value of the corresponding tendency of the crime categoryi of the known gangn ,
Figure SMS_22
It is interpreted as the characteristic value of the corresponding tendency of the gang crime categoryi to be predicted, with the superscript
Figure SMS_23
Indicates the group to be predicted; calculates the gang-intelligence network diagram of the group to be predicted based on the criminal tendency prediction formula
Figure SMS_24
Propensity to crime categories with known criminal gangs:
Figure SMS_25

其中，

Figure SMS_26
表示犯罪类别i倾向度，
Figure SMS_27
表示任意已知团伙犯罪类别i对应倾向的特征值，
Figure SMS_28
表示待预测团伙犯罪类别i对应倾向的特征值，上标
Figure SMS_29
表示待预测，G表示输入的关系网络图，W与
Figure SMS_30
表示可学习权重，b表示标量偏置，exp指以自然常数e为底的指数函数，I表示犯罪类别数量；根据犯罪类别倾向度评分，比较得出待预测团伙的犯罪趋势的预测结果。in,
Figure SMS_26
represents the tendency of crime categoryi ,
Figure SMS_27
represents the characteristic value of the corresponding tendency of any known gang crime categoryi ,
Figure SMS_28
Indicates the characteristic value of the corresponding tendency of the gang crime categoryi to be predicted, with superscript
Figure SMS_29
represents the prediction to be made, G represents the input relationship network diagram,W and
Figure SMS_30
represents the learnable weight,b represents the scalar bias, exp refers to the exponential function with the natural constant e as the base, and I represents the number of crime categories. According to the crime category tendency score, the prediction result of the crime trend of the gang to be predicted is obtained by comparison.

和现有技术相比，本发明的有益效果为：本发明能够充分挖掘团伙与其他线索、案件等信息的相关性，预测其他团伙的犯罪趋势，能够加快案件侦破；通过已知的团伙社交关系预测、推理团伙中成员与其他线索和案件的关系、通过对已知的团伙定义犯罪属性，预测其他未知团伙的犯罪趋势；能够对犯罪团伙进行信息挖掘、分析团伙内的交互信息并挖掘团伙间的隐性关系。Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention can fully explore the correlation between gangs and other clues, cases and other information, predict the crime trends of other gangs, and speed up case detection; predict the crime trends of other unknown gangs by predicting known gang social relationships, inferring the relationship between gang members and other clues and cases, and defining criminal attributes for known gangs; and can conduct information mining on criminal gangs, analyze interactive information within gangs and mine implicit relationships between gangs.

附图说明BRIEF DESCRIPTION OF THE DRAWINGS

图1是本发明的步骤流程图。FIG. 1 is a flow chart of the steps of the present invention.

图2为本发明的实施例1社交关系及情报信息示意图。FIG. 2 is a schematic diagram of social relationships and intelligence information according to Example 1 of the present invention.

实施方式Implementation

下面结合附图和具体实施例对本发明作进一步详细说明，但本发明的保护范围并不限于所述内容。The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments, but the protection scope of the present invention is not limited to the contents described above.

实施例1Example 1

Step1：获取待预测团伙的社交关系及已知情报信息，将团伙中的已知线索、案件、情报与成员关系相关信息建模为节点，如图2中所示，现有群体含有4个相关信息节点：[信息1，信息2，信息3，信息4]，通过图论建模团伙-情报网络图，将其定义为：

Figure SMS_31
，其中V表示团伙-情报网络图中的相关信息节点，E表示节点间已知关系，如：图2中（a）图的信息1与信息4之间有已知关系，信息2与信息3有已知关系，用实线表示；定义不同的数值用以表示关系的强度，已知关系的关系强度可以人为定义。Step 1: Obtain the social relationships and known intelligence information of the gang to be predicted, and model the known clues, cases, intelligence and member relationship information in the gang as nodes. As shown in Figure 2, the existing group contains 4 related information nodes: [Information 1,Information 2,Information 3, Information 4]. The gang-intelligence network diagram is modeled by graph theory and defined as:
Figure SMS_31
, whereV represents the relevant information nodes in the gang-intelligence network diagram, andE represents the known relationship between nodes. For example, in Figure 2 (a), there is a known relationship betweeninformation 1 andinformation 4, and there is a known relationship betweeninformation 2 andinformation 3, which are represented by solid lines; different numerical values are defined to represent the strength of the relationship, and the strength of the known relationship can be defined artificially.

