Graph algorithm anti-malicious debt avoidance identification method

Abstract

The present application relates to the field of computer. The graph algorithm anti-malicious debt collection recognition method comprises group feature analysis, and the group feature analysis comprises the following steps: step one, classifying overdue personnel; step two, extracting features of the community by using the graph algorithm; and step three, analyzing by using the features by the computer to judge the tendency and probability of the overall transformation of the community. First, the present application analyzes the overdue personnel by the computer, is not affected by personal experience and emotion, and has higher accuracy compared with the manual judgment method in the background technology. Secondly, the present application associates the overdue personnel with the community and extracts features, which are used for the collection personnel to judge the collection strategy.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of computer, in particular to computer application. BACKGROUND

[0002] In recent years, affected by multiple factors such as macro economy, the non-performing loan ratio of banks has increased, among which the non-performing loan ratio of retail loans has increased, and the non-performing loan ratio of credit cards has also increased. Various "anti-recovery" organizations, under the guise of being able to reduce the debts of debtors, instigate and incite them to actively evade debts, through means such as forging medical records, poverty certificates, fabricating difficult circumstances, and maliciously complaining against financial institutions and loan assistance platforms to evade debts, and make profits from it.

[0003] Currently, the identification of the "anti-recovery alliance" mainly relies on the manual judgment of customer service personnel based on dialogue, which is easily affected by personal experience, emotions and other states, resulting in misjudgment or missed reports. SUMMARY

[0004] The purpose of the present application is to provide a graph algorithm anti-malicious debt avoidance identification method to solve the above technical problems.

[0005] The technical problem solved by the present application can be realized by the following technical solutions:

[0006] The graph algorithm anti-malicious debt avoidance identification method comprises group feature analysis, characterized in that the group feature analysis comprises the following steps:

[0007] Step one, the computer classifies overdue personnel according to the relevant information of the overdue personnel, and divides the overdue personnel that can be associated through relevant information into the same community;

[0008] Step two, the computer extracts the following seven features by using a graph algorithm to extract the features of the community: feature 1, the number of neighbors of the customer nodes in the community; feature 2, the number of members in the community; feature 3, the shortest distance between each customer node in the community and the blacklisted customer nodes is calculated using the shortest path algorithm, and the result is written as an attribute for subsequent neighbor blacklist number calculation; feature 4, the influence of the nodes in the community network is calculated using PageRank; feature 5, the number of neighbor blacklist users of each customer node in the community; feature 6, the proportion of long and short account period customers in the community; feature 7, the maximum width of the customer's community.

[0009] Step three, the computer uses feature 1 to judge the tendency and probability of a single node being affected by the environment, uses feature 2 to judge the community influence degree, uses features 3 and 5 to calculate the risk coefficient of the anti-malicious debt avoidance alliance, uses feature 4 to find the most influential node in the community, and uses feature 6 to judge the tendency and probability of the overall conversion of the community.

[0010] Firstly, the application analyzes overdue personnel by computer, which is not affected by personal experience and emotion, and has higher accuracy than the manual judgment method in the background art. Secondly, the application associates the overdue personnel with the community and extracts features for the collection personnel to judge the collection strategy.

[0011] The graph algorithm anti-malicious debt avoidance identification method further includes an anti-malicious debt avoidance alliance identifier, which includes the following steps:

[0012] Step 1, collect the relevant information of the anti-malicious debt avoidance alliance, including the name and member list of the anti-malicious debt avoidance alliance, and also including the supervision complaint information and legal litigation information related to the anti-malicious debt avoidance;

[0013] Step 2, the computer matches and associates the relevant information of the anti-malicious debt avoidance alliance with the relevant information of the overdue personnel, and classifies the relationship between the overdue personnel and the anti-malicious debt avoidance alliance.

