Method, device, electronic device and storage medium for predicting illegal accounts

Abstract

The present application provides a prediction method, device, electronic device and storage medium for illegal accounts, wherein, in the prediction method for illegal accounts, firstly, the account information of N accounts to be predicted is obtained. Then, for each account to be predicted, the account information is extracted to obtain account features. Based on the account features of each account to be predicted, an account feature matrix and an account relationship matrix are constructed. Finally, a pre-constructed illegal account prediction model is called to process the account feature matrix and the account relationship matrix to obtain a prediction result. It can be seen that the method of the present application uses the account information of each account to be predicted obtained to generate account features, and construct an account feature matrix and an account relationship matrix. The account feature matrix and the account relationship matrix are processed by the pre-constructed illegal account prediction model to automatically obtain the prediction result, without the need for special auditors to audit each account.

Description

Technical Field

[0001] The present application relates to the field of data analysis technology, and in particular to a method, device, electronic device and storage medium for predicting illegal accounts. Background Art

[0002] In the daily operation of each unit, many projects need to be executed. Before executing a project, it is usually necessary to determine the budget of the project and allocate the budget through the budget unit. However, in the case of budget execution, there will be problems such as false reporting of projects, repeated reporting, misappropriation of funds, concealment of income, etc., which will cause certain losses to the unit.

[0003] In the prior art, the screening of trading accounts with these violations usually requires manual screening by skilled auditors, which consumes a lot of manpower and time. In addition, it is difficult to achieve 100% full coverage screening manually, resulting in unsatisfactory screening results. Summary of the invention

[0004] In view of this, the present application provides a prediction method, device, electronic device and storage medium for illegal accounts to solve the problem that the screening of trading accounts with these illegal behaviors in the prior art usually requires manual screening by skilled auditors, which consumes a lot of manpower and time, and it is difficult to achieve 100% full coverage screening manually, resulting in unsatisfactory screening results.

[0005] To achieve the above objectives, this application provides the following technical solutions:

[0006] The first aspect of the present application discloses a method for predicting a violation account, comprising:

[0007] Obtaining account information of N accounts to be predicted, where N is a preset positive integer;

[0008] For each of the accounts to be predicted, extracting features from the account information to obtain account features;

[0009] Based on the account characteristics of each of the accounts to be predicted, construct an account characteristic matrix and an account relationship matrix;

[0010] The pre-built illegal account prediction model is called to process the account feature matrix and the account relationship matrix to obtain a prediction result, wherein the illegal account prediction model is a model obtained after training with sample data.

[0011] Optionally, in the above method, the process of constructing the illegal account prediction model includes:

[0012] Obtain sample account data with marked account categories;

[0013] For each of the sample accounts, extracting features from the account information to obtain account features;

[0014] Based on the account characteristics of each of the sample accounts, an account characteristic matrix and an account relationship matrix are constructed;

[0015] Applying the account feature matrix and the account relationship matrix to the initial model for operation to obtain the account category of the sample account;

[0016] Determining whether the account type of the sample account is consistent with the account type of the sample account actually marked;

[0017] If the account category of the sample account is consistent with the account category of the sample account actually marked, the construction of the illegal account prediction model is completed;

[0018] If the account category of the sample account is inconsistent with the account category of the sample account actually marked, the error function is calculated and the parameters of the neural network model are adjusted using the error function until the account category of the sample account output is consistent with the account category of the sample account actually marked, and the construction of the illegal account prediction model is completed.

[0019] Optionally, in the above method, extracting features from the account information to obtain account features further includes:

[0020] The account characteristics are normalized.

[0021] Optionally, the above method, after calling the pre-built illegal account prediction model to process the account feature matrix and the account relationship matrix to obtain the prediction result, further includes:

[0022] If the prediction result is that there is a violation account, an alarm message of the violation account is generated.

