CN111444437A - Mobile social network-based criminal and loan unlink relationship closeness cyclic search method - Google Patents

Abstract

The invention discloses a mobile social network-based criminal and loan unlink contact relationship tightness cyclic search method, which is characterized by comprising the following steps of: constructing a mobile social network of the lost people; constructing an interaction attribute matrix of the lost person; establishing a logarithmic weight calculation formula; obtaining a logarithm weighting interaction attribute matrix; then defining a maximum interactive user and a minimum interactive user, and establishing a distance calculation formula and a relationship compactness calculation formula; respectively solving the relationship closeness of each interactive user and the unconnectorized person; and (4) carrying out circular sequence search on the interaction users of the unconnected persons according to the sequence of the relationship closeness from large to small, and providing a circular search algorithm of the unconnected person relationship closeness based on the mobile social network to search clues of the unconnected persons. The invention has practical help and operability for public security tracking of evacuees, bank pursuit of evacuees and financial platform tracking of loan unlink persons, including society and family search of sudden unlink persons.

Description

Mobile social network-based criminal and loan unlink relationship closeness cyclic search method

Technical Field

The invention belongs to the technical field of searching of unlink persons, and particularly relates to a mobile social network-based criminals and loan unlink person relationship tightness cyclic searching method.

Background

At present, literature reports are shown in research and corresponding search and search methods for researching the closeness of the relationship between evacuees caught by police and loan unlink clients of banks and financial institutions and various large loan platforms, such as unlink people in families or society.

Firstly, in the research of related problems such as network density, group relations and the like, Zhang Yongyuan (2011) analyzes the network density, small group and other aspects of characteristics of a social network formed by respondent users in sampled data by utilizing a social network analysis tool UCINET, further analyzes the user relation network characteristics of the question-answer community and provides suggestions for improving the question-answer community. Xu xiong et al (2014) explore the similarity degree of the attention list and the fan list of the microblog users, and consider that the similarity degree represents the relation degree between the users. Zhengqiong and the like (2019) construct a trust evaluation model based on real mobile social interaction data of the Xinlang microblog and considering four influence factors such as social relationship strength and social influence range for embodying relationship interaction, information value and information transmission control force for embodying information interaction and the like. Wangliang et al (2018) propose a method for social task distribution of 'platform-community-user' based on social attributes of group of users participating in crowd-sourcing perception. The method comprises the steps of carrying out dynamic community division clustering on users based on approximation calculation of space-time movement characteristic distribution of the participating users, setting roles of a community organizer and a community slave in a community, introducing models such as a social affinity connection relation network and the like, and constructing a method for distributing a mobile swarm intelligence perception task primary community and selecting secondary users. Introducing a social relation graph model (2018) by the aid of a rock-entering level model and the like, comprehensively considering the relation between users and articles, fusing two important indexes of diversity and relevance by the aid of a linear model, and realizing the algorithm by the aid of a Spark GraphX parallel graph computing frame. Mimao pupils, et al (2018) summarize the main research content of social networking services, give a general definition of social networking services,the Hemian et al (2018) divides the network into four layers of super networks of 'space-time-user-position-category' aiming at the heterogeneity of the network in L BSN and the characteristics of the space-time relationship among users, defines and quantifies the edge weight of a subnet by mining the influence, the hidden relation, the user preference and the node degree information of the users, mines the multi-element association relationship among the users to predict, the army and the Liu administration (2017) construct an online social network multi-dimensional network structure based on two mechanisms of social influence and selection, integrates the topological structures of all dimensional nodes and edges to realize the aggregation of a multi-dimensional social network prediction matrix, carries out link prediction on the network after dimensionality reduction (2017) and the like from the aspect of measuring the cause of the behavior, the influence of the behavior and the behavior expression of the behavior, and 3 aspects of the influence on the social network behavior, and carries out comprehensive analysis on the influence of the social network behavior, wherein the influence of the social network behavior comprises two aspects of the influence of the social behavior and the social network behavior^[]. Real-time dynamic interactions require building collections depending on how often the user contacts are compared to a single interaction. The nodes for recording and identifying the contact will generate different interaction frequencies. Todorov, Claudio (1973) studied the effect of the degree of frequency change of past behavior on the current behavior, finding that when the user does one more frequently, the event is reinforced and more likely to occur in the future. The flood fighting and the like (2018) research the characteristics of online information interaction behavior based on the human behavior dynamics law, firstly, a Hurst index and V statistics of a time interval sequence in a topic discussion are calculated by using a re-standard polar difference method, the Hurst index of each topic time interval sequence is larger than 0.5, and a V statistics curve shows an ascending trend, so that the sequence has self-similarity and long-range related fractal characteristics; then, the relation between the turning point of the statistic trend of the sequence V and the Hurst index is analyzed, and the sequence with the larger Hurst index is found, the later the turning point is, namely the non-cyclic period of the sequence is longer. Research results show that the longer the interaction time interval, the later the next interaction start time. Zhang (2001) byInvite 320 adults in china to randomly visit 8 laboratories and read the information of 2 colleagues and distribute the prize. The 8 experimental conditions are different in relation type, frequency of past interaction between colleagues and possibility of future interaction, and the result shows that the past interaction and the future interaction between colleagues have great influence on the allocation decision by researching how the interaction frequency influences the bonus allocation decision. And the interactive happiness level is difficult to measure directly, and the learner adopts the interactive duration to reflect the psychological state in the interactive process. Kahanda and Neville (2009) used online community interaction data within a certain university on a facebook to explore how interactions between users affect user relationships, and thought that interactions between users play a very important role in maintaining closeness of user relationships. Phithakkitnukoo and Smireda (2016) analyze daily activities and interpersonal relationship characteristics of users through mobile call data, measure position information through types, differences and positions of the user movements, and measure interpersonal relationship by measuring call frequency and call time of each time. It is found that the stronger the relationship strength between users, the closer their respective measurement metrics are, the more similar the geographic location movement information is. Jujuchunhua et al (2016) measure how close a user relationship is by how often and for how long the user interacts. Korean loyalty et al (2016) studied the passiveness and activeness of user interaction and proposed a directed relationship closeness degree calculation method. Zhang Xiang et al (2006) propose a fuzzy support vector machine method based on compactness. The relationship between the samples in the class is described by the compactness between the samples, and the compactness between the samples is measured by the size of the minimum spherical radius surrounding the samples in the same class. Experimental results show that compared with a traditional support vector machine method and a fuzzy support vector machine method based on the relation between a sample and a class center, the fuzzy support vector machine method based on the compactness has better anti-noise performance and classification capability. Wanlianxi et al^[10](2012) Explaining the concept of microblog user relation mining according to the characteristics of microblog users; then, two important research contents mined according to the relationship of the microblog users are taken as main lines to respectively carry out detailed introduction and key user identification on the community analysis of the microblog users andand (6) analyzing.