[0014] The graph algorithm anti-malicious debt avoidance identification method further includes personal feature analysis, which includes the following steps:

[0015] Step a, collect the personal basic information, transaction feature data, and loan information of the overdue personnel;

[0016] Step b, count the overdue amount of the overdue personnel in the recent time, whether the user number of the community where the overdue personnel is located requires long-term, the number of days of disconnection, the number of associated neighbors, and the repayment ratio in the past 6 months.

[0017] The computer matches and analyzes the overall conversion tendency and probability of the community where the overdue personnel is located obtained by the group feature analysis, whether the overdue personnel has the possibility of anti-malicious debt avoidance obtained by the anti-malicious debt avoidance alliance identifier, and the repayment ability of the overdue personnel obtained by the personal feature analysis, and displays the current total number of overdue personnel, the total number of suspected anti-malicious debt avoidance personnel, the number of convertible short account period personnel, and the total number of communities where the overdue personnel is located in a statistical summary manner through the display. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 The statistical analysis graph of the overdue personnel for the software platform;

[0019] Figure 2 The list of communities / groups;

[0020] Figure 3 The feature population proportion graph;

[0021] Figure 4 The group feature analysis graph;

[0022] Figure 5 The social relationship graph of the overdue personnel;

[0023] Figure 6 a detailed data graph for a user;

[0024] Figure 7 a personal feature analysis graph;

[0025] Figure 8 a group feature analysis graph. DETAILED DESCRIPTION

[0026] In order to make the technical means, creative features, purposes and effects of the present application easy to understand, the present application will be further described below in combination with specific drawings.

[0027] The graph algorithm anti-malicious debt avoidance identification method includes group feature analysis, and the group feature analysis includes the following steps:

[0028] Step one, classify overdue personnel: the computer classifies overdue personnel according to related information of the overdue personnel, and divides the overdue personnel that can be associated through the related information into the same community;

[0029] Step two, the computer extracts the following seven features by using a graph algorithm to extract features of the community: feature 1, the number of neighbors of the customer node in the community; feature 2, the number of members in the community; feature 3, the shortest distance between each customer node in the community and the blacklisted customer node is calculated by using the shortest path algorithm, and the result is written as an attribute for subsequent neighbor blacklist number calculation; feature 4, the influence of the node in the community network is calculated by using PageRank; feature 5, the number of neighbor blacklist users of each customer node in the community; feature 6, the proportion of long and short account period customer groups in the community; feature 7, the maximum width of the community where the customer is located.

[0030] Step three, the computer uses feature 1 to judge the tendency and probability of a single node being affected by the environment, uses feature 2 to judge the community influence degree, uses features 3 and 5 to calculate the risk coefficient of the anti-malicious debt avoidance alliance, uses feature 4 to find the most influential node in the community, and uses feature 6 to judge the tendency and probability of the overall conversion of the community.

[0031] Firstly, the present application analyzes the overdue personnel by computer, which is not affected by personal experience and emotion, and has higher accuracy than the manual judgment method in the background art. Secondly, the present application associates the overdue personnel with the community and extracts features, which are used to judge the collection strategy for the collection personnel.

[0032] The relevant information of the overdue personnel includes personal basic information (including name, gender, age, etc.), contact information (such as telephone number, WeChat number, QQ number, email, address, etc.), emergency contact information (such as the name, telephone number, address of the emergency contact, and the relationship with the overdue personnel, etc.), and the like. The computer can divide the overdue personnel that can be associated through the relevant information into the same community, and the relationship here includes strong relationship association, weak relationship association, indirect relationship association, and the like. For example, the addresses of two overdue personnel are the same community, and the telephone numbers of the two overdue personnel are consistent in the place of origin, so they can be considered as strong relationship association. The telephone numbers of two overdue personnel are consistent in the place of origin, and other relevant information is not associated, so they can be considered as weak relationship association. The addresses of the emergency contacts of two overdue personnel are the same community, so they can be considered as indirect association.