[0023] The second aspect of the present application discloses a prediction device for illegal accounts, comprising:

[0024] An acquisition unit, used to acquire account information of N accounts to be predicted, wherein N is a preset positive integer;

[0025] A feature extraction unit, configured to extract features from the account information for each of the accounts to be predicted, to obtain account features;

[0026] A construction unit, configured to construct an account feature matrix and an account relationship matrix based on the account feature of each of the accounts to be predicted;

[0027] The prediction unit is used to call a pre-built illegal account prediction model to process the account feature matrix and the account relationship matrix to obtain a prediction result, wherein the illegal account prediction model is a model obtained after training with sample data.

[0028] Optionally, in the above device, the prediction unit includes:

[0029] An acquisition subunit is used to acquire sample account data with marked account categories;

[0030] A feature extraction subunit, configured to extract features from the account information for each of the sample accounts to obtain account features;

[0031] A first construction subunit is used to construct an account feature matrix and an account relationship matrix based on the account feature of each of the sample accounts;

[0032] A first operation subunit is used to perform operations on the account feature matrix and the account relationship matrix in an initial model to obtain the account category of the sample account;

[0033] A judging subunit, used to judge whether the account category of the sample account is consistent with the account category of the sample account actually marked;

[0034] A second construction subunit is used to complete the construction of the illegal account prediction model if the account category of the sample account is consistent with the account category of the sample account actually marked;

[0035] The parameter adjustment subunit is used to calculate the error function if the account category of the sample account is inconsistent with the account category of the sample account actually marked, and use the error function to adjust the parameters of the neural network model until the account category of the sample account output is consistent with the account category of the sample account actually marked, thereby completing the construction of the illegal account prediction model.

[0036] Optionally, the above device further includes:

[0037] A processing unit is used to perform standardization processing on the account characteristics.

[0038] Optionally, the above device further includes:

[0039] The alarm unit is used to generate alarm information of the illegal account if the prediction result shows that there is an illegal account.

[0040] The third aspect of the present application discloses an electronic device, comprising:

[0041] one or more processors;

[0042] a storage device having one or more programs stored thereon;

[0043] When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of the first aspects of the present invention.

[0044] A fourth aspect of the present application discloses a computer storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the method described in any one of the first aspects of the present invention is implemented.

[0045] It can be seen from the above technical scheme that in a prediction method for illegal accounts provided by the present application, the account information of N accounts to be predicted is first obtained. Then, for each account to be predicted, the account information is feature extracted to obtain account features. Based on the account features of each account to be predicted, an account feature matrix and an account relationship matrix are constructed. Finally, the pre-built illegal account prediction model is called to process the account feature matrix and the account relationship matrix to obtain the prediction result, wherein the illegal account prediction model is a model obtained after training with sample data. It can be seen that the method of the present application uses the account information of each account to be predicted obtained to generate account features, and constructs an account feature matrix and an account relationship matrix, and processes the account feature matrix and the account relationship matrix through the pre-built illegal account prediction model to automatically obtain the prediction result, without the need for special auditors to audit each account. The use of machine learning algorithms can greatly reduce regulatory costs and solve the problem that the screening of trading accounts with these illegal behaviors in the prior art usually requires manual screening by skilled auditors, which consumes a lot of manpower and time, and it is difficult to achieve 100% full coverage screening manually, resulting in unsatisfactory screening results. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0047] Figure 1 A flowchart of a method for predicting a violation account disclosed in another embodiment of the present application;

[0048] Figure 2 An account feature matrix disclosed in another embodiment of the present application;

[0049] Figure 3 The account relationship matrix disclosed in another embodiment of the present application;

[0050] Figure 4 This is a schematic diagram of information flow for calculating a violation account prediction model disclosed in another embodiment of the present application;

[0051] Figure 5 This is a flowchart of an implementation of a process for constructing a violation account prediction model in step S104 disclosed in another embodiment of the present application;

[0052] Figure 6 A schematic diagram of a device for predicting illegal accounts disclosed in another embodiment of the present application;

[0053] Figure 7 A schematic diagram of an electronic device disclosed in another embodiment of the present application. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0055] In this application, the terms "comprises", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element.

