Behavior recognition system and method based on relational network, and computer program product

Abstract

The invention discloses a behavior recognition system and method based on a relation network and a computer program product, and the system comprises a searching module which is used for searching an associated person according to the relation network of a current user; the first screening module is used for screening initial strong associated persons according to the account states of the associated persons and / or the contact list interaction information of the associated persons and the user; the second screening module is used for sorting the initial strong association persons to obtain final strong association persons; the derivation module is used for deriving the user characteristics of the final strongly-associated persons; and the identification module is used for identifying the current user behavior according to the user characteristics and the derivative characteristics of the current user. According to the method, the associated persons strongly associated with the user are mined from numerous associated persons of the user through two times of screening, a large number of associated persons with small relevance are excluded, illegal behaviors are recognized by integrating derivative features of the strongly associated persons and user features, computing resources can be effectively saved, and data processing is accelerated; meanwhile, the accuracy of illegal behavior recognition can be improved, and the platform information security is guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of computer application technology, and in particular to a behavior recognition system, method and computer program product based on a relationship network. Background Art

[0002] At present, there are some violations that affect the information security of Internet service platforms (such as online shopping platforms, short video platforms, online car-hailing platforms, financial platforms, etc.) (such as data modification, data theft, fraud, breach of contract, etc.). The behavioral characteristics of related persons are often mined with the help of the user's relationship network, and based on the behavioral characteristics of related persons, it is identified whether the user has committed these violations.

[0003] In reality, the user's relationship network is complex and the data is huge, which leads to a large number of related persons found directly through the association network, and a large part of the related persons have little relevance to the user. On the one hand, this will waste computing resources and reduce the data processing speed. On the other hand, since the behavior of low-related related persons has a low correlation with the user's behavior, and their proportion is too high, it will seriously affect the accuracy of identifying the target user's behavior and pose a threat to the platform's information security. Summary of the invention

[0004] In view of this, the main purpose of the present invention is to propose a behavior recognition system, method and computer program product based on a relationship network, in order to at least partially solve at least one of the above-mentioned technical problems.

[0005] In order to solve the above technical problems, the first aspect of the present invention proposes a behavior recognition system based on a relationship network, the system comprising: A search module is used to search for related persons based on the current user's relationship network; A first screening module, used to screen out initial strong associated persons from the associated persons according to the associated persons' account status and / or the associated persons' contact book interaction information with the user; A second screening module is used to sort the initial strong associated persons to obtain final strong associated persons; A derivation module, used to derive the user features of the final strongly associated person to obtain derived features; The recognition module is used to input the user characteristics of the current user and the derived characteristics into the trained behavior recognition model to recognize the current user behavior.

[0006] According to a preferred embodiment of the present invention, the first screening module comprises: The first query module is used to query the account status of each associated person; The first sub-screening module is used to screen out the associated persons whose account status is in the rental state and / or has been returned and whose return time is less than a threshold as the initial strong associated persons.

[0007] According to a preferred embodiment of the present invention, the first screening module comprises: The second query module is used to query the address books of the associated persons and the user respectively, and select the first associated persons from the associated persons where both the associated persons and the user appear in each other's address books; The third query module queries the user's call records and selects initial strong associated persons who have call records with the user from the first associated persons.

[0008] According to a preferred embodiment of the present invention, the second screening module comprises: A sub-query module is used to query the call duration and / or call number between each initial strong associated person and the user; A sub-determination module, used to determine the call relevance according to the call duration and / or the number of calls; A sub-sorting module, used for sorting the initial strongly associated persons according to the call association degree; The sub-selection module is used to select the top N initial strong associated persons as the final strong associated persons; Wherein: N is a natural number.

[0009] According to a preferred embodiment of the present invention, the derivation module derives the user characteristics in one or more of the following ways: Input the user features of the final strong associated persons into a preset model, output the scores of each final strong associated person, and extract the user features of the strongest associated person corresponding to the nth quantile of the score as derived features; Select the maximum, minimum, or average value of the user features of all final strongly associated persons as derived features; The proportion of the call duration and / or call number between all final strongly associated persons and the current user in the call duration and / or call number between all associated persons and the current user is used as a derived feature.