The search of the lost people plays an important role in various fields:

(1) in public security, it is often reported that a criminal runs away after a crime and then recovers after more than ten years or even two decades as if the criminal evaporates.

(2) On the bank side, some businesses cheat the bank and some roll money escapes.

(3) In the aspect of financial platform, some enterprises or borrowers may flee for a loan (or loan) after the loan is successful.

(4) In real estate, a common developer rolls over to run away, damaging house purchasers.

From this knowledge, tracking and locating an unlink person is critical! Because the mobile social data contains a large amount of information of the lost people, how to construct a mobile social network of the lost people is to analyze the social circle of the lost people through the data information contained in the mobile social network nodes and analyze the correlation degree and the relationship closeness degree between the lost people and other people in the society through the historical interaction data of the mobile social network of the lost people so as to track and search the lost people, which is one of the research directions of technicians in the field.

Disclosure of Invention

The invention mainly aims to overcome the defects and shortcomings of the prior art and provide a circulating searching method for the relationship closeness of criminals and loan unlink persons based on a mobile social network.

In order to achieve the purpose, the invention adopts the following technical scheme:

a criminal and loan unlink contact relationship closeness circular searching method based on a mobile social network comprises the following steps:

constructing a mobile social network of the lost people;

based on a mobile social network of the lost contacts, considering interaction attributes of active interaction frequency, passive interaction frequency, active interaction duration, passive interaction duration, interaction interval time and interaction intensity, and constructing an interaction attribute matrix of the lost contacts;

carrying out normalization processing on the interaction attribute matrix to obtain a normalized interaction attribute matrix;

establishing a logarithmic weight calculation formula, and then multiplying the logarithmic weight by the normalized interaction attribute matrix correspondingly to obtain a logarithmic weight interaction attribute matrix;

in the logarithmic weighting interaction attribute matrix, solving the average value of each weighting attribute group, wherein the maximum average value corresponds to the maximum interaction user of the loss contact person, and the minimum average value corresponds to the minimum interaction user of the loss contact person;

respectively calculating the distance between each interactive user and the large interactive user and the distance between each interactive user and the minimum interactive user based on the average value, wherein the distance represents the relationship closeness between each interactive user and the unconnectorized person;

correspondingly obtaining the maximum interactive user, the 2 nd large interactive user and the 3 rd large interactive user of the unconnectorized according to the sequence of the relationship closeness from large to small, and so on, and finally obtaining the minimum interactive user;

and according to the closeness of the relationship of the unconnected persons from large to small, the interactive users are searched in a circulating sequence correspondingly so as to find out the clues of the unconnected persons.

As a preferred technical solution, the constructing of the mobile social network of the unconnector specifically includes:

in the considered time period T, all the contacts directly calling out and directly calling in the lost person are called as interaction users of the lost person, the lost person and all the interaction users of the lost person and/or the lost person are represented by nodes, then the lost person and the interaction users of the lost person are connected by a directed arrow, and the lost person points to all the interaction users of the lost person, so that the mobile social network of the lost person is obtained.

As a preferred technical solution, the interaction attribute matrix is:

X＝(x_ij)_n×mis defined as:

wherein x_ijFor the user pair c_iI is more than or equal to 1 and less than or equal to n, and j is more than or equal to 1 and less than or equal to m;

changing X to (X)_ij)_n×mThe interaction attribute matrix is called because the matrix is constructed based on the mobile social network interaction data of the losers.

As a preferred technical solution, then, normalization processing is performed on the interaction attribute matrix by using a certain normalization processing method, so as to obtain a normalized interaction matrix Y ═ (Y ═ Y)_ij)_m×n。

As a preferred technical solution, the log-weighted interaction attribute matrix specifically includes:

let the interaction attribute weight vector of the node be W ═ ω₁,ω₂,L,ω_mWhere ω is_jAccording to the formula

To calculate, for the normalized matrix Y ═ Y (Y)_ij)_n×m(ii) a Order to

z_ij＝ω_jy_ij

The matrix Z is (Z)_ij)_n×mA log-weighted interaction attribute matrix;

if the logarithm weighted interaction attribute matrix Z is equal to (Z)_ij)_n×mRepresents an inline vector Z ═ Z₁,Z₂,L,Z_n) Then Z is_i＝(z_i1,z_i2,L,z_in) For the attribute of the ith interactive user, 1 ≦ i ≦ n, so z_ij＝ω_jy_ijIs the jth weighted interaction attribute for the ith user.

As a preferred technical scheme, the maximum interactive user and the minimum interactive user specifically are:

taking the logarithm-weighted interaction attribute matrix Z as (Z)_ij)_n×mEach row Z of_iSelf-squaring mean

Namely, it is

I is more than or equal to 1 and less than or equal to n, which is called as weighted interaction attribute mean;

comparing Z ═ Z_ij)_n×mMean value Z of middle n weighted interaction attributes_iThe largest weighted interaction attribute mean is recorded as

Then

Then the minimum weighted interaction attribute mean is recorded as

Then

Changing Z to (Z)_ij)_n×mMean of the largest weighted interaction attributes

The maximum interactive user called the loser is v_maxIndicating that the corresponding interaction attribute group is called the maximum interaction attribute group, using Z_maxRepresents;

changing Z to (Z)_ij)_n×mWeighted interaction attribute mean of medium minimum

The corresponding interactive user is called the minimum interactive user of the lost contact person, and is expressed by v_minRepresenting, with Z, the corresponding set of interaction attributes, called the minimum set of interaction attributes_minAnd (4) showing.