Step2：为了获取团伙-情报网络图中未知的关系，根据四元空间空域图卷积计算团伙信息节点四元空间特征，如下公式所示：

Figure SMS_33
其中，上标DQ表示对偶四元空间d，k表示卷积迭代次数；
Figure SMS_37
表示非线性激活函数；v和u表示实体或者关系节点，
Figure SMS_41
表示所有节点的集合，
Figure SMS_34
是拉普拉斯算子归一化的邻接矩阵
Figure SMS_38
中的节点v和u之间的边常数，其中
Figure SMS_39
，
Figure SMS_42
是
Figure SMS_32
的对角线节点度矩阵，
Figure SMS_35
表示输入的加权邻接矩阵；W(k),DQ表示对偶四元数权重矩阵；
Figure SMS_36
表示对偶四元数乘法；
Figure SMS_40
表示在对偶四元空间中第k代实体或关系节点。Step 2: In order to obtain the unknown relationships in the gang-intelligence network graph, the four-dimensional space features of the gang information nodes are calculated according to the four-dimensional space domain graph convolution, as shown in the following formula:
Figure SMS_33
Wherein, the superscriptDQ represents the dual quaternion spaced , and k represents the number of convolution iterations;
Figure SMS_37
represents a nonlinear activation function;v andu represent entity or relationship nodes,
Figure SMS_41
represents the set of all nodes,
Figure SMS_34
is the Laplacian-normalized adjacency matrix
Figure SMS_38
The edge constant between nodesv andu in
Figure SMS_39
,
Figure SMS_42
yes
Figure SMS_32
The diagonal node degree matrix of
Figure SMS_35
represents the input weighted adjacency matrix;W(k ),DQ represents the dual quaternion weight matrix;
Figure SMS_36
Represents dual quaternion multiplication;
Figure SMS_40
Represents the k-th generation entity or relationship node in the dual quaternion space.

根据更新后的信息节点四元空间特征

Figure SMS_43
，将待预测的团伙-情报网络图建模为待预测犯罪团伙的情报-网络知识图谱，将知识图谱头节点表示为vhQ，将知识图谱关系节点表示为vrQ，将知识图谱尾节点表示为vtQ；其中，h表示知识图谱中的头，r表示知识图谱中的关系，t表示知识图谱中的尾，上标Q表示对偶四元空间d；According to the updated information node quaternary space features
Figure SMS_43
, the gang-intelligence network graph to be predicted is modeled as the intelligence-network knowledge graph of the criminal gang to be predicted, the head node of the knowledge graph is represented asvhQ , the relationship node of the knowledge graph is represented asvrQ , and the tail node of the knowledge graph is represented asvtQ ; whereh represents the head in the knowledge graph,r represents the relationship in the knowledge graph,t represents the tail in the knowledge graph, and the superscriptQ represents the dual quaternion space d;