[0010] In order to solve the above technical problems, the second aspect of the present invention provides a behavior recognition method based on a relationship network, the method comprising: Find related people based on the current user's relationship network; Filtering initial strong associated persons from the associated persons according to the associated persons' account status and / or the associated persons' contact book interaction information with the user; Sorting the initial strong associated persons to obtain final strong associated persons; Deriving the user features of the final strongly associated person to obtain derived features; The user features of the current user and the derived features are input into a trained behavior recognition model to recognize the current user behavior.

[0011] According to a preferred embodiment of the present invention, the screening of initial strong associated persons from the associated persons according to the associated persons' account status comprises: Query the account status of each associated person; The associated persons whose account status is "renting" and / or "returned" and whose return time is less than the threshold are selected as the initial strong associated persons.

[0012] According to a preferred embodiment of the present invention, the step of selecting initial strong associated persons from the associated persons according to the contact information of the associated persons and the user comprises: Search the address books of the associated persons and the user respectively, and select the first associated person from the associated persons whose associated persons and the user both appear in each other's address books; The user's call records are queried, and initial strong associated persons who have call records with the user are screened out from the first associated persons.

[0013] According to a preferred implementation manner of the present invention, the sorting of the initial strong associated persons to obtain the final strong associated persons includes: Query the call duration and / or call number between each initial strong associated person and the user; Determining call relevance based on the call duration and / or the number of calls; sorting the initial strongly associated persons according to the call association degree; Select the top N initial strong associated persons as the final strong associated persons; Wherein: N is a natural number.

[0014] According to a preferred embodiment of the present invention, the user characteristics are derived by one or more of the following methods: Input the user features of the final strong associated persons into a preset model, output the scores of each final strong associated person, and extract the user features of the strongest associated person corresponding to the nth quantile of the score as derived features; Select the maximum, minimum, or average value of the user features of all final strongly associated persons as derived features; The proportion of the call duration and / or call number between all final strongly associated persons and the current user in the call duration and / or call number between all associated persons and the current user is used as a derived feature.

[0015] In order to solve the above technical problems, the third aspect of the present invention provides an electronic device, including: Processor; and A memory storing computer executable instructions, which when executed cause the processor to perform any of the methods described above.

[0016] In order to solve the above technical problem, the fourth aspect of the present invention provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements any of the methods described above.

[0017] In summary, the present invention selects initial strong associated persons from the associated persons according to the account status of the user's associated persons and / or the interactive information between the associated persons and the user's address book; then sorts the initial strong associated persons to obtain the final strong associated persons; after deriving the user characteristics of the final strong associated persons, the behavior recognition model is trained based on the derived characteristics. The present invention mines the associated persons who are strongly associated with the user from the user's numerous associated persons through two screenings, excludes a large number of associated persons with relatively small correlations, and identifies illegal behaviors by combining the derived characteristics of the strongly associated persons and the user characteristics. On the one hand, it reduces the amount of data of the associated persons, can save computing resources, and speed up data processing; on the other hand, because the strong associated persons are more strongly associated with the user, their derived characteristics are more correlated with the user's behavior, which can improve the accuracy of identifying illegal behaviors and ensure the information security of the platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to make the technical problems solved by the present invention, the technical means adopted and the technical effects achieved more clearly, the specific embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, it should be noted that the drawings described below are only drawings of exemplary embodiments of the present invention, and those skilled in the art can obtain drawings of other embodiments based on these drawings without creative work.

[0019] Figure 1 It is a schematic diagram of the structural framework of a behavior recognition system based on a relationship network provided by an embodiment of the present invention; Figure 2 It is a flowchart of a behavior recognition method based on a relationship network provided by an embodiment of the present invention; Figure 3 is a structural block diagram of an exemplary embodiment of an electronic device according to the present invention; Figure 4 is a schematic diagram of an embodiment of a computer readable medium of the present invention. DETAILED DESCRIPTION

[0020] Under the premise of conforming to the technical concept of the present invention, the structure, performance, effect or other characteristics described in a specific embodiment may be combined with one or more other embodiments in any appropriate manner.