[0033] The graph algorithm anti-malicious debt avoidance identification method further includes an anti-malicious debt avoidance alliance identification, which includes the following steps:

[0034] Step 1, collect the relevant information of the anti-malicious debt avoidance alliance, including the name and member list of the anti-malicious debt avoidance alliance, and also including the supervision complaint information and legal litigation information related to the anti-malicious debt avoidance;

[0035] Step 2, the computer matches and associates the relevant information of the anti-malicious debt avoidance alliance with the relevant information of the overdue personnel, and classifies the relationship between the overdue personnel and the anti-malicious debt avoidance alliance.

[0036] The name and member list of the anti-malicious debt avoidance alliance include the name and member list of the anti-malicious debt avoidance alliance entity, the name and member list of the anti-malicious debt avoidance social group, and the like. The supervision complaint information and legal litigation information related to the anti-malicious debt avoidance include the name of the complainant, the telephone number used by the complainant when filing a complaint, the name of the complained object, the name of the plaintiff, the address of the plaintiff, the telephone number of the plaintiff, the name of the defendant, the address of the defendant, the telephone number of the defendant, the name of the litigation agent, the name of the law firm where the litigation agent is located, the telephone number of the litigation agent, and the like. The anti-malicious debt avoidance alliance identification is used to determine whether the overdue personnel has the possibility of anti-malicious debt avoidance.

[0037] The graph algorithm anti-malicious debt avoidance identification method further includes personal feature analysis, which includes the following steps:

[0038] Step a, collect the personal basic information, transaction feature data, and loan information of the overdue personnel;

[0039] Step b, count the overdue amount of the most recent time, whether the long-term is required within the past 90 days, the number of users in the community, the number of days of disconnection, the number of associated neighbors, and the repayment ratio in the past 6 months of the overdue personnel.

[0040] The personal feature analysis is used to determine the repayment ability of the overdue personnel.

[0041] The computer performs matching analysis on the overall conversion tendency and probability of the community where the delinquent persons are located, obtained from the group characteristic analysis; whether the delinquent persons are likely to engage in anti-malicious debt avoidance, obtained from the anti-malicious debt avoidance alliance identification; and the repayment ability of the delinquent persons obtained from the individual characteristic analysis. The computer displays the total number of current delinquent persons, the total number of suspected anti-malicious debt avoidance persons, the number of persons who can be converted to short-term debt, and the total number of communities where the delinquent persons are located in a statistical summary manner on the display screen.

[0042] The core functions of the software platform employing the graph algorithm-based anti-malicious debt evasion identification method of the present invention are as follows:

[0043] Overdue Person Statistics: Displays statistics on all overdue persons in the current collection pool, categorized by the number of overdue persons, suspected malicious debt evasion, convertible short-term payment terms, and community / group.

[0044] Group Feature Analysis: After clustering and classifying a large number of people, it is very easy to identify the characteristics of the group. This section lists the feature vectors for group classification, along with the magnitudes of the feature vector values. This information is used by debt collectors to determine collection strategies.

[0045] Analysis of interpersonal relationship characteristics: There are a large number of relationship characteristics among overdue individuals, including strong relationships, weak relationships, and indirect relationships.

[0046] Overdue personnel category list: It is necessary to classify overdue personnel to facilitate the mining and aggregation of target groups.

[0047] Overdue personnel information: A list of all overdue personnel, categorized by month and level.

[0048] List of Individual Characteristics of Overdue Individuals: Each overdue individual has distinct individual characteristics. Analyzing and summarizing these characteristics facilitates further segmentation of the population based on these characteristics.

[0049] like Figure 1 The software platform using the graph algorithm for identifying malicious debt evasion of the present invention has four statistical analysis dimensions for overdue individuals: the total number of current overdue individuals, the total number of suspected malicious debt evasion individuals, the number of individuals who can be converted to short-term debt, and the total number of communities where overdue individuals reside.