根据四元分数评价公式计算新建关系的得分：

Figure SMS_44
其中，h表示知识图谱中的头，r表示知识图谱中的关系，t表示知识图谱中的尾，上标Q表示对偶四元空间d，
Figure SMS_45
表示Hamilton乘法，
Figure SMS_46
表示归一化的四元数，·表示四元数内积；
Figure SMS_47
表示对偶四元空间头节点，
Figure SMS_48
表示归一化的对偶四元空间关系节点，
Figure SMS_49
表示对偶四元空间尾节点；结合图1，新建关系的得分计算结果示意如表1所示，其中“-”表示已知的无关系或确定关系。The score of the newly created relationship is calculated according to the four-element score evaluation formula:
Figure SMS_44
Among them,h represents the head in the knowledge graph,r represents the relationship in the knowledge graph,t represents the tail in the knowledge graph, and the superscriptQ represents the dual quaternion space d.
Figure SMS_45
represents Hamilton multiplication,
Figure SMS_46
represents the normalized quaternion, · represents the inner product of the quaternion;
Figure SMS_47
represents the dual quaternion space head node,
Figure SMS_48
represents the normalized dual quaternion space relationship node,
Figure SMS_49
Represents the tail node of the dual quaternion space; combined with Figure 1, the score calculation results of the newly created relationship are shown in Table 1, where "-" represents a known no relationship or a definite relationship.

表1 新建关系的得分计算结果示意

Figure SMS_50
Table 1 Schematic diagram of the score calculation results of the newly created relationship
Figure SMS_50

选取评分前2名的新建关联作为挖掘出的新建有效关系，保留已知有效关系与新建有效关系，对应犯罪团伙的团伙-情报网络图中的边，更新犯罪团伙的团伙-情报网络图，如图2（a）中所示，实线表示已知的有效关系，虚线表示新添加的有效关系。The newly created associations with the top two scores are selected as the newly mined valid relationships, and the known valid relationships and the newly created valid relationships are retained, corresponding to the edges in the gang-intelligence network diagram of the criminal gang. The gang-intelligence network diagram of the criminal gang is updated, as shown in Figure 2 (a). The solid line represents the known valid relationship, and the dotted line represents the newly added valid relationship.

Step3：构建已破获的犯罪团伙的团伙-情报网络图，如图2中的（b）（c）所示；Step 3: Construct the gang-intelligence network diagram of the cracked criminal gang, as shown in (b) and (c) in Figure 2;

Step4：根据犯罪倾向预测公式计算待预测团伙的团伙-情报网络图

Figure SMS_51
与已知犯罪团伙犯罪类别倾向度：
Figure SMS_52
其中，
Figure SMS_53
表示犯罪类别i倾向度，
Figure SMS_54
表示任意已知团伙犯罪类别i对应倾向的特征值，
Figure SMS_55
表示待预测团伙犯罪类别i对应倾向的特征值，上标
Figure SMS_56
表示待预测，G表示输入的关系网络图，W与
Figure SMS_57
表示可学习权重，b表示标量偏置，exp指以自然常数e为底的指数函数，I表示犯罪类别数量。Step 4: Calculate the gang-intelligence network diagram of the gang to be predicted based on the criminal tendency prediction formula
Figure SMS_51
Propensity to crime categories with known criminal gangs:
Figure SMS_52
in,
Figure SMS_53
represents the tendency of crime categoryi ,
Figure SMS_54
represents the characteristic value of the corresponding tendency of any known gang crime categoryi ,
Figure SMS_55
Indicates the characteristic value of the corresponding tendency of the gang crime categoryi to be predicted, with superscript
Figure SMS_56
represents the prediction to be made, G represents the input relationship network diagram,W and
Figure SMS_57
represents the learnable weight,b represents the scalar bias, exp refers to the exponential function with the natural constant e as the base, and I represents the number of crime categories.