As a preferred technical solution, the distance calculation formula is as follows:

in an unconnectorized mobile social network, assume Z ═ Z (Z)_ij)_n×mIs a logarithmically weighted interaction attribute matrix for which any two interacting users b_iAnd b_jTheir weighted interaction attribute groups are respectively Z_iAnd Z_jEstablishing any two interactive users b of the lost person_iAnd b_jFrom b to b_iTo b_jThe distance calculation formula of (c) is as follows:

d (Z)_i,Z_j) Called "Pangsulin-distance 1", where

And

respectively represent weighted interaction attribute groups respectively are Z_iAnd Z_jAverage value of (Z)_i-Z_j)^TIs Z_i-Z_jI is not less than 1, n is not less than j, i is not equal to j, ∑ is a weighted reciprocal matrix Z ═ Z (Z)_ij)_n×mOf the user, then, any interactive user v of the loser_iWith maximum interaction user b_maxAnd minimum interactive user v_minThe distances between are respectively:

as a preferred technical scheme, the relationship compactness calculation model is as follows:

formula r_iReferred to as "Pangsulin-relationship compactness of 1", wherein

And

it cannot be simultaneously 0 at the same time,

r_irepresenting the closeness of the relation between the unconnector and the ith interactive user, and r is more than or equal to 0_i≤1；r _i0, the relationship closeness between the unconnector and the ith interactive user is 0; r is_i1, the closeness of relationship between the unconnector and the ith interactive user is 1.

As a preferred technical scheme, through an unconnectorized-person relationship compactness calculation model, an interactive user relationship compactness set R ═ R of the unconnectorized person can be calculated and obtained₁,r₂,...,r_m}；

Setting the interactive user relationship compactness set R as { R ═ R₁,r₂,...,r_mElement r in (1)₁,r₂,...,r_mSorting is carried out in the order from big to small, and the element sequence of the sorting is assumed to be r'₁,r′₂,...,r′_m(r′₁,r′₂,...,r′_mIs r₁,r₂,...,r_mA combination of) then r'₁≥r′₂≥...≥r′_m。

As an optimal technical scheme, a relation closeness cyclic search algorithm is adopted to search the unconnected person, and the method specifically comprises the following steps:

step 1: gathering n mobile social users v of an unconnected person u₁,v₂,...,v_nThe interaction data of (2);

step 2: establishing a set of all user pairs C ═ { C) of the lost person u₁,c₂,...,c_mAnd establishing an interaction attribute matrix X of the user pairs (X ═ X)_ij)_m×n；

And 3, step 3: changing the interaction attribute matrix X to (X)_ij)_m×nCarrying out normalization processing to obtain a normalized interaction attribute matrix Y ═ Y_ij)_m×n；

And 4, step 4: calculating the logarithmic weight of each node interaction attribute to obtain a node interaction attribute weight set W ═ ω { (ω) }₁,ω₂,...,ω_n}；

And 5, step 5: establishing a logarithm weighting interactive attribute matrix Z ═ (Z ═_ij)_n×m；

And 6, step 6: calculating Z_maxAnd Z_min；

And 7, step 7: respectively calculating each interactive user v_iAttribute group Z_iAnd Z_maxAnd Z_minA distance d between_i ^maxAnd d_i ^min；

And 8, step 8: calculating the closeness of the unconnectorized relation:

obtaining a relation compactness set R ═ R of the unconnector₁,r₂,...,r_n}；

Step 9: set R-R of closeness of relationship of all interaction users of the lost contact person₁,r₂,...,r_nElement r in (1)₁,r₂,...,r_nSequencing the elements in the descending order to obtain the sequence of the sequenced elements of r'₁,r′₂,...,r′_nR 'is satisfied'₁≥r′₂≥...≥r′_n，r′₁,r′₂,...,r′_nIs r₁,r₂,...,r_nTo establish a new ordered set R '═ { R'₁,r′₂,...,r′_n}；

Step 10: from i to 1, i is more than or equal to 0 and less than or equal to n, and the relation compactness R ' is taken out of the ordered set R ' in sequence '_iPropose corresponding interactive users v_i(ii) a If can pass v_iIf the unconnected person u is found, the algorithm is ended; otherwise, the next step is carried out;

and 11, step 11: letting i equal to i +1, and executing the step 10 in a circulating manner; when i is equal to n, the algorithm ends.

Compared with the prior art, the invention has the following advantages and beneficial effects:

1. the invention provides a method for searching evasion of public security pursuit, loan unlink persons of financial institutions, unlink persons of families, society and the like by adopting a method based on established mathematical model and algorithm design for the first time;

2. the invention establishes a log-weighted interaction attribute matrix by utilizing the interaction data information of the mobile social network of the lost contact person for the first time. Defining the maximum interactive user and the minimum interactive user of the lost person by using the maximum average value and the minimum average value of the interactive attribute group; correspondingly defining the maximum relationship compactness and the minimum relationship compactness;

3. the invention establishes a logarithmic weight calculation formula of an interaction attribute matrix of an unconnectorized person for the first time:

wherein j is more than or equal to 1 and less than or equal to m, and omega is more than or equal to 0_j≤1，

4. The invention establishes a distance formula for the first time:

the distance formula is named as Pangsulin-1 st distance by the Chinese pinyin "Pangsulin" of the inventor.

5. The invention establishes a relationship compactness calculation model of the unconnectorized person for the first time:

the closeness of the relation is named by the Chinese pinyin 'Pangsulin' of the inventor, namely 'Pangsulin-1 st relation closeness'

6. The method is characterized in that a cyclic searching method for the closeness of the relation of the unconnectorized based on the mobile social network is designed based on a logarithmic weight calculation formula innovatively established by the method, a logarithmic weighting interaction attribute matrix of the unconnectorized, a Pangsulini-1 st distance and a Pangsulini-1 st relation closeness calculation formula for the first time;

7. the mobile social network-based offline contact relation closeness circular searching method can be used for helping evacuees caught by public security, loan offline contacts of financial institutions, family offline contacts including social offline contacts and the like to track, search and search.

Drawings

FIG. 1 is a flow chart of the method of the apparatus of the present invention.

FIG. 2 is a diagram of a mobile social network of an unconnector with only directly interacting users according to the present invention.

FIG. 3 is a diagram of relationship closeness, corresponding interactive users and search sequence according to the present invention.

Detailed Description

The present invention will be described in further detail with reference to examples and drawings, but the present invention is not limited thereto.