[0050] Clicking on different analysis dimensions allows you to view different lists of overdue individuals. Clicking on a community / group displays a list of communities / groups, such as... Figure 2 As shown.

[0051] In addition to displaying detailed group information in a list format, the system also provides characteristic group percentages, primarily in the form of a pie chart, for reference. Figure 3 This helps managers adjust collection strategies in a timely manner.

[0052] The software platform tests the characteristics of overdue personnel as follows:

[0053] After testing, it was found that among the 62442 investigated collection objects, 10000+ objects could form hundreds of communities / groups of various sizes through mobile phone numbers, transfer associations, etc., with the largest group having 146 people. By applying graph algorithms to the collection object communities / groups for feature extraction, the following 9 features were successfully extracted, such as Figure 8 : The number of neighbors of customer nodes within the community (business significance - the tendency and probability of a single node being affected by the environment); the number of members in the community; (business significance - community influence); using the shortest path algorithm to calculate the shortest distance between each customer node and the blacklist customer node within the community, the result is written as an attribute for subsequent 1, 2, and 3-degree neighbor blacklist number calculation; (business significance - calculate the risk coefficient of anti-malicious debt avoidance alliance); use PageRank to calculate the influence of nodes in the community network; (business significance - find the most influential nodes within the community to drive the conversion of the customer group or dissolve potential anti-malicious debt groups); the number of 1-degree neighbor blacklist users of each customer node within the community; (business significance - calculate the risk coefficient of anti-malicious debt alliance); the number of 2-degree neighbor blacklist users of each customer node within the community; (business significance - calculate the risk coefficient of anti-malicious debt alliance); the number of 3-degree neighbor blacklist users of each customer node within the community; (business significance - calculate the risk coefficient of anti-malicious debt alliance); the proportion of long and short account period customer groups in the community; (business significance - the tendency and probability of group conversion); the maximum width of the customer's community.

[0054] Based on the customer association community, a more obvious pattern / rule was found: the small customer group within the community is divided by the intermediary node, and most of them have the same label. That is, the potential anti-malicious debt tendency and negotiation willingness of the group members can be directly judged through this pattern.

[0055] Group feature analysis, refer to Figure 4 , 62442 investigated collection objects were analyzed, and 26 feature vectors were decomposed. Based on the length of these feature vectors, the feature table was sorted. Clicking on each feature vector, the social relationship graph can be further analyzed for that feature, refer to Figure 5 . The social relationship graph shows the data analysis based on mobile phone number, bank account number, and voiceprint, analyzing the key association relationships and importance of each object. Clicking on the user icon can display the relationship vector analysis centered on the user.

[0056] Users can perform data mining analysis from the mobile phone number dimension to observe the surrounding relationship data. Double-clicking the center user image can view the detailed data of the user, such as Figure 6Double-click the list of users, you can view the specified person detailed personal information from the user's perspective. Overdue personnel details are mainly targeted at individuals, analyzing user transaction feature data, user basic information, loan information, feature information. For the user's transaction information, analyze the overdue amount in the recent time, whether to require long-term in the past 90 days, the number of users in the sample community, the number of customers lost contact days, the number of neighbors associated with the sample 1 degree, the repayment ratio in the past 6 months. For example Figure 7 User basic information shows ID card, social security information, mobile phone number, fixed-line number, home address, emergency contact, company name, and rates and defines the user's repayment ability. User loan information shows loan limit, installment number and collection record. Each collection process recording can be played back to listen. Display user feature information, use the analysis capability of the graph database to analyze user feature dimensions and feature vectors. For reference by the collection personnel. Display the social relationship graph of the specified user, conduct data mining and analysis from the three dimensions of mobile phone number, bank account number and voiceprint, understand the relationship and distance between the user and other objects, and analyze whether the user has the possibility of anti-malicious debt avoidance.

[0057] Note: The personal information and data involved in this application are authorized or fully authorized information or data, and the collection, use and processing of related data comply with the relevant laws and standards of the country and region.