例如，假定已知团伙1[犯罪类别1，犯罪类别2]的具体特征值为

Figure SMS_59
，已知团伙2[犯罪类别1，犯罪类别2]的具体特征值为
Figure SMS_61
，待预测团伙[犯罪类别1，犯罪类别2]的具体特征值为
Figure SMS_63
；假定犯罪倾向预测公式中，可学习权重的值为
Figure SMS_58
，
Figure SMS_62
，偏置
Figure SMS_64
；那么，通过犯罪倾向预测公式可以计算得到
Figure SMS_65
；显然，通过比较可以得到结果
Figure SMS_60
，表明待预测团伙更加倾向于与已知犯罪团伙1进行犯罪类别1活动，第二倾向于与犯罪团伙2进行犯罪类别2活动，以此类推，得到预测的所有犯罪倾向。For example, suppose the specific feature values of gang 1 [crime category 1, crime category 2] are known to be
Figure SMS_59
, the specific feature value of gang 2 [crime category 1, crime category 2] is known to be
Figure SMS_61
, the specific feature value of the gang to be predicted [crime category 1, crime category 2] is
Figure SMS_63
; Assume that in the crime tendency prediction formula, the value of the learnable weight is
Figure SMS_58
,
Figure SMS_62
, bias
Figure SMS_64
; Then, the crime tendency prediction formula can be used to calculate
Figure SMS_65
; Obviously, the result can be obtained by comparison
Figure SMS_60
, indicating that the gang to be predicted is more inclined to engage incrime category 1 activities with knowncriminal gang 1, and is secondly inclined to engage incrime category 2 activities withcriminal gang 2, and so on, to obtain all the predicted criminal tendencies.

以上结合附图对本发明的具体实施方式作了详细说明，但是本发明并不限于上述实施方式，在本领域普通技术人员所具备的知识范围内，还可以在不脱离本发明宗旨的前提下作出各种变化。The specific implementation modes of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above implementation modes, and various changes can be made within the knowledge scope of ordinary technicians in this field without departing from the purpose of the present invention.

Claims (4)