Examples

The embodiment researches the mobile social network of the lost contact person based on the mobile social network based on a relation closeness circular searching method of criminals, loan lost contacts and the like, and aims to provide public security, banks, financial institutions, real estate, insurance, various loan big platforms and the like for searching and tracking evacuees, bank lost contacts, lost contact platforms, family lost contacts, social lost contacts and the like. The invention mainly studies such a search situation: the method comprises the steps of establishing a corresponding mobile social network according to historical mobile communication data of a lost contact person and a data flow track of the lost contact person, analyzing the first generation time, the occurrence frequency, the node flow direction (flowing in or out), the interaction time, the adjacent two-time interaction interval time, the node interaction strength, the closest contact person and the like of each network node by means of the mobile social network, tracking and searching the nodes of the mobile social network of the lost contact person according to the relationship closeness of each node and the lost contact person, and searching a 'route running' clue of the lost contact person.

The criminals and loan unlink persons in the application comprise family and social unlink persons and the like, the unlink persons refer to criminals caught by public security, money carrying fleeing persons caught by banks, loan unlink persons caught by financial platforms, platform unlink persons caught by financial departments, and various suddenly-missing unlink persons found by families, society and the like.

As shown in fig. 1, the method of the present invention comprises the steps of:

s1, constructing a mobile social network of the lost contact person, wherein the method comprises the following steps:

in an investigation time period T (T > 0 is an integer), all contacts directly calling out and directly calling in from the lost person are collectively called as interaction users of the lost person, the lost person and all the interaction users of the lost person are represented by nodes, then the lost person and all the interaction users of the lost person are connected by a directed arrow, and the lost person points to all the interaction users of the lost person, so that a mobile social network of the lost person can be obtained, as shown in FIG. 2.

In the mobile social network of the lost contact person, the node of the lost contact person is called a main node, nodes of all interactive users of the lost contact person are called sub-nodes, and the mobile social network of the lost contact person is a single-layer network diagram formed by the main node, the sub-nodes and an arrow pointing to the sub-nodes from the main node. In fig. 2, there is one master node, 27 child nodes, indicating that the loser has 27 interactive users.

S2, constructing an interaction attribute matrix:

in the mobile social network of the lost contact person, the lost contact person is represented by u, and an interaction user who has an interaction behavior with the lost contact person is represented by v; if an interactive behavior exists between the loser u and the interactive user v, representing that c is (u, v); if the loser u actively interacts with the interactive user v, O is used_uvRepresenting an active interaction frequency; if the person u with lost connection passively interacts with the interactive user v, O is used_vuRepresenting a passive interaction frequency; reuse I_uvThe interactive duration of the active interaction between the lost contact person u and the interactive user v is represented and is called as active interactive duration; by means of I_vuRepresenting the interaction duration of the active interaction between the lost person u and the interactive user v, which is called the passive interaction duration; with R_uvRepresenting the activity intensity of the interaction of the loser u with the interactive user v.

If an observation time interval when the user u and the user v interact is divided into a plurality of time intervals, if the observation time interval is divided into k time intervals, the interaction node sequence o is used₁,o₂,...,o_kTo show thatMiddle o_iThe total number of the interaction between the two from the moment i-1 to the moment i each day.

Definition 1: if an interactive node sequence o exists between the user u and the user v in a certain observation period₁,o₂,...,o_kThen the strength of the interaction node between the two users is defined as

Wherein theta is_iIs a time period [ i-1, i]Weight of (0) theta_i≤1，

Weight θ_iThe interactive behaviors in different time periods have different values and reflect different characteristics of the interactive behaviors in different time periods, and then the value is changed according to theta_iThe difference of the interaction strength can be reflected. For example,

example 1 assume a master node user u and a child node user v_iThe total number of interactions in 24 periods of a day is respectively as follows: {20, 13, 14,0,1,3,2, 14, 45, 53, 66, 72, 73, 53, 75, 88, 85, 87, 89, 96, 106, 197, 199, 49}. If 8:00 am to 18:00 pm and 18:00 pm to 23:00 pm and 23:00 pm to 8:00 am are regarded as working time, leisure time and sleeping time respectively, the weight of the working time is theta₁0.5, weight of off-hours θ₂0.3, weight of off-hours θ₃When the sum equals 0.2, the master node user u and the child node user v are known from equation (1)_iThe strength of the interaction node is:

defining 2, referring various interaction behaviors between the lost contact person u and the interaction user v, such as active interaction frequency, passive interaction frequency, active interaction time length, passive interaction time length, interaction node strength and the like, to be referred to as node interaction attributes.

In absence ofIn a mobile social network of contacts, assuming that a contact lost person u has n interactive users, the set of interactive users may be represented as V ═ { V ═ V₁,v₂,...,v_n}. Thus, the number pairs formed by the loser u and its n interactive users can be represented as c_i＝(u,v_i) I is more than or equal to 1 and less than or equal to n, the same applies below. Then C ═ C₁,c₂,...,c_nRepresents the set of all user pairs in the mobile social network. Suppose each user pair c_i＝(u,v_i) There are m node interaction attributes.

Definition 3. interaction attribute matrix X ═ X (X) of the loser u and its n-bit interaction users_ij)_n×mIs defined as:

wherein x_ijFor the user pair c_iThe node interaction attribute is that i is more than or equal to 1 and less than or equal to n, j is more than or equal to 1 and less than or equal to m, and the same is used below.

From definition 3, X_i＝[x_i1x_i2L x_im]Is that c_iThe m nodes of (1) interacting with the attributes. Since the node interaction attribute does not have uniform calculation and statistical functions, the interaction attribute matrix X ═ X (X)_ij)_n×mThere is no arithmetic function, and therefore, it is necessary to set (X) to the interaction attribute matrix X_ij)_n×mAnd (6) carrying out normalization processing. When the temperature of the molten metal is higher than the set temperature,

then, normalization is performed as follows:

a normalized interaction matrix Y ═ Y is then obtained (Y)_ij)_m×n. But if

No normalization is required.