[0056] Furthermore, in this document, relational terms such as first and second, etc. are used only to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0057] As can be seen from the background technology, in the prior art, the screening of trading accounts with these violations usually requires manual screening by skilled auditors, which consumes a lot of manpower and time, and it is difficult to achieve 100% full coverage screening manually, resulting in unsatisfactory screening results.

[0058] In view of this, the present application provides a prediction method, device, electronic device and storage medium for illegal accounts to solve the problem that the screening of trading accounts with these illegal behaviors in the prior art usually requires manual screening by skilled auditors, which consumes a lot of manpower and time, and it is difficult to achieve 100% full coverage screening manually, resulting in unsatisfactory screening results.

[0059] The present application embodiment provides a method for predicting illegal accounts, specifically, Figure 1 As shown, including:

[0060] S101. Obtain account information of N accounts to be predicted, where N is a preset positive integer.

[0061] It should be noted that first, the account information of N accounts to be predicted is obtained, and the account information at least includes the account type (zero balance account / actual fund account), unit name, account login information (login IP, login time, login times), account transaction information (daily transaction number, direct payment number, authorized payment number, actual payment number, total transaction amount, transaction remarks, current transfer mark, peer mark, settlement method, voucher type), etc. Wherein, N is a preset positive integer, such as 10.

[0062] S102: For each account to be predicted, extract features from the account information to obtain account features.

[0063] It should be noted that after the account information of N accounts to be predicted is obtained, the account information is subjected to feature extraction for each account to be predicted to obtain account features.

[0064] Optionally, in another embodiment of the present application, after executing step S102, the following may also be included:

[0065] Standardize account characteristics.

[0066] It should be noted that after extracting the features of account information and obtaining the account features, the account features are standardized because the value ranges of different features are not the same. For example, the daily transaction amount may reach millions, but the number of daily transactions is generally only hundreds. This will cause different weights of different features, and too large values ​​may cause calculations to slow down. Therefore, the feature data is standardized so that all values ​​are in the range of [0,1]. At the same time, since features such as peer marks are not digital information, it is necessary to convert the information into digital information, such as 1 for the same line and 0 for the cross-line, to facilitate subsequent application of the data.

[0067] S103: Based on the account characteristics of each account to be predicted, an account characteristic matrix and an account relationship matrix are constructed.

[0068] It should be noted that based on the account features of each account to be predicted, an account feature matrix is ​​constructed. Specifically, a matrix of size N*D is established, where N is the number of accounts and D is the number of features extracted for each account. Figure 2 As shown, x1, x2, x3 represent accounts, f1, f2 represent account features, label [0, 1] represents a normal account, and [1, 0] represents an illegal account.

[0069] At the same time, based on the account characteristics of each account to be predicted, an account relationship matrix is ​​constructed. Specifically, a matrix of size N*N is established to represent the relationship between accounts. The traditional GCN algorithm only uses 0 and 1 as the values ​​of matrix elements. If accounts i and j have transactions, the matrix value A(i,j) is 1, otherwise it is 0. For the banking system, if this method is adopted, most of the values ​​in the matrix will be empty and the data cannot be effectively utilized. In order to utilize the data more effectively, we use the formula A(i,j)=w*C(i,j)+(1–w)*B(i,j), where C indicates whether accounts i and j have transactions, B indicates the similarity of the behavior patterns between accounts i and j, and w is a constant between 0 and 1, as shown in the following example. Figure 3 In this way, the behavioral relationship between accounts can be more accurately represented.

[0070] S104. Call a pre-built illegal account prediction model to process the account feature matrix and the account relationship matrix to obtain a prediction result, wherein the illegal account prediction model is a model obtained after training with sample data.