[0021] In the process of introducing specific embodiments, the detailed description of the structure, performance, effect or other features is to enable those skilled in the art to fully understand the embodiments. However, it does not exclude that those skilled in the art can implement the present invention with a technical solution that does not contain the above-mentioned structure, performance, effect or other features under certain circumstances. The figures in the accompanying drawings are only an exemplary demonstration, and do not mean that the solution of the present invention must include all the contents, operations and steps in the figures, nor do they mean that they must be executed in the order shown in the figures.

[0022] refer to Figure 1 FIG1 is a schematic diagram of a structural framework of a behavior recognition system based on a relationship network provided by an embodiment of the present invention. As shown in FIG1 , the system includes: A search module 11 is used to search for associated persons according to the relationship network of the current user; A first screening module 12, configured to screen out initial strong associated persons from the associated persons according to the associated persons' account status and / or the associated persons' contact list interaction information with the user; The second screening module 13 is used to sort the initial strong associated persons to obtain the final strong associated persons; A derivation module 14, used to derive the user features of the final strongly associated person to obtain derived features; The identification module 15 is used to input the user characteristics of the current user and the derived characteristics into the trained behavior identification model to identify the current user behavior.

[0023] In an optional implementation, the first screening module 12 may include: The first query module is used to query the account status of each associated person; The first sub-screening module is used to screen out the associated persons whose account status is in the rental state and / or has been returned and whose return time is less than a threshold as the initial strong associated persons.

[0024] In another optional implementation, the first screening module 12 may include: The second query module is used to query the address books of the associated persons and the user respectively, and select the first associated persons from the associated persons where both the associated persons and the user appear in each other's address books; The third query module queries the user's call records and selects initial strong associated persons who have call records with the user from the first associated persons.

[0025] Furthermore, the second screening module 13 may include: A sub-query module is used to query the call duration and / or call number between each initial strong associated person and the user; A sub-determination module, used to determine the call relevance according to the call duration and / or the number of calls; A sub-sorting module, used for sorting the initial strongly associated persons according to the call association degree; The sub-selection module is used to select the top N initial strong associated persons as the final strong associated persons; Wherein: N is a natural number.

[0026] Exemplarily, the derivation module 14 derives the user features in one or more of the following ways: Input the user features of the final strong associated persons into a preset model, output the scores of each final strong associated person, and extract the user features of the strongest associated person corresponding to the nth quantile of the score as derived features; Select the maximum, minimum, or average value of the user features of all final strongly associated persons as derived features; The proportion of the call duration and / or call number between all final strongly associated persons and the current user in the call duration and / or call number between all associated persons and the current user is used as a derived feature.

[0027] based on Figure 1 The behavior recognition system based on relationship network, the embodiment of the present invention also provides a behavior recognition method based on relationship network, such as Figure 2 , the model training method based on the relationship network includes: S1. Find related persons based on the current user's relationship network; In this embodiment, the relationship network is a data structure used to represent social network results, which regards users as nodes and the relationships between users as edges to form a graph. The nodes in the graph can represent various attributes of users, such as age, gender, interests, behaviors, etc.; the edges can represent the type and strength of the relationship between the two users connected by it.

[0028] Preferably, before this step, one or more social media platforms may be selected to establish the relationship network based on the user interaction data on the platform. For example, when there is interaction between two users, the nodes corresponding to the two users may be connected to form an edge.

[0029] In this embodiment, all node users connected to the node corresponding to the current user in the relationship network can be regarded as the current user's associated persons, including both first-degree associated persons and second-degree associated persons. The first-degree associated person refers to the node user directly connected to the current user in the relationship network, and the second-degree associated person refers to the node user connected to the current user through other nodes in the relationship network.

[0030] S2. Filtering out initial strong associated persons from the associated persons according to the associated persons' account status and / or the associated persons' contact list interaction information with the user; Considering that users who have not registered on the platform, or have registered but have not used platform services for a long time (for example, purchasing, renting or using products on the platform) have very little behavior data on the platform, these associated users need to be filtered out. Therefore, the initial strong associated persons can be filtered out based on the account status of the associated person to filter out the above users. In addition, the initial strong associated persons can also be filtered out based on the interactive information of the address book.