S3, providing a logarithmic weight calculation method:

before a logarithmic weight calculation formula is given, the mean and variance calculation processing needs to be carried out on the normalized interaction matrix (4) firstly. I.e. for the normalized interaction matrix Y ═ Y_ij)_m×nRespectively (i.e. normalized values of node interaction attributes) to obtain a mean value mu_jVariance, variance

And standard deviation σ_jThe following were used:

wherein j is more than or equal to 1 and less than or equal to m.

Order to

α′_j＝μ_j+|lgσ_j| (8)

Wherein j is more than or equal to 1 and less than or equal to m, and then α'_jNormalized, we establish a logarithmic interaction weight calculation formula for the node interaction attributes as follows:

wherein j is more than or equal to 1 and less than or equal to m, and omega is more than or equal to 0_j≤1，

From the formula (9), there is a weight ω_jNot equal to 0 (j is more than or equal to 1 and less than or equal to m). This means that the weight of the node interaction attribute established herein is not 0, which indicates that the interaction behavior of the loser in any past time period is very important and is not negligible 0. Let us say omega_jCalled the log weight of the jth node user interaction attribute.

S4, constructing a log-weighted interaction attribute matrix:

let the node interaction attribute weight vector be W ═ ω₁,ω₂,L,ω_mWhere ω is_jCalculated according to equation (9) or equation (9). For normalized matrix Y ═ Y_ij)_n×m。

Definition 4. order

z_ij＝ω_jy_ij(10)

The matrix Z is (Z)_ij)_n×mThe interaction attribute matrix is logarithmically weighted.

If the logarithm weighted interaction attribute matrix Z is equal to (Z)_ij)_n×mRepresents an inline vector Z ═ Z₁,Z₂,L,Z_n) Then Z is_i＝(z_i1,z_i2,L,z_in) The attribute of the ith interactive user (i is more than or equal to 1 and less than or equal to n). Thus, z_ij＝ω_jy_ijIs the jth weighted interaction attribute for the ith user.

S5, relation compactness and distance calculation formula:

s5.1, defining maximum and minimum interactive users:

log-weighted interaction attribute matrix Z ═ Z (Z)_ij)_n×mThe row of (1) is the interaction attribute value of one user of the loser to the node. We will determine the largest, smallest interactive users of the lost person in the following way:

step 1: taking the logarithm-weighted interaction attribute matrix Z as (Z)_ij)_n×mEach row Z of_i(1. ltoreq. i. ltoreq. n, the same applies hereinafter) self-averaging the squares

Namely, it is

Referred to as the weighted interaction attribute mean.

Step 2: comparing Z ═ Z_ij)_n×mMean of n weighted interaction attributes

The largest weighted interaction attribute mean is recorded as

Then

Then the minimum weighted interaction attribute mean is recorded as

Then

Definition 5. converting Z to (Z)_ij)_n×mMean of the largest weighted interaction attributes

The maximum interactive user called the loser is v_maxIndicating that the corresponding interaction attribute group is called the maximum interaction attribute group, using Z_maxAnd (4) showing.

Definition 6. converting Z to (Z)_ij)_n×mWeighted interaction attribute mean of medium minimum

The corresponding interactive user is called the minimum interactive user of the lost contact person, and is expressed by v_minRepresenting, with Z, the corresponding set of interaction attributes, called the minimum set of interaction attributes_minAnd (4) showing.

S5.2, defining the maximum and minimum relationship compactness:

based on the maximum interactive users and the minimum interactive users, the maximum relationship closeness users and the minimum relationship closeness users can be further defined.

And 7, calling the maximum interactive user of the unconnectorized as the maximum relationship closeness user. Log-weighted interaction attribute matrix Z-Z (Z-Z) for the lost person_ij)_n×mIn, the weighted interaction attribute group Z corresponding to the maximum interaction user_maxReferred to as the maximum interaction property group.

Definition of8. The minimum interactive user of the unconnectorized person is referred to as the maximum relationship affinity user. Log-weighted interaction attribute matrix Z-Z (Z-Z) for the lost person_ij)_n×mThe weighted interaction attribute group Z corresponding to the minimum interaction user_minReferred to as the minimum interaction property group.

S5.3, a distance calculation formula:

in an unconnectorized mobile social network, assume Z ═ Z (Z)_ij)_n×mIs a logarithmically weighted interaction attribute matrix for which any two interacting users b_iAnd b_jTheir weighted interaction attribute groups are respectively Z_iAnd Z_jThen we establish any two interactive users b of the lost person_iAnd b_jFrom b to b_iTo b_jThe distance calculation formula of (c) is as follows:

we call equation (11) as "Pangsulin-distance 1". Wherein

And

respectively represent weighted interaction attribute groups respectively are Z_iAnd Z _j1 ≦ i, j ≦ n, i ≠ j, ∑ is the weighted cross-membership matrix Z ═ Z (Z)_ij)_n×mThe covariance matrix of (2). Thus, any interactive user v of the lost person_i(1 ≦ i ≦ n) and maximum interaction user b_maxAnd minimum interactive user v_minThe distances between are respectively:

s5.4, constructing a relationship compactness calculation model:

in a mobile social network, the magnitude of closeness of relationship between an unconnector and its interacting user represents the person the unconnector is most likely to contact after an unconnection. Normally, the greater the closeness of relationship, the more likely the loser is to contact the interactive users after the loss of contact; the smaller the closeness of relationship, the less likely the loser will contact the interactive users after the loss of contact. The search for an unconnectorized person may therefore be tracked by the magnitude of closeness of relationship of their interacting users.

Establishing an interactive user relationship closeness calculation model of the unconnected person as follows:

equation (14) is referred to as "Pangsulin-tightness of relationship 1". Wherein

And

cannot be 0 at the same time. In fact, if

And

and 0 at the same time, the logarithm-weighted interaction attribute matrix Z is (Z)_ij)_n×mIs the same for each element or row vector, in either case, Z ═ Z (Z_ij)_n×mAre both degradation matrices, which in reality will not be the case, so both degradation cases are not considered.

In the formula (14), r_i(0≤r_i≦ 1) represents the closeness of relationship between the unconnector and its ith interactive user. r is_i0, the relationship closeness between the unconnector and the ith interactive user is 0; r is_i1, the closeness of relationship between the unconnector and the ith interactive user is 1.