[0071] It should be noted that after constructing the account feature matrix and account relationship matrix based on the account features of each account to be predicted, the pre-built illegal account prediction model is called to process the account feature matrix and account relationship matrix to obtain the prediction result. The information flow of the model operation is as follows: Figure 4 As shown. Among them, the illegal account prediction model is a model obtained after training with sample data. The illegal account prediction model pre-built in this embodiment adopts a two-layer convolutional neural network model. The more layers of the neural network, the more accurate the calculation will be, but the amount of calculation will increase accordingly, and the calculation time will increase. The two-layer neural network can achieve a good balance between accuracy and calculation time. The activation function used in this model is:

[0072] Z=f(X,A)=softmax(A ReLU(AXW 0 )W 1 )

[0073] Where X is the account feature matrix, A is the account relationship matrix. W0 is the weight from the input layer to the hidden layer, and W1 is the weight from the hidden layer to the output layer. ReLU is the rectified linear unit function, and softmax is the normalized exponential function.

[0074] In a prediction method for a violation account provided by an embodiment of the present application, first, the account information of N accounts to be predicted is obtained. Then, for each account to be predicted, the account information is feature extracted to obtain account features. Based on the account features of each account to be predicted, an account feature matrix and an account relationship matrix are constructed. Finally, a pre-built violation account prediction model is called to process the account feature matrix and the account relationship matrix to obtain a prediction result, wherein the violation account prediction model is a model obtained after training with sample data. It can be seen that the method of the present application uses the account information of each account to be predicted obtained to generate account features, and constructs an account feature matrix and an account relationship matrix, and processes the account feature matrix and the account relationship matrix through the pre-built violation account prediction model to automatically obtain the prediction result, without the need for a special auditor to audit each account. The use of machine learning algorithms can greatly reduce regulatory costs, solve the problem that the screening of transaction accounts with these violations in the prior art usually requires manual screening by skilled auditors, thus consuming a lot of manpower and time, and it is difficult to achieve 100% full coverage screening manually, resulting in unsatisfactory screening results.

[0075] Optionally, in another embodiment of the present application, an implementation method of the process of constructing the illegal account prediction model in the above step S104 is as follows: Figure 5 As shown, it may include:

[0076] S501. Obtain sample account data with marked account categories.

[0077] It should be noted that first, sample account data with labeled account categories are obtained, and these data are used as training data and input into the neural network model for training.

[0078] S502: For each sample account, extract the account information to obtain account features.

[0079] S503: Based on the account characteristics of each sample account, an account characteristic matrix and an account relationship matrix are constructed.

[0080] It should be noted that the specific implementation of step S502 and step S503 can refer to the embodiment of the above-mentioned step S102 and step S103, which will not be repeated here.

[0081] S504: Apply the account feature matrix and the account relationship matrix to the initial model for operation to obtain the account categories of the sample accounts.

[0082] S505: Determine whether the account category of the sample account is consistent with the account category of the actually marked sample account.

[0083] It should be noted that after the account category of the sample account is calculated in the initial calculation model, the account category of the sample account calculated is compared with the account category of the actually marked sample account to determine whether the account category of the sample account calculated in the initial calculation model is accurate.

[0084] S506. If the account category of the sample account is consistent with the account category of the actually marked sample account, the construction of the illegal account prediction model is completed.

[0085] S507. If the account category of the sample account is inconsistent with the account category of the actually marked sample account, the error function is calculated and the parameters of the neural network model are adjusted using the error function until the account category of the output sample account is consistent with the account category of the actually marked sample account, and the construction of the illegal account prediction model is completed.

[0086] It should be noted that if the calculated account category of the sample account is inconsistent with the account category of the actually marked sample account, the error function of the neural network model is calculated, the parameters of the neural network model are adjusted using the error function, and a new round of calculations is performed again until the account category of the output sample account is consistent with the account category of the actually marked sample account, and the construction of the illegal account prediction model is completed.

[0087] Optionally, in another embodiment of the present application, after executing step S104, the following may also be included:

[0088] If the prediction result shows that there is a violation account, an alarm message of the violation account is generated.

[0089] It should be noted that if the prediction result is that there is a violation account, an alarm message will be generated based on the information of the violation account to remind the staff that the account is at risk.

[0090] Another embodiment of the present application also discloses a prediction device for illegal accounts, such as Figure 6 As shown, specifically including:

[0091] The acquisition unit 601 is used to acquire account information of N accounts to be predicted, where N is a preset positive integer.