Translated fromChinese
1.一种基于图论的线索推理与情报预测方法，其特征在于：具体步骤为：1. A clue reasoning and intelligence prediction method based on graph theory, characterized in that: the specific steps are:Step1：获取待预测团伙的社交关系及已知情报信息，根据已知线索、案件、情报与成员关系相关信息构建团伙-情报网络图；Step 1: Obtain the social relationships and known intelligence information of the gang to be predicted, and build a gang-intelligence network diagram based on known clues, cases, intelligence and member relationship information;Step2：预测团伙中的成员与其他线索、案件信息的相关性，更新犯罪团伙的团伙-情报网络图；Step 2: Predict the correlation between gang members and other clues and case information, and update the gang-intelligence network diagram of the criminal gang;Step3：构建已破获的犯罪团伙的团伙-情报网络图；Step 3: Construct a gang-intelligence network diagram of the uncovered criminal gangs;Step4：比较待预测犯罪趋势团伙与已破获团伙的相关性，预测团伙的犯罪趋势。Step 4: Compare the correlation between the gangs whose criminal trends are to be predicted and the gangs that have been cracked, and predict the criminal trends of the gangs.2.根据权利要求1所述的基于图论的线索推理与情报预测方法，其特征在于：Step1中，通过图论建模犯罪团伙的团伙-情报网络图，将其定义为G=(V, E)，其中V表示图中节点集合，具体包含团伙中已知线索、案件、情报与成员关系相关信息节点，E表示节点间关系；2. The clue reasoning and intelligence prediction method based on graph theory according to claim 1 is characterized in that: in Step 1, the gang-intelligence network graph of the criminal gang is modeled by graph theory, and is defined asG= (V, E ), whereV represents the node set in the graph, specifically including nodes related to known clues, cases, intelligence and member relationships in the gang, andE represents the relationship between nodes;特别地，基于图论的犯罪团伙团伙-情报网络图G中，节点V的特征以四元数形式表示，即
Figure QLYQS_1
，其中H表示特征空间，M为节点维度，N为特征维度，d为对偶四元空间。In particular, in the criminal gang-intelligence network graphG based on graph theory, the characteristics of the nodeV are expressed in the form of quaternions, that is,
Figure QLYQS_1
, where H represents the feature space,M is the node dimension,N is the feature dimension,and d is the dual quaternion space.3.根据权利要求1所述的基于图论的线索推理与情报预测方法，其特征在于，在Step2中更新犯罪团伙-情报网络图的具体步骤为：3. The clue reasoning and intelligence prediction method based on graph theory according to claim 1 is characterized in that the specific steps of updating the criminal gang-intelligence network diagram in Step 2 are:Step2.1：构建四元空间空域图卷积运算，将表示团伙中的成员、相关线索节点的V建模为四元空间中的节点特征向量集，在四元空间中进行图卷积，即采用空域图卷积算子构建四元空间空域图卷积：
Figure QLYQS_2
；Step 2.1: Construct a quaternary spatial domain graph convolution operation, modelV representing the members of the gang and related clue nodes as a set of node feature vectors in the quaternary space, and perform graph convolution in the quaternary space, that is, use the spatial domain graph convolution operator to construct the quaternary spatial domain graph convolution:
Figure QLYQS_2
;其中，上标DQ表示对偶四元空间d，k表示卷积迭代次数；
Figure QLYQS_5
表示非线性激活函数；v和u表示实体或者关系节点，
Figure QLYQS_7
表示所有节点的集合，
Figure QLYQS_10
是拉普拉斯算子归一化的邻接矩阵
Figure QLYQS_4
中的节点v和u之间的边常数，其中
Figure QLYQS_6
,
Figure QLYQS_9
是
Figure QLYQS_12
的对角线节点度矩阵，
Figure QLYQS_3
表示输入的加权邻接矩阵；W(k),DQ表示对偶四元数权重矩阵；
Figure QLYQS_8
表示对偶四元数乘法；
Figure QLYQS_11
表示在对偶四元空间中第k代实体或关系节点；Wherein, the superscriptDQ represents the dual quaternion spaced , and k represents the number of convolution iterations;
Figure QLYQS_5
represents a nonlinear activation function;v andu represent entity or relationship nodes,
Figure QLYQS_7
represents the set of all nodes,
Figure QLYQS_10
is the Laplacian-normalized adjacency matrix
Figure QLYQS_4
The edge constant between nodesv andu in
Figure QLYQS_6
,
Figure QLYQS_9
yes
Figure QLYQS_12
The diagonal node degree matrix of
Figure QLYQS_3
represents the input weighted adjacency matrix;W(k ),DQ represents the dual quaternion weight matrix;
Figure QLYQS_8
Represents dual quaternion multiplication;
Figure QLYQS_11
Represents the k-th generation entity or relationship node in the dual quaternion space;Step2.2：根据四元空间空域图卷积计算团伙节点成员四元空间特征和其相关关系节点四元空间特征，将其连接起来产生新的四元空间特征，将待预测的团伙-情报网络图建模为待预测犯罪团伙的情报-网络知识图谱，将知识图谱头节点表示为vhQ，将知识图谱关系节点表示为vrQ，将知识图谱尾节点表示为vtQ；其中，h表示知识图谱中的头，r表示知识图谱中的关系，t表示知识图谱中的尾，上标Q表示对偶四元空间d；Step 2.2: Calculate the quaternary space features of gang node members and the quaternary space features of their related relationship nodes according to the quaternary space spatial domain graph convolution, connect them to generate new quaternary space features, model the gang-intelligence network graph to be predicted as the intelligence-network knowledge graph of the criminal gang to be predicted, represent the head node of the knowledge graph asvhQ , represent the relationship node of the knowledge graph asvrQ , and represent the tail node of the knowledge graph asvtQ ; whereh represents the head in the knowledge graph,r represents the relationship in the knowledge graph,t represents the tail in the knowledge graph, and the superscriptQ represents the dual quaternary space d;Step2.3：为待预测犯罪团伙的情报-网络知识图谱构建新的关系，根据四元分数评价公式计算新建关系的得分：