Through calculation of the unconnectorized-person relationship compactness calculation model (14), an interactive user relationship compactness set R ═ { R ═ R of unconnectorized persons can be obtained₁,r₂,...,r_m}。

Setting the interactive user relationship compactness set R as { R ═ R₁,r₂,...,r_mElement r in (1)₁,r₂,...,r_mSorting is carried out in the order from big to small, and the element sequence of the sorting is assumed to be r'₁,r′₂,...,r′_m(r′₁,r′₂,...,r′_mIs r₁,r₂,...,r_mA combination of) then r'₁≥r′₂≥...≥r′_m。

S6, searching the unconnectorized person by using a relation compactness circular search algorithm:

step 1: gathering n mobile social users v of an unconnected person u₁,v₂,...,v_nThe interaction data of (2);

step 2: establishing a set of all user pairs C ═ { C) of the lost person u₁,c₂,...,c_mAnd (3) establishing an interaction attribute matrix X ═ X of the user pairs_ij)_m×n；

And 3, step 3: the interaction attribute matrix X is expressed by equation (4) as (X)_ij)_m×nNormalization processing is carried out, and the processing method comprises the following steps:

(1) when in use

Then, normalization is performed as follows:

thus, a normalized interaction-attribute matrix Y ═ Y (Y) is obtained_ij)_m×n；

(2) When in use

Then, the interaction attribute matrix X ═ X (X)_ij)_m×nNo normalization is required.

And 4, step 4: calculating the weight of each node interaction attribute by using formula (9) (exponential weight formula) or (9) (logarithmic weight formula) to obtain a node interaction attribute weight set W ═ ω { (ω) }₁,ω₂,...,ω_n}；

And 5, step 5: from equation (10), a log-weighted interaction attribute matrix Z ═ Z is established_ij)_n×m；

And 6, step 6: calculating Z from Definitions 5 and 6, respectively_maxAnd Z_min；

And 7, step 7: calculating each interactive user v from equation (12) and equation (13) respectively_iAttribute group Z_iAnd Z_maxAnd Z_minThe distance between

And

and 8, step 8: calculating the unconnectorized relationship closeness according to equation (14):

(wherein

And

not all can be 0) to get the unconnectorized relation compactness set R ═ R₁,r₂,...,r_n}。

Step 9: set R-R of closeness of relationship of all interaction users of the lost contact person₁,r₂,...,r_nElement r in (1)₁,r₂,...,r_nSequencing the elements in the descending order to obtain the sequence of the sequenced elements of r'₁,r′₂,...,r′_nSatisfy r₁′≥r₂′≥...≥r′_n(r′₁,r′₂,...,r′_nIs r₁,r₂,...,r_nA combination of R) to create a new ordered set R '═ { R'₁,r′₂,...,r′_n}；

Step 10: sequentially taking out the relationship compactness R from the ordered set R' from i ═ 1 (i is more than or equal to 0 and less than or equal to n)_i' propose a corresponding interactive user v_i(ii) a If can pass v_iIf the unconnected person u is found, the algorithm is ended; otherwise, the next step is carried out;

step 12: making i equal to i +1, and executing the step 11 circularly; when i is equal to n, the algorithm ends.

The following is a further description of a specific application example:

suppose in the mobile social network of the unconnector u, the unconnector u has 12 interaction users A, B, C, D, E, F, G, H, I, J, K, L in total in a certain important monitoring period, and 6 interaction attributes between the unconnector u and the interaction users v are considered, namely active interaction frequency O_uvPassive interaction frequency O_vuActive interaction duration L_uvDuration of passive interaction L_vuInteraction interval time I_uvInteraction intensity R_uvAnd collecting relevant data as shown in a matrix (16) (note: the matrix data is generated by random numbers and is used for verifying algorithm), and establishing an interaction attribute matrix X-X (X)_ij)_m×nComprises the following steps:

the attributes are as follows: o is_uvO_vuL_uvL_vuI_uvR_uv

Then, according to the formula (4), the interaction attribute matrix X is (X)_ij)_m×nCarrying out normalization processing to obtain a normalized interaction matrix Y ═ Y_ij)_m×nComprises the following steps:

attribute O_uvO_uvL_uvL_vuI_uvR_uv

The logarithmic interaction weight is calculated by equation (9):

ω′＝[0.4045 0.0645 0 0.4823 1 0.989](18)

the exponential interaction weight omega' formula (10) is substituted, and a logarithm weighting interaction attribute matrix is obtained as follows:

attribute O_uvO_vuL_uvL_vuI_uvR_uv

In the log-weighted interaction attribute matrix, the individual weighted attribute groups are averaged (see table 1). As seen from the mean of the respective weighted attribute groups of Table 1, Z_max0.3821 and Z_min0.174; and Z_maxCorresponding to interactive users H, Z_minCorresponds to interactive user B, so H and B are the maximum interactive user and the minimum interactive user, respectively, of the loser u. Then, using equations (12) and (13), the distance between each interactive user and the large interactive user H and the distance between the minimum interactive user B are obtained, respectively, as shown in table 1. And then, solving the relationship closeness of each interactive user and the unconnectorized u according to a formula (14), and finally, sequencing the relationship closeness from large to small (see table 1).

TABLE 1 closeness of relationship and ordering of unconnected persons u

According to the relationship closeness calculated by the formula (14), the maximum interactive user, the 2 nd maximum interactive user and the minimum interactive user of the unconnector u can be obtained according to the values of the relationship closeness, the sequence of the interactive users is shown as H, F, C, D, E, L, A, G, J, K, I and B in the table 1, in order to see the sequence of the relationship closeness and the sequence of the corresponding interactive users more clearly, a sequence corresponding table of the relationship closeness and the corresponding interactive users is constructed, and the sequence corresponding table is shown in the table 2.