[0058] The above shows and describes the basic principles and main features of the present application and the advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above examples, and the above examples and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A graph algorithm anti-malicious debt avoidance identification method, comprising a group feature analysis, characterized in that, The population feature analysis comprises the following steps: Step 1: The computer classifies overdue personnel according to the relevant information of the overdue personnel, and divides the overdue personnel that can be associated through relevant information into the same community; Step 2: The computer extracts the following seven features by using a graph algorithm to extract the features of the community: Feature 1: the number of neighbors of the customer node in the community; Feature 2: the number of members in the community; Feature 3: the shortest distance between each customer node in the community and the blacklisted customer node is calculated using the shortest path algorithm, and the result is written as an attribute for subsequent neighbor blacklist number calculation; Feature 4: the influence of the node in the community network is calculated using PageRank; Feature 5: the number of neighbor blacklist users of each customer node in the community; Feature 6: the proportion of long and short account period customers in the community; Feature 7: the maximum width of the community where the customer is located; Step 3: The computer uses Feature 1 to judge the tendency and probability of a single node being affected by the environment; uses Feature 2 to judge the community influence degree; uses Features 3 and 5 to calculate the risk coefficient of the anti-malicious debt avoidance alliance; uses Feature 4 to find the most influential node in the community; and uses Feature 6 to judge the tendency and probability of the overall conversion of the community; It also includes an anti-malicious debt avoidance alliance identifier, which comprises the following steps: Step 1: Collect relevant information of the anti-malicious debt avoidance alliance, including the name and member list of the anti-malicious debt avoidance alliance, as well as supervision complaint information and legal litigation information related to anti-malicious debt avoidance; Step 2: The computer matches and associates the relevant information of the anti-malicious debt avoidance alliance with the relevant information of the overdue personnel, and classifies the relationship between the overdue personnel and the anti-malicious debt avoidance alliance.

2. The graph algorithm anti-evil debt avoidance identification method according to claim 1, characterized in that, The relevant information of the overdue personnel includes personal basic information, contact information, and emergency contact information.

3. The graph algorithm anti-evil debt avoidance identification method according to claim 1, characterized in that, The name and member list of the anti-malicious debt avoidance alliance includes the name and member list of the anti-malicious debt avoidance alliance entity and the name and member list of the anti-malicious debt avoidance social group.

4. The graph algorithm anti-evil debt avoidance identification method according to claim 1, characterized in that, The supervision complaint information and legal litigation information related to anti-malicious debt avoidance include the name of the complainant, the phone number used by the complainant when filing a complaint, the name of the person complained against, the name of the plaintiff, the address of the plaintiff, the phone number of the plaintiff, the name of the defendant, the address of the defendant, the phone number of the defendant, the name of the litigation agent, the name of the law firm where the litigation agent is located, and the phone number of the litigation agent.

5. The graph algorithm anti-evil debt avoidance identification method according to claim 1, characterized in that, It also includes personal feature analysis, which comprises the following steps: Step a: Collect the personal basic information, transaction feature data, and loan information of the overdue personnel; Step b: Statistics of the overdue amount of the overdue personnel in the latest time, whether the long-term is required within the past 90 days, the number of users in the community, the number of days of disconnection, the number of associated neighbors, and the repayment ratio in the past 6 months.

6. The graph algorithm anti-evil debt avoidance identification method according to claim 5, characterized in that, The computer matches and analyzes the tendency and probability of the overall conversion of the community where the overdue personnel is located obtained by the population feature analysis, whether the overdue personnel has the possibility of anti-malicious debt avoidance obtained by the anti-malicious debt avoidance alliance identifier, and the repayment ability of the overdue personnel obtained by the personal feature analysis, and displays the current total number of overdue personnel, the total number of suspected anti-malicious debt avoidance personnel, the number of convertible short account period personnel, and the total number of communities where the overdue personnel is located in a statistical summary manner through the display.

Images