[0092] The feature extraction unit 602 is used to extract features from the account information for each account to be predicted, and obtain account features.

[0093] The construction unit 603 is used to construct an account feature matrix and an account relationship matrix based on the account features of each account to be predicted.

[0094] The prediction unit 604 is used to call a pre-built illegal account prediction model to process the account feature matrix and the account relationship matrix to obtain a prediction result, wherein the illegal account prediction model is a model obtained after training with sample data.

[0095] In this embodiment, the specific execution process of the acquisition unit 601, the feature extraction unit 602, the construction unit 603 and the prediction unit 604 can be found in the corresponding Figure 1 The content of the method embodiment will not be repeated here.

[0096] In a prediction device for illegal accounts provided by an embodiment of the present application, first, an acquisition unit 601 acquires account information of N accounts to be predicted. Then, the feature extraction unit 602 extracts features from the account information for each account to be predicted to obtain account features. The construction unit 603 constructs an account feature matrix and an account relationship matrix based on the account features of each account to be predicted. Finally, the prediction unit 604 calls a pre-built illegal account prediction model to process the account feature matrix and the account relationship matrix to obtain a prediction result, wherein the illegal account prediction model is a model obtained after training with sample data. It can be seen that the method of the present application uses the acquired account information of each account to be predicted to generate account features, and construct an account feature matrix and an account relationship matrix, and automatically obtains a prediction result by processing the account feature matrix and the account relationship matrix through the pre-built illegal account prediction model, without the need for special auditors to audit each account. The use of machine learning algorithms can significantly reduce regulatory costs and solve the problem of existing technologies for screening trading accounts that have these violations. Manual screening usually requires skilled auditors, which consumes a lot of manpower and time. In addition, it is difficult to achieve 100% coverage screening manually, resulting in unsatisfactory screening results.

[0097] Optionally, in another embodiment of the present application, an implementation of the prediction unit 604 includes:

[0098] The acquisition subunit is used to obtain sample account data with marked account categories.

[0099] The feature extraction subunit is used to extract features from account information for each sample account to obtain account features.

[0100] The first construction subunit is used to construct an account feature matrix and an account relationship matrix based on the account features of each sample account.

[0101] The first operation subunit is used to operate the account feature matrix and the account relationship matrix in the initial model to obtain the account category of the sample account.

[0102] The judgment subunit is used to judge whether the account category of the sample account is consistent with the account category of the actually marked sample account.

[0103] The second construction subunit is used to complete the construction of the illegal account prediction model if the account category of the sample account is consistent with the account category of the actually marked sample account.

[0104] The parameter adjustment subunit is used to calculate the error function if the account category of the sample account is inconsistent with the account category of the actually marked sample account, and use the error function to adjust the parameters of the neural network model until the account category of the output sample account is consistent with the account category of the actually marked sample account, thereby completing the construction of the illegal account prediction model.

[0105] In this embodiment, the specific execution process of the acquisition subunit, the feature extraction subunit, the first construction subunit, the first operation subunit, the judgment subunit, the second construction subunit and the parameter adjustment subunit can be referred to in the corresponding Figure 6 The content of the method implementation example will not be repeated here.

[0106] Optionally, in another embodiment of the present application, the above-mentioned illegal account prediction device may further include:

[0107] A processing unit used to standardize account characteristics.

[0108] In this embodiment, the specific execution process of the processing unit can refer to the corresponding method embodiment content above, which will not be repeated here.

[0109] Optionally, in another embodiment of the present application, the above-mentioned illegal account prediction device may further include:

[0110] The alarm unit is used to generate alarm information of the illegal account if the prediction result is that there is an illegal account.

[0111] In this embodiment, the specific execution process of the alarm unit can refer to the corresponding method embodiment content above, which will not be repeated here.

[0112] Another embodiment of the present application further provides an electronic device, such as Figure 7 As shown, specifically including:

[0113] One or more processors 701.

[0114] The storage device 702 stores one or more programs.

[0115] When one or more programs are executed by one or more processors 701 , the one or more processors 701 implement any one of the methods in the above embodiments.