[0031] In one example, the initial strong associated persons may be screened only according to the associated person's account status, and this step may include: S21. Query the account status of each associated person; The account status is used to indicate the user's use of the platform services, which may include: purchased, not purchased, in the collection, rented, not rented, returned and the return time is less than a threshold, returned and the return time is greater than or equal to a threshold, not registered, etc.

[0032] S22. Filter out associated persons whose account status is "renting" and / or "returned" and whose return time is less than a threshold as initial strong associated persons.

[0033] Exemplarily, the account status of each associated person of the current user on the platform can be queried, and the associated persons whose account status is rented and / or returned and whose return time is less than a threshold are taken as initial strong associated persons. Furthermore, the initial strong associated persons can be stored for subsequent use.

[0034] In another example, the initial strong associated persons can be screened out based on the interaction information between the associated persons and the user's address book, where the address book interaction information refers to the information of the interaction between the user and the associated persons in their respective address books, such as: both parties have stored the other party in the address book, the two parties have had a call, etc. Then this step may include: S201, querying the address books of the associated persons and the current user respectively, and selecting the first associated person from the associated persons where both the associated person and the user appear in each other's address book; Through this step, the first associated person who is in the other party's address book with the current user can be screened out. Obviously, the association degree between the first associated person and the current user is greater than the association degree between the associated person and the current user.

[0035] S202: Query the call records of the user, and select initial strong associated persons who have call records with the user from the first associated persons.

[0036] This step can filter out the second associated person who is in the other party's contact book and has call records with the current user. Obviously, the association degree between the second associated person and the current user is greater than the association degree between the first associated person and the current user.

[0037] In another example, the initial strong associated persons can be screened out based on both the associated person's account status and the associated person's contact list interaction information with the user. For example, the associated persons whose account status is being rented and / or returned and whose return time is less than a threshold can be screened out through steps S21-S22, and then the associated persons who have contact list interaction information with the user can be screened out through steps S201-S202 as the initial strong associated persons. Of course, steps S201-S202 can also be performed first, and then steps S21-S22 are performed. The steps of the present invention are specifically limited.

[0038] S3, sorting the initial strong associated persons to obtain final strong associated persons; In this embodiment, the initial strongly associated persons may be sorted according to the call duration and / or the number of calls, and this step may include: S31, querying the call duration and / or call number between each initial strongly associated person and the current user; In order to save computing resources, a query period (for example, the last six months) can be configured to query the call duration and / or call number between each initial strong associated person and the current user within the query period.

[0039] S32, determining the call relevance according to the call duration and / or the number of calls; In principle, the longer the call duration and the more calls are made, the greater the call correlation. In one example, the call correlation L can be determined by the following formula: L = at + bs; Where: t is the call duration, s is the number of calls, and a and b are constants.

[0040] S33, sorting the initial strongly associated persons according to the call association degree; For example: sort the initial strong associated persons in descending order according to the call relevance.

[0041] S34, selecting the top N initial strong associated persons in the ranking as the final strong associated persons; Wherein: N is a natural number, such as 60, 100, etc., which can be configured according to actual needs.

[0042] S4, deriving the user features of the final strongly associated person to obtain derived features; Exemplarily, the user features may be derived in one or more of the following ways: Method 1: Input the user characteristics of the final strong associated person into the preset model, output the scores of each final strong associated person, and extract the user characteristics of the strongest associated person corresponding to the nth quantile of the score as derived features; wherein: the preset model can be a model for predicting a certain task, such as: behavior recognition model, demand recognition model, etc.

[0043] Method 2: Select the maximum, minimum, or average value of the user features of all final strongly associated persons as the derived features; Method three: taking the ratio of the call duration between all final strongly associated persons and the current user to the call duration between all associated persons and the current user as a derived feature; and / or taking the ratio of the number of calls between all final strongly associated persons and the current user to the number of calls between all associated persons and the current user as a derived feature.

[0044] In this embodiment, the user feature may be any device-related data that the device user chooses to make public or anonymize. It may include at least one of: user purchase or return records, user communication records, user information, and user behavior information; Among them, user purchase records refer to records of users purchasing goods on the platform. User return records refer to records of whether the user returns the goods on time after applying for platform goods. The goods can be physical goods, virtual goods, services, etc., and the present invention does not make specific limitations. The user communication records may include communication-related information such as address book contacts stored in the user's device, communication records stored in the user's device, etc. The user information may include: gender, age, education, fraud records, illegal and irregular records, data change records, etc. The user behavior information refers to the user's operation information on the platform, such as: browsing, visiting, collecting, clicking, etc.