Figure QLYQS_13
；Step 2.3: Build a new relationship for the intelligence-network knowledge graph of the criminal gang to be predicted, and calculate the score of the newly created relationship according to the four-element score evaluation formula:
Figure QLYQS_13
;其中，h表示知识图谱中的头，r表示知识图谱中的关系，t表示知识图谱中的尾，上标Q表示对偶四元空间d，
Figure QLYQS_14
表示Hamilton乘法，
Figure QLYQS_15
表示归一化的四元数，·表示四元数内积；vhQ表示对偶四元空间头节点，
Figure QLYQS_16
表示归一化的对偶四元空间关系节点，vtQ表示对偶四元空间尾节点；Among them,h represents the head in the knowledge graph,r represents the relationship in the knowledge graph,t represents the tail in the knowledge graph, and the superscriptQ represents the dual quaternion space d.
Figure QLYQS_14
represents Hamilton multiplication,
Figure QLYQS_15
represents the normalized quaternion, · represents the inner product of the quaternion;vhQ represents the dual quaternion space head node,
Figure QLYQS_16
represents the normalized dual quaternion space relationship node,vtQ represents the dual quaternion space tail node;Step2.4：选取评分前2名的新建关系作为挖掘出的有效关系，保留有效关系，对应犯罪团伙的团伙-情报网络图中的边，更新犯罪团伙的团伙-情报网络图。Step 2.4: Select the top two newly created relationships as the mined valid relationships, retain the valid relationships, correspond to the edges in the gang-intelligence network graph of the criminal gang, and update the gang-intelligence network graph of the criminal gang.4.根据权利要求1所述的基于图论的线索推理与情报预测方法，其特征在于，在Step4中预测待预测团伙的犯罪趋势，具体过程为：4. The clue reasoning and intelligence prediction method based on graph theory according to claim 1 is characterized in that the crime trend of the gang to be predicted is predicted in Step 4, and the specific process is:将Step3得到的已知团伙的团伙-情报网络图记为
Figure QLYQS_20
；将待预测团伙的团伙-情报网络图记为
Figure QLYQS_23
；其中，
Figure QLYQS_26
解释为已知团伙n犯罪类别i对应倾向的特征值，
Figure QLYQS_19
解释为待预测团伙犯罪类别i对应倾向的特征值，上标
Figure QLYQS_22
表示待预测；根据犯罪倾向预测公式计算待预测团伙的团伙-情报网络图
Figure QLYQS_25
与已知犯罪团伙犯罪类别倾向度：
Figure QLYQS_28
其中，
Figure QLYQS_17
表示犯罪类别i倾向度，
Figure QLYQS_21
表示任意已知团伙犯罪类别i对应倾向的特征值，
Figure QLYQS_24
表示待预测团伙犯罪类别i对应倾向的特征值，上标
Figure QLYQS_27
表示待预测，G表示输入的关系网络图，W与
Figure QLYQS_18
表示可学习权重，b表示标量偏置，exp指以自然常数e为底的指数函数，I表示犯罪类别数量；根据犯罪类别倾向度评分，比较得出待预测团伙的犯罪趋势的预测结果。The gang-intelligence network graph of the known gang obtained in Step 3 is denoted as
Figure QLYQS_20
; The gang-intelligence network graph of the gang to be predicted is recorded as
Figure QLYQS_23
;in,
Figure QLYQS_26
It is interpreted as the characteristic value of the corresponding tendency of the crime categoryi of the known gangn ,
Figure QLYQS_19
It is interpreted as the characteristic value of the corresponding tendency of the gang crime categoryi to be predicted, with the superscript
Figure QLYQS_22
Indicates the group to be predicted; calculates the gang-intelligence network diagram of the group to be predicted based on the criminal tendency prediction formula
Figure QLYQS_25
Propensity to crime categories with known criminal gangs:
Figure QLYQS_28
in,
Figure QLYQS_17
represents the tendency of crime categoryi ,
Figure QLYQS_21
represents the characteristic value of the corresponding tendency of any known gang crime categoryi ,
Figure QLYQS_24
Indicates the characteristic value of the corresponding tendency of the gang crime categoryi to be predicted, with superscript
Figure QLYQS_27
represents the prediction to be made, G represents the input relationship network diagram,W and
Figure QLYQS_18
represents the learnable weight,b represents the scalar bias, exp refers to the exponential function with the natural constant e as the base, and I represents the number of crime categories. According to the crime category tendency score, the prediction result of the crime trend of the gang to be predicted is obtained by comparison.
CN202310151256.1A2023-02-222023-02-22Graph theory-based clue reasoning and intelligence prediction methodPendingCN115858946A (en)