[0116] Another embodiment of the present application further provides a computer storage medium on which a computer program is stored, wherein when the computer program is executed by a processor, any one of the methods in the above embodiments is implemented.

[0117] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can refer to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment. The system and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without creative work.

[0118] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0119] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting illegal accounts, characterized in that: include: Obtaining account information of N accounts to be predicted, where N is a preset positive integer; For each of the accounts to be predicted, extracting features from the account information to obtain account features; Based on the account characteristics of each of the accounts to be predicted, construct an account characteristic matrix and an account relationship matrix; Calling a pre-built illegal account prediction model to process the account feature matrix and the account relationship matrix to obtain a prediction result, wherein the illegal account prediction model is a model obtained after training with sample data; The process of constructing the illegal account prediction model includes: Obtain sample account data with marked account categories; For each of the sample accounts, extracting features from the account information to obtain account features; Based on the account characteristics of each of the sample accounts, construct an account characteristic matrix and an account relationship matrix; Inputting the account feature matrix and the account relationship matrix into the initial model for calculation to obtain the account category of the sample account; Determining whether the account type of the sample account is consistent with the account type of the sample account actually marked; If the account category of the sample account is consistent with the account category of the sample account actually marked, the construction of the illegal account prediction model is completed; If the account category of the sample account is inconsistent with the account category of the sample account actually marked, the error function is calculated and the parameters of the initial model are adjusted using the error function until the account category of the sample account output is consistent with the account category of the sample account actually marked, and the construction of the illegal account prediction model is completed.

2. The method according to claim 1, characterized in that The extracting features from the account information to obtain account features also includes: The account characteristics are normalized.

3. The method according to claim 1, characterized in that The calling of the pre-built illegal account prediction model to process the account feature matrix and the account relationship matrix to obtain the prediction result also includes: If the prediction result is that there is a violation account, an alarm message of the violation account is generated.

4. A device for predicting illegal accounts, characterized in that: include: An acquisition unit, used to acquire account information of N accounts to be predicted, wherein N is a preset positive integer; A feature extraction unit, configured to extract features from the account information for each of the accounts to be predicted, to obtain account features; A construction unit, configured to construct an account feature matrix and an account relationship matrix based on the account feature of each of the accounts to be predicted; A prediction unit, configured to call a pre-built illegal account prediction model to process the account feature matrix and the account relationship matrix to obtain a prediction result, wherein the illegal account prediction model is a model obtained after training with sample data; Wherein, the prediction unit comprises: An acquisition subunit is used to acquire sample account data with marked account categories; A feature extraction subunit, configured to extract features from the account information for each of the sample accounts to obtain account features; A first construction subunit is used to construct an account feature matrix and an account relationship matrix based on the account feature of each of the sample accounts; A first operation subunit is used to input the account feature matrix and the account relationship matrix into an initial model for operation to obtain the account category of the sample account; A judging subunit, used to judge whether the account category of the sample account is consistent with the account category of the sample account actually marked; A second construction subunit is used to complete the construction of the illegal account prediction model if the account category of the sample account is consistent with the account category of the sample account actually marked; The parameter adjustment subunit is used to calculate the error function if the account category of the sample account is inconsistent with the account category of the sample account actually marked, and use the error function to adjust the parameters of the initial model until the account category of the sample account output is consistent with the account category of the sample account actually marked, thereby completing the construction of the illegal account prediction model.

5. The device according to claim 4, characterized in that Also includes: A processing unit is used to perform standardization processing on the account characteristics.

6. The device according to claim 4, characterized in that Also includes: The alarm unit is used to generate alarm information of the illegal account if the prediction result shows that there is an illegal account.

7. An electronic device, characterized in that: include: one or more processors; a storage device having one or more programs stored thereon; When the one or more programs are executed by the one or more processors, the one or more processors are enabled to implement the method according to any one of claims 1 to 3.

8. A computer storage medium, characterized in that: A computer program is stored thereon, wherein when the computer program is executed by a processor, the method according to any one of claims 1 to 3 is implemented.

Images