[0045] S5. Input the user features of the current user and the derived features into the trained behavior recognition model to recognize the current user behavior.

[0046] Exemplarily, this step can collect the above-mentioned user features of the current user, input the user features of the current user and the derived features into a trained behavior recognition model to identify the probability that the current user has violated the rules, and determine whether the current user has violated the rules based on the probability value.

[0047] Among them: the behavior recognition model can be pre-trained through a large amount of historical user behaviors and the derived features of the historical users' final strongly associated persons.

[0048] The present invention mines out associates who are strongly associated with the user from among numerous associates of the user through two screenings, excludes a large number of associates with relatively small associations, and identifies illegal behaviors by combining the derived characteristics of the strongly associated persons and the user characteristics. On the one hand, the amount of data of the associated persons is reduced, which can save computing resources and speed up data processing; on the other hand, since the strongly associated persons are more strongly associated with the user, their derived characteristics are more highly correlated with the user behavior, thereby improving the accuracy of identifying illegal behaviors and ensuring the information security of the platform.

[0049] Those skilled in the art will appreciate that the modules in the above system embodiments may be distributed in the system as described, or may be changed accordingly and distributed in one or more systems different from the above embodiments. The modules in the above embodiments may be combined into one module, or may be further split into multiple sub-modules.

[0050] The electronic device embodiment of the present invention is described below, and the electronic device can be regarded as a physical implementation of the method and device embodiments of the present invention described above. The details described in the electronic device embodiment of the present invention should be regarded as a supplement to the above method or device embodiments; details not disclosed in the electronic device embodiment of the present invention can be implemented with reference to the above method or device embodiments.

[0051] Figure 3 is a structural block diagram of an exemplary embodiment of an electronic device according to the present invention. Figure 3 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0052] like Figure 3 As shown, the electronic device 300 of this exemplary embodiment is in the form of a general data processing device. The components of the electronic device 300 may include, but are not limited to: at least one processing unit 310, at least one storage unit 3320, a bus 330 connecting different electronic device components (including the storage unit 320 and the processing unit 310), a display unit 340, etc.

[0053] The storage unit 320 stores a computer-readable program, which may be a source program or a code of a read-only program. The program may be executed by the processing unit 310, so that the processing unit 310 performs the steps of various embodiments of the present invention. For example, the processing unit 310 may perform the following steps: Figure 2 Steps shown.

[0054] Bus 330 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0055] The electronic device 300 may also communicate with one or more external devices 100 (e.g., keyboard, display, network device, Bluetooth device, etc.), so that a user can interact with the electronic device 300 via these external devices 100, and / or the electronic device 300 can communicate with one or more other data processing devices (e.g., router, modem, etc.). Such communication may be performed through an input / output (I / O) interface 350, and may also be performed through a network adapter 360 with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network). The network adapter 360 may communicate with other modules of the electronic device 300 through a bus 330.

[0056] Figure 4 Schematic diagram of a computer readable medium embodiment of the present invention. Figure 4 As shown, the computer program can be stored on one or more computer-readable media. The computer-readable medium can be a readable signal medium or a readable storage medium. When the computer program is executed by one or more data processing devices, the computer-readable medium can implement the above method of the present invention, namely: searching for associated persons based on the relationship network of the current user; screening out initial strong associated persons from the associated persons based on the account status of the associated persons and / or the interaction information between the associated persons and the user's address book; sorting the initial strong associated persons to obtain the final strong associated persons; deriving the user features of the final strong associated persons to obtain derived features; inputting the user features of the current user and the derived features into the trained behavior recognition model to identify the current user behavior.

[0057] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, Figure 2 The method described.

[0058] In summary, the present invention can be implemented by executing a computer program method, system, electronic device or computer readable medium. In practice, a general data processing device such as a microprocessor or a digital signal processor (DSP) can be used to implement some or all functions of the present invention.