Priority Applications (2)

Application NumberPriority DateFiling DateTitle
CN202310151256.1ACN115858946A (en)2023-02-222023-02-22Graph theory-based clue reasoning and intelligence prediction method
CN202310544005.XACN116578719A (en)2023-02-222023-05-15Clue reasoning and information analysis method based on graph theory

Applications Claiming Priority (1)

Application NumberPriority DateFiling DateTitle
CN202310151256.1ACN115858946A (en)2023-02-222023-02-22Graph theory-based clue reasoning and intelligence prediction method

Publications (1)

Publication NumberPublication Date
CN115858946Atrue CN115858946A (en)2023-03-28

Family

ID=85658662

Family Applications (2)

Application NumberTitlePriority DateFiling Date
CN202310151256.1APendingCN115858946A (en)2023-02-222023-02-22Graph theory-based clue reasoning and intelligence prediction method
CN202310544005.XAPendingCN116578719A (en)2023-02-222023-05-15Clue reasoning and information analysis method based on graph theory

Family Applications After (1)

Application NumberTitlePriority DateFiling Date
CN202310544005.XAPendingCN116578719A (en)2023-02-222023-05-15Clue reasoning and information analysis method based on graph theory

Country Status (1)

CountryLink
CN (2)CN115858946A (en)

Citations (12)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
US20070067211A1 (en)*2005-09-222007-03-22Kaplan Craig AGraphical forecasting interface
CN110458373A (en)*2019-08-222019-11-15电子科技大学成都学院A kind of method of crime prediction and system of the fusion of knowledge based map
CN110609903A (en)*2019-08-012019-12-24华为技术有限公司 Information presentation method and device
US20200096998A1 (en)*2016-05-092020-03-26Strong Force Iot Portfolio 2016, LlcMethods and systems for detection in an industrial internet of things data collection environment with distributed data processing in a marginal network
CN112328801A (en)*2020-09-282021-02-05西南电子技术研究所（中国电子科技集团公司第十研究所）Method for predicting group events by event knowledge graph
CN112365171A (en)*2020-11-162021-02-12广州杰赛科技股份有限公司Risk prediction method, device and equipment based on knowledge graph and storage medium
CN112463983A (en)*2020-12-012021-03-09天翼电子商务有限公司Anti-money laundering criminal identification method based on community division and graph convolution
CN113051440A (en)*2021-04-122021-06-29北京理工大学Link prediction method and system based on hypergraph structure
CN114860854A (en)*2022-05-052022-08-05中国人民解放军国防科技大学 Time-series knowledge graph reasoning method, device and device based on attention mechanism
CN114915468A (en)*2022-05-102022-08-16广州数智网络科技有限公司Intelligent analysis and detection method for network crime based on knowledge graph
CN115221432A (en)*2022-08-022022-10-21武汉烽火技术服务有限公司 A method and device for predicting crime hotspots based on spatiotemporal features
CN115243268A (en)*2022-07-202022-10-25中国联合网络通信集团有限公司 Method, system and readable storage medium for identifying group fraud relationship