TABLE 2 relationship closeness and Interactive user sequence correspondence table (based on logarithmic weighting)

And then, according to the closeness of the relation of the unconnected person u from large to small, the interactive users are sequentially searched. The searching method comprises the following steps:

(1) extracting the maximum relationship compactness r 'from Table 2'₁1 corresponds to the maximum interactive user H. Searching relevant information of the unlink person u from H, and if the unlink person u can be searched, finishing the algorithm; if the search is not available, switching to the following (2);

(2) extracting the relation compactness r 'of ranking 2 from the table 2'₂0.990 corresponding maximum interactive user F. Searching the related information of the unlink person u from the F, and if the related information can be searched, finishing the algorithm; if the search is not available, switching to the following step (3);

(3) extracting the relation compactness r 'of ranking 3 from the table 2'₃0.9826 corresponding maximum interactive user C. Searching the related information of the unlink person u from the C, and if the related information can be searched, finishing the algorithm; if the search is not available, switching to the following (4);

(4) extracting the relation compactness r 'of ranking 4 from the table 2'₄0.9580 for the maximum interactive user D. Searching the related information of the unlink person u from the D, and if the related information can be searched, finishing the algorithm; if the search is not available, switching to the following (5);

(5) extracting the relation compactness r 'of ranking 5 from the table 2'₅0.8604 for the maximum interactive user E. Searching the related information of the unlink person u from E, and if the related information can be searched, finishing the algorithm; if the search is not available, switching to the following (6);

(6) extracting the relation compactness r 'of ranking 6 from the table 2'₆0.7846 corresponding maximum interactive user L, searching relevant information of the person u lost connection from L, if the relevant information can be searched, finishing the algorithm, and if the relevant information cannot be searched, turning to the following step (7);

(7) extraction of the 7 th order from Table 2Density r'₇0.5882 corresponding maximum interactive user a. Searching the related information of the unlink person u from the A, and if the related information can be searched, finishing the algorithm; if the search is not available, switching to the following (8);

(8) extracting relation compactness r 'of ranking 8 from table 2'₈0.4502 for the maximum interactive user G. Searching the relevant information of the unlink person u from G, and if the unlink person u can be searched, finishing the algorithm; if the search is not available, switching to the following (9);

(9) extracting the relation compactness r 'of ranking 9 from the table 2'₉0.4020 for the maximum interactive user J. Searching the relevant information of the unlink person u from J, and if the unlink person u can be searched, finishing the algorithm; if the search is not available, switching to the following (10);

(10) extracting relationship compactness r 'of ranking 10 from Table 2'₁₀0.3317 corresponding maximum interactive user K. Searching the relevant information of the unlink person u from the K, and if the unlink person u can be searched, finishing the algorithm; if the search is not available, switching to the following (11);

(11) extracting the relationship compactness r 'of ranking 11 from Table 2'₁₁0.0856 corresponding maximum interactive user I. Searching the related information of the unlink person u from the I, and if the information can be searched, finishing the algorithm; if the search is not available, switching to the following (12);

(12) extracting minimum relationship compactness r 'from Table 2'₁₂Maximum interactive user B corresponding to 0. Searching the related information of the unlink person u from the B, and if the related information can be searched, finishing the algorithm; if the search is not available, the algorithm is also ended.

The relationship closeness, corresponding interactive users and search sequence of the unlink person u are shown in fig. 3. In fig. 3, (a, 7) indicates that the search order of the interactive user a is ranked at the 7 th position. (B, 12) shows that the search order of the interactive user B is ranked at the 12 th position. And so on for others.

The above embodiments are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments, and any other changes, modifications, substitutions, combinations, and simplifications which do not depart from the spirit and principle of the present invention should be construed as equivalents thereof, and all such changes, modifications, substitutions, combinations, and simplifications are intended to be included in the scope of the present invention.

Claims (10)

1. A criminal and loan unlink contact relation closeness circular searching method based on a mobile social network is characterized by comprising the following steps:

constructing a mobile social network of the lost people;

based on a mobile social network of the lost contacts, considering interaction attributes of active interaction frequency, passive interaction frequency, active interaction duration, passive interaction duration, interaction interval time and interaction intensity, and constructing an interaction attribute matrix of the lost contacts;

carrying out normalization processing on the interaction attribute matrix to obtain a normalized interaction attribute matrix;

establishing a logarithmic weight calculation formula, and then multiplying the logarithmic weight by the normalized interaction attribute matrix correspondingly to obtain a logarithmic weight interaction attribute matrix;

in the logarithmic weighting interaction attribute matrix, solving the average value of each weighting attribute group, wherein the maximum average value corresponds to the maximum interaction user of the loss contact person, and the minimum average value corresponds to the minimum interaction user of the loss contact person;

respectively calculating the distance between each interactive user and the large interactive user and the distance between each interactive user and the minimum interactive user based on the average value, wherein the distance represents the relationship closeness between each interactive user and the unconnectorized person;

correspondingly obtaining the maximum interactive user, the 2 nd large interactive user and the 3 rd large interactive user of the unconnectorized according to the sequence of the relationship closeness from large to small, and so on, and finally obtaining the minimum interactive user;

and according to the closeness of the relationship of the unconnected persons from large to small, the interactive users are searched in a circulating sequence correspondingly so as to find out the clues of the unconnected persons.

2. The mobile social network-based criminal and loan unlink relationship closeness circular search method according to claim 1, wherein the method for constructing the mobile social network of the unlink is specifically as follows:

in the considered time period T, all the contacts directly calling out and directly calling in the lost person are called as interaction users of the lost person, the lost person and all the interaction users of the lost person and/or the lost person are represented by nodes, then the lost person and the interaction users of the lost person are connected by a directed arrow, and the lost person points to all the interaction users of the lost person, so that the mobile social network of the lost person is obtained.

3. The mobile social network-based criminal and loan unlink relationship closeness circular search method according to claim 1, wherein the interaction attribute matrix is:

X＝(x_ij)_n×mis defined as:

wherein x_ijFor the user pair c_iI is more than or equal to 1 and less than or equal to n, and j is more than or equal to 1 and less than or equal to m;

changing X to (X)_ij)_n×mThe interaction attribute matrix is called because the matrix is constructed based on the mobile social network interaction data of the losers.

4. The mobile social network-based criminal and loan unlink relationship closeness circular search method according to claim 1, wherein a normalization processing method is then used to normalize the interaction attribute matrix, so as to obtain a normalized interaction matrix Y ═ Y (Y ═ Y)_ij)_m×n。

5. The mobile social network-based criminal and loan unlink contact relationship closeness circular search method according to claim 1, wherein the log-weighted interaction attribute matrix is specifically:

let the interaction attribute weight vector of the node be W ═ ω₁,ω₂,L,ω_mWhere ω is_jAccording to the formula

To calculate, for the normalized matrix Y ═ Y (Y)_ij)_n×m(ii) a Order to

z_ij＝ω_jy_ij

The matrix Z is (Z)_ij)_n×mA log-weighted interaction attribute matrix;

if the logarithm weighted interaction attribute matrix Z is equal to (Z)_ij)_n×mRepresents an inline vector Z ═ Z₁,Z₂,L,Z_n) Then Z is_i＝(z_i1,z_i2,L,z_in) For the attribute of the ith interactive user, 1 ≦ i ≦ n, so z_ij＝ω_jy_ijIs the jth weighted interaction attribute for the ith user.