[0059] The specific embodiments described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the present invention is not inherently related to any specific computer, virtual device or electronic device, and various general devices can also implement the present invention. The above description is only a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A behavior recognition system based on a relationship network, characterized in that: The system comprises: A search module is used to search for related persons based on the current user's relationship network; A first screening module, used to screen out initial strong associated persons from the associated persons according to the associated persons' account status and / or the associated persons' contact book interaction information with the user; A second screening module is used to sort the initial strong associated persons to obtain final strong associated persons; A derivation module, used to derive the user features of the final strongly associated person to obtain derived features; The recognition module is used to input the user characteristics of the current user and the derived characteristics into the trained behavior recognition model to recognize the current user behavior.

2. The system according to claim 1, characterized in that The first screening module comprises: The first query module is used to query the account status of each associated person; The first sub-screening module is used to screen out the associated persons whose account status is in the rental state and / or has been returned and whose return time is less than a threshold as the initial strong associated persons.

3. The system according to claim 1, characterized in that The first screening module comprises: The second query module is used to query the address books of the associated persons and the user respectively, and select the first associated persons from the associated persons where both the associated persons and the user appear in each other's address books; The third query module queries the user's call records and selects initial strong associated persons who have call records with the user from the first associated persons.

4. The system according to claim 1, characterized in that The second screening module comprises: A sub-query module is used to query the call duration and / or call number between each initial strong associated person and the user; A sub-determination module, used to determine the call relevance according to the call duration and / or the number of calls; A sub-sorting module, used for sorting the initial strongly associated persons according to the call association degree; The sub-selection module is used to select the top N initial strong associated persons as the final strong associated persons; Wherein: N is a natural number.

5. The method according to any one of claims 1 to 4, characterized in that: The derivation module derives the user features in one or more of the following ways: Input the user features of the final strong associated persons into a preset model, output the scores of each final strong associated person, and extract the user features of the strongest associated person corresponding to the nth quantile of the score as derived features; Select the maximum, minimum, or average value of the user features of all final strongly associated persons as derived features; The proportion of the call duration and / or call number between all final strongly associated persons and the current user in the call duration and / or call number between all associated persons and the current user is used as a derived feature.

6. A behavior recognition method based on a relationship network, characterized in that: The method comprises: Find related people based on the current user's relationship network; Filtering initial strong associated persons from the associated persons according to the associated persons' account status and / or the associated persons' contact book interaction information with the user; Sorting the initial strong associated persons to obtain final strong associated persons; Deriving the user features of the final strongly associated person to obtain derived features; The user features of the current user and the derived features are input into a trained behavior recognition model to recognize the current user behavior.

7. The method according to claim 6, characterized in that The step of selecting initial strong associated persons from the associated persons according to the associated persons' account status includes: Query the account status of each associated person; The associated persons whose account status is "renting" and / or "returned" and whose return time is less than the threshold are selected as the initial strong associated persons.

8. The method according to claim 6, characterized in that The step of selecting initial strong associated persons from the associated persons according to the contact information of the associated persons and the user comprises: Search the address books of the associated persons and the user respectively, and select the first associated person from the associated persons whose associated persons and the user both appear in the other's address book; The user's call records are queried, and initial strong associated persons who have call records with the user are screened out from the first associated persons.

9. The method according to claim 6, characterized in that The initial strong associated persons are sorted to obtain the final strong associated persons including: Query the call duration and / or call number between each initial strong associated person and the user; Determining call relevance based on the call duration and / or the number of calls; sorting the initial strongly associated persons according to the call association degree; Select the top N initial strong associated persons as the final strong associated persons; Wherein: N is a natural number.

10. The method according to any one of claims 6 to 8, characterized in that: Derive user features in one or more of the following ways: Input the user features of the final strong associated persons into a preset model, output the scores of each final strong associated person, and extract the user features of the strongest associated person corresponding to the nth quantile of the score as derived features; Select the maximum, minimum, or average value of the user features of all final strongly associated persons as derived features; The proportion of the call duration and / or call number between all final strongly associated persons and the current user in the call duration and / or call number between all associated persons and the current user is used as a derived feature.

11. An electronic device, comprising: processor; as well as A memory storing computer executable instructions which, when executed, cause the processor to perform a method according to any one of claims 6 to 10.

12. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 6 to 10 is implemented.

Images