Patent Citations (12)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
US20070067211A1 (en)*2005-09-222007-03-22Kaplan Craig AGraphical forecasting interface
US20200096998A1 (en)*2016-05-092020-03-26Strong Force Iot Portfolio 2016, LlcMethods and systems for detection in an industrial internet of things data collection environment with distributed data processing in a marginal network
CN110609903A (en)*2019-08-012019-12-24华为技术有限公司 Information presentation method and device
CN110458373A (en)*2019-08-222019-11-15电子科技大学成都学院A kind of method of crime prediction and system of the fusion of knowledge based map
CN112328801A (en)*2020-09-282021-02-05西南电子技术研究所（中国电子科技集团公司第十研究所）Method for predicting group events by event knowledge graph
CN112365171A (en)*2020-11-162021-02-12广州杰赛科技股份有限公司Risk prediction method, device and equipment based on knowledge graph and storage medium
CN112463983A (en)*2020-12-012021-03-09天翼电子商务有限公司Anti-money laundering criminal identification method based on community division and graph convolution
CN113051440A (en)*2021-04-122021-06-29北京理工大学Link prediction method and system based on hypergraph structure
CN114860854A (en)*2022-05-052022-08-05中国人民解放军国防科技大学 Time-series knowledge graph reasoning method, device and device based on attention mechanism
CN114915468A (en)*2022-05-102022-08-16广州数智网络科技有限公司Intelligent analysis and detection method for network crime based on knowledge graph
CN115243268A (en)*2022-07-202022-10-25中国联合网络通信集团有限公司 Method, system and readable storage medium for identifying group fraud relationship
CN115221432A (en)*2022-08-022022-10-21武汉烽火技术服务有限公司 A method and device for predicting crime hotspots based on spatiotemporal features

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
ANDREW J. COLLINS 等: "Agent-based model of criminal gang formation"*
孙权 等: "金融交易数据驱动的图谱网络智能化欺诈侦测"*

Also Published As

Publication numberPublication date
CN116578719A (en)2023-08-11

Similar Documents

PublicationPublication DateTitle
CN113792937B (en)Social network influence prediction method and device based on graph neural network
CN110334742B (en) A Reinforcement Learning-Based Graph Adversarial Example Generation Method by Adding Fake Nodes for Document Classification
CN111475838B (en)Deep neural network-based graph data anonymizing method, device and storage medium
CN110390017B (en)Target emotion analysis method and system based on attention gating convolutional network
CN109697512B (en) Personal data analysis method and computer storage media based on Bayesian network
CN113220866A (en)Recommendation method based on user social network and product similarity
CN117056970B (en)Privacy feature protection method and system based on graph neural network
CN111814066A (en) Dynamic social user alignment method and system based on heuristic algorithm
CN113395172B (en)Important user discovery and behavior prediction method based on communication network
CN115225373A (en)Network space security situation expression method and device under incomplete information condition
CN115858946A (en)Graph theory-based clue reasoning and intelligence prediction method
CN117194687B (en) Multimedia recommendation method, device, equipment and storage medium
CN116361759B (en) An intelligent compliance control method based on quantitative authority guidance
CN112163170A (en) A method and system for improving social network alignment based on virtual nodes and meta-learning
CN115904464B (en) Core developer identification method, device and equipment based on developer collaboration network
CN118171184A (en) A graph node classification method, device and medium based on security policy
CN111090781A (en) A Link Prediction Method in Dynamic Social Networks
CN116702231A (en) A frequent itemsets mining method for Byzantine untrustworthy multi-party cooperative differential privacy data
CN113065073B (en) An effective path set search method for cities
CN115269867A (en) A Knowledge Graph Node Attribute Completion Method Based on Higher-Order Neighbor Attribute Propagation
CN113886715A (en) A friend recommendation method and system
CN107018027A (en)Link prediction method based on Bayesian estimation and common neighbor node degree
CN111859164A (en)Local structure based microblog network important node discovery method, device and medium
Li et al.Critical Nodes Identification Algorithm Based on ResNet-CBAM
CN117150321B (en)Equipment trust evaluation method and device, service equipment and storage medium

Legal Events

DateCodeTitleDescription
PB01Publication
PB01Publication
SE01Entry into force of request for substantive examination
SE01Entry into force of request for substantive examination
WD01Invention patent application deemed withdrawn after publication
WD01Invention patent application deemed withdrawn after publication

Application publication date:20230328
[8]ページ先頭
©2009-2025 Movatter.jp