6. The mobile social network-based criminal and loan unlink contact relationship closeness circular search method according to claim 1, wherein the maximum interactive users and the minimum interactive users are specifically:

taking the logarithm-weighted interaction attribute matrix Z as (Z)_ij)_n×mEach row Z of_iSelf-squaring mean

Namely, it is

I is more than or equal to 1 and less than or equal to n, which is called as weighted interaction attribute mean;

comparing Z ═ Z_ij)_n×mMean of n weighted interaction attributes

The largest weighted interaction attribute mean is recorded as

Then

Then the minimum weighted interaction attribute mean is recorded as

Then

Changing Z to (Z)_ij)_n×mMean of the largest weighted interaction attributes

The maximum interactive user called the loser is v_maxIndicating that the corresponding interaction attribute group is called the maximum interaction attribute group, using Z_maxRepresents;

changing Z to (Z)_ij)_n×mWeighted interaction attribute mean of medium minimum

The corresponding interactive user is called the minimum interactive user of the lost contact person, and is expressed by v_minRepresenting, with Z, the corresponding set of interaction attributes, called the minimum set of interaction attributes_minAnd (4) showing.

7. The mobile social network-based criminal and loan unlink contact relationship closeness circular search method according to claim 1, wherein the distance calculation formula is as follows:

in an unconnectorized mobile social network, assume Z ═ Z (Z)_ij)_n×mIs a logarithmically weighted interaction attribute matrix for which any two interacting users b_iAnd b_jTheir weighted interaction attribute groups are respectively Z_iAnd Z_jEstablishing any two interactive users b of the lost person_iAnd b_jFrom b to b_iTo b_jThe distance calculation formula of (c) is as follows:

d (Z)_i,Z_j) Called "Pangsulin-distance 1", where

And

respectively represent weighted interaction attribute groups respectively are Z_iAnd Z_jAverage value of (Z)_i-Z_j)^TIs Z_i-Z_jI is not less than 1, n is not less than j, i is not equal to j, ∑ is a weighted reciprocal matrix Z ═ Z (Z)_ij)_n×mOf the user, then, any interactive user v of the loser_iWith maximum interaction user b_maxAnd minimum interactive user v_minThe distances between are respectively:

8. the mobile social network-based criminal and loan unlink contact relation closeness cyclic search method according to claim 1, wherein the relation closeness calculation model is as follows:

formula r_iReferred to as "Pangsulin-relationship compactness of 1", wherein

And

it cannot be simultaneously 0 at the same time,

r_irepresenting the closeness of the relation between the unconnector and the ith interactive user, and r is more than or equal to 0_i≤1；r_i0 denotes lostThe closeness of the relationship between the contact and the ith interactive user is 0; r is_i1, the closeness of relationship between the unconnector and the ith interactive user is 1.

9. The mobile social network-based criminal and loan unconnector relationship closeness cyclic search method as claimed in claim 8, wherein through the unconnector relationship closeness calculation model, the interactive user relationship closeness set R ═ R { R } of the unconnector can be calculated₁,r₂,...,r_m}；

Setting the interactive user relationship compactness set R as { R ═ R₁,r₂,...,r_mElement r in (1)₁,r₂,...,r_mSorting is carried out in the order from big to small, and the element sequence of the sorting is assumed to be r'₁,r′₂,...,r′_m(r′₁,r′₂,...,r′_mIs r₁,r₂,...,r_mA combination of) then r'₁≥r′₂≥...≥r′_m。

10. The mobile social network-based criminal and loan unlink contact person relation tightness cyclic search method according to claim 1, wherein the relation tightness cyclic search algorithm is adopted to search unlink contacts, and specifically comprises the following steps:

step 1: gathering n mobile social users v of an unconnected person u₁,v₂,...,v_nThe interaction data of (2);

step 2: establishing a set of all user pairs C ═ { C) of the lost person u₁,c₂,...,c_mAnd establishing an interaction attribute matrix X of the user pairs (X ═ X)_ij)_m×n；

And 3, step 3: changing the interaction attribute matrix X to (X)_ij)_m×nCarrying out normalization processing to obtain a normalized interaction attribute matrix Y ═ Y_ij)_m×n；

And 4, step 4: calculating the logarithmic weight of each node interaction attribute to obtain a node interaction attribute weight set W ═ ω { (ω) }₁,ω₂,...,ω_n}；

And 5, step 5: establishing a logarithm weighting interactive attribute matrix Z ═ (Z ═_ij)_n×m；

And 6, step 6: calculating Z_maxAnd Z_min；

And 7, step 7: respectively calculating each interactive user v_iAttribute group Z_iAnd Z_maxAnd Z_minThe distance between

And

and 8, step 8: calculating the closeness of the unconnectorized relation:

obtaining a relation compactness set R ═ R of the unconnector₁,r₂,...,r_n}；

Step 9: set R-R of closeness of relationship of all interaction users of the lost contact person₁,r₂,...,r_nElement r in (1)₁,r₂,...,r_nSequencing the elements in the descending order to obtain the sequence of the sequenced elements of r'₁,r′₂,...,r′_nR 'is satisfied'₁≥r′₂≥...≥r′_n，r′₁,r′₂,...,r′_nIs r₁,r₂,...,r_nTo establish a new ordered set R '═ { R'₁,r′₂,...,r′_n}；

Step 10: from i to 1, i is more than or equal to 0 and less than or equal to n, and the relation compactness R ' is taken out of the ordered set R ' in sequence '_iPropose corresponding interactive users v_i(ii) a If can pass v_iIf the unconnected person u is found, the algorithm is ended; otherwise, the next step is carried out;

and 11, step 11: letting i equal to i +1, and executing the step 10 in a circulating manner; when i is equal to n, the algorithm ends.

Images