Account identification method and device, computer readable medium and electronic equipment

Abstract

The invention discloses an account identification method and device, a computer readable medium and electronic equipment, and the method comprises the steps: carrying out the feature extraction of first account information having a transaction relation with a to-be-identified account according to the historical transaction data of the to-be-identified account, and obtaining first feature information; performing feature extraction on second account information having a management relationship with the to-be-identified account to obtain second feature information; performing feature cross derivation processing on the historical transaction data to generate third feature information based on feature cross combination; and determining whether the to-be-identified account is a normal account or not according to the first feature information, the second feature information and the third feature information. According to the technical scheme of the invention, through the combination of the transaction relationship, the management relationship and the feature cross derivation, the data dimensions referred to by account identification are greatly enriched, and the accuracy of account identification is improved.

Description

Technical Field

[0001] The present application belongs to the field of Internet technology, and specifically relates to an account identification method, device, computer-readable medium and electronic device. Background Art

[0002] It is a common phenomenon to conduct transactions based on the Internet. People's identities on the Internet are usually reflected through accounts on various activity platforms. Since online transactions cannot intuitively determine the true identities of transaction participants like face-to-face transactions, online transactions are usually accompanied by certain risks. In order to improve the security of online transactions, various platforms will test the security of accounts. However, in some cases, this security test operation may be wrong, and normal accounts may be identified as abnormal accounts, which will have an adverse impact on the platform activities of normal accounts. Therefore, it is necessary to improve the accuracy of account identification. Summary of the invention

[0003] The purpose of this application is to provide an account identification method, device, computer-readable medium and electronic device to improve the accuracy of account identification.

[0004] Other features and advantages of the present application will become apparent from the following detailed description, or may be learned in part by the practice of the present application.

[0005] According to one aspect of an embodiment of the present application, there is provided an account identification method, including:

[0006] Extracting features of first account information having a transaction relationship with the account to be identified based on historical transaction data of the account to be identified, to obtain first feature information;

[0007] Extracting features of second account information that has a management relationship with the account to be identified to obtain second feature information;

[0008] Performing feature cross-derivation processing on the historical transaction data to generate third feature information based on the feature cross-combination;

[0009] Determine whether the account to be identified is a normal account according to the first feature information, the second feature information, and the third feature information.

[0010] According to one aspect of an embodiment of the present application, there is provided an account identification device, including:

[0011] A first feature extraction module is used to extract features of first account information that has a transaction relationship with the account to be identified based on historical transaction data of the account to be identified, so as to obtain first feature information;

[0012] A second feature extraction module is used to extract features from second account information that has a management relationship with the account to be identified, to obtain second feature information;

[0013] A third feature extraction module, used for performing feature cross-derivation processing on the historical transaction data to generate third feature information based on feature cross-combination;

[0014] An account identification module is used to determine whether the account to be identified is a normal account based on the first characteristic information, the second characteristic information and the third characteristic information.

[0015] In one embodiment of the present application, the first feature extraction module includes:

[0016] A network construction unit, configured to construct a transaction pair network corresponding to the account to be identified based on the historical transaction data of the account to be identified; the nodes in the transaction pair network are used to represent the account to be identified and a first account having a transaction relationship with the account to be identified;

[0017] The first feature extraction unit is used to extract features from the transaction pair network to obtain the first feature information.

[0018] In one embodiment of the present application, the first feature extraction unit is specifically used to:

[0019] According to the transaction, the account name corresponding to each node in the network is identified as the account type, and the account type corresponding to each node is obtained;

[0020] Among the nodes representing the first account, count the number of nodes corresponding to a specified account type and the proportion of nodes corresponding to the specified account type;

[0021] The first feature information is generated according to the number of nodes, the node ratio, and the account type corresponding to the node representing the account to be identified.

[0022] In one embodiment of the present application, the first feature extraction unit is specifically used to:

[0023] Performing noise reduction processing on the account names corresponding to the respective nodes to obtain pre-processed name texts;

[0024] Performing first feature processing on the preprocessed name text through a bidirectional transformation network to obtain a name feature;

[0025] The second feature processing is performed on the name feature through a multi-layer perceptron to obtain the account type corresponding to each node.

[0026] In one embodiment of the present application, the second feature extraction module is specifically used for:

[0027] Comparing the common transaction address of the account to be identified with the common transaction address of the second account in the second account information for consistency, to obtain a transaction location consistency feature;

[0028] Identify the industry category corresponding to the second account, and calculate the industry similarity between the second account and the account to be identified based on the industry category;

[0029] The second feature information is generated according to the consistency feature of the transaction location, the industry category and the industry similarity.

[0030] In one embodiment of the present application, the second feature extraction module is specifically used for:

[0031] Performing vectorization processing on the industry category and the account type of the account to be identified respectively to obtain an industry category vector and an account type vector;

[0032] The cosine similarity between the industry category vector and the account type vector is calculated, and the cosine similarity is used as the industry similarity between the second account and the account to be identified.

[0033] In one embodiment of the present application, the second feature extraction module is specifically used for:

[0034] extracting the risk status and risk screening data of the second account from the second account information; the risk status is used to indicate whether the second account is a normal account, and the risk screening data includes the number of times the second account is identified;

[0035] The second characteristic information is generated according to the risk status and the risk screening data.

[0036] In one embodiment of the present application, the third feature extraction module includes:

[0037] A preprocessing unit, configured to perform numerical anomaly processing on structured data in the historical transaction data, and perform category mapping processing on unstructured data in the historical transaction data, so as to obtain preprocessing feature data corresponding to the historical transaction data;

[0038] The third feature extraction unit is used to perform feature cross-derivation processing on the pre-processed feature data based on multiple feature dimensions, so as to generate the third feature information based on feature cross-combination.

[0039] In one embodiment of the present application, the third feature extraction unit is specifically used to:

[0040] Obtaining classification features corresponding to each feature dimension, wherein the classification features are used to classify the preprocessed feature data;

[0041] Cross-combining the classification features corresponding to each feature dimension to obtain multiple combined classification features;

[0042] Performing statistical processing on the preprocessed feature data according to a plurality of combined classification features to obtain statistical features corresponding to each combined classification feature;

[0043] The statistical features corresponding to each combined classification feature are used as the third feature information.

[0044] In one embodiment of the present application, the device further includes:

[0045] a fourth characteristic information extraction module, configured to divide the historical transaction data according to a plurality of time intervals, obtain the historical transaction data in each time interval, and perform statistical processing on the historical transaction data in each time interval to obtain the historical transaction statistical value of each time interval; for a plurality of time nodes, respectively calculate the transaction data change amount between the historical transaction data within a first preset time length before each time node and the historical transaction data within a second preset time length after each time node, to obtain the transaction data change amount corresponding to each time node; and use the historical transaction statistical value of each time interval and the transaction data change amount corresponding to each time node as the fourth characteristic information;

[0046] Among them, the account identification module is specifically used for:

[0047] Determine whether the account to be identified is a normal account according to the first feature information, the second feature information, the third feature information, and the fourth feature information.

[0048] In one embodiment of the present application, the account identification module is specifically used to:

[0049] Inputting the first feature information, the second feature information, and the third feature information into an account identification model to obtain probability data output by the account identification model, wherein the probability data indicates a probability that the account to be identified is a normal account;

[0050] When the probability data is greater than a preset threshold, it is determined that the account to be identified is a normal account.

[0051] In one embodiment of the present application, the device further includes:

[0052] A model building module is used to obtain, for multiple time nodes, sample transaction data within a first preset time period before each time node, and sample transaction data within a second preset time period after each time node, the sample transaction data including transaction data of at least one sample account; sample the sample transaction data within the first preset time period to obtain training data corresponding to each time node; generate test data corresponding to each time node based on the sample transaction data within the second preset time period; train the account recognition model through the training data corresponding to each time node, and test the trained account recognition model through the test data corresponding to each time node.

[0053] According to one aspect of an embodiment of the present application, a computer-readable medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the account identification method in the above technical solution is implemented.

[0054] According to one aspect of an embodiment of the present application, an electronic device is provided, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor executes the executable instructions so that the electronic device performs an account identification method as in the above technical solution.

[0055] According to one aspect of the embodiments of the present application, a computer program product or a computer program is provided, the computer program product or the computer program including computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the account identification method in the above technical solution.

[0056] In the technical solution provided in the embodiment of the present application, first, based on the historical transaction data of the account to be identified, feature extraction is performed on the first account information that has a transaction relationship with the account to be identified to obtain the first feature information; at the same time, feature extraction is performed on the second account information that has a management relationship with the account to be identified to obtain the second feature information; and, feature cross-derivation processing is performed on the historical transaction data to generate third feature information based on the feature cross-combination; finally, based on the first feature information, the second feature information and the third feature information, it is determined whether the account to be identified is a normal account, the first feature information and the second feature information respectively extract account features from the transaction relationship and the management relationship, and the third type of feature information expands the account features through feature cross-derivation. The combination of these three types of feature information greatly enriches the data dimensions referenced by account identification, thereby improving the accuracy of account identification.

[0057] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present application, and together with the specification are used to explain the principles of the present application. Obviously, the drawings described below are only some embodiments of the present application, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0059] Figure 1 The exemplary system architecture block diagram applying the technical solution of the present application is schematically shown.

[0060] Figure 2 The following is a flowchart schematically showing an account identification method provided by an embodiment of the present application.

[0061] Figure 3 The following is a flowchart schematically showing an account identification method provided by an embodiment of the present application.

[0062] Figure 4 The structural diagram of a bidirectional conversion network provided by an embodiment of the present application is schematically shown.

[0063] Figure 5 The structure diagram of a multi-layer perceptron provided by an embodiment of the present application is schematically shown.

[0064] Figure 6 A flowchart schematically illustrates the input-output relationship of the ReLU function provided by an embodiment of the present application.

[0065] Figure 7 A schematic diagram of a word vector represented by GloVe provided in an embodiment of the present application is schematically shown.

[0066] Figure 8 A schematic diagram of an account identification method provided by an embodiment of the present application is schematically shown.

[0067] Fig. 9 The structural block diagram of the account identification device provided in an embodiment of the present application is schematically shown.

[0068] Fig.10 The structure block diagram of a computer system suitable for implementing an electronic device of an embodiment of the present application is schematically shown. DETAILED DESCRIPTION

[0069] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this application will be more comprehensive and complete and fully convey the concept of the example embodiments to those skilled in the art.

[0070] In addition, described feature, structure or characteristic can be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present application. However, those skilled in the art will appreciate that the technical scheme of the present application can be put into practice without one or more of the specific details, or other methods, components, devices, steps, etc. can be adopted. In other cases, known methods, devices, realizations or operations are not shown or described in detail to avoid blurring the various aspects of the application.

[0071] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities may be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0072] The flowcharts shown in the accompanying drawings are only exemplary and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps can be decomposed, and some operations / steps can be combined or partially combined, so the actual execution order may change according to actual conditions.

[0073] In the embodiments of the present application, the term "module" or "unit" refers to a computer program or a part of a computer program with a predetermined function, and works together with other related parts to achieve a predetermined goal, and can be implemented in whole or in part by using software, hardware (such as processing circuits or memories) or a combination thereof. Similarly, a processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be part of an overall module or unit that includes the function of the module or unit.

[0074] It is understandable that in the specific implementation of this application, related data such as user information (such as transaction information) is involved. When the above embodiments of this application are applied to specific products or technologies, customer permission or consent is required, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0075] Figure 1 The exemplary system architecture block diagram applying the technical solution of the present application is schematically shown.

[0076] like Figure 1 As shown, the system architecture 100 may include a terminal device 110, a network 120 and a server 130. The terminal device 110 may include a smart phone, a tablet computer, a laptop computer, an intelligent voice interaction device, a smart home appliance, a vehicle-mounted terminal, etc. The server 130 may be an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The network 120 may be a communication medium of various connection types that can provide a communication link between the terminal device 110 and the server 130, for example, a wired communication link or a wireless communication link.

[0077] According to the implementation requirements, the system architecture in the embodiment of the present application can have any number of terminal devices, networks and servers. For example, the server 130 can be a server group composed of multiple server devices. In addition, the technical solution provided in the embodiment of the present application can be applied to the terminal device 110, can also be applied to the server 130, or can be implemented by the terminal device 110 and the server 130 together, and the present application does not make any special restrictions on this.

[0078] The account identification method provided by this application is described in detail below in conjunction with specific implementation methods.

[0079] Figure 2 The following schematically shows a flow chart of an account identification method provided by an embodiment of the present application, which can be implemented by a terminal device or a server. Figure 2 As shown, the account identification method provided in this embodiment includes steps 210 to 240, which are specifically as follows:

[0080] Step 210: Based on the historical transaction data of the account to be identified, feature extraction is performed on the first account information that has a transaction relationship with the account to be identified to obtain first feature information.

[0081] Specifically, the historical transaction data of the account to be identified may include transaction time, transaction amount, transaction product, transaction counterparty, transaction address, transaction scenario and other transaction-related data. Other accounts that have conducted transactions with the account to be identified are the first accounts that have a transaction relationship with the account to be identified. In other words, the first account is the transaction counterparty of the account to be identified. The first account information may include transaction information between the first account and the account to be identified and basic information of the first account. For example, the basic information of the first account includes information such as the name of the first account and the identifier of the first account. Feature extraction of the first account information is actually to extract feature information of the transaction counterparty of the account to be identified. For example, the name (or nickname) of the first account, the type of transaction product provided by the first account, the industry in which the first account is located, etc. can be extracted. These feature information are recorded as first feature information.

[0082] In one embodiment of the present application, in online transaction activities, users usually use created accounts to conduct transactions. Therefore, an account usually refers to a user, and an account can also be called a merchant. The account to be identified can be the account of a target user who needs to undergo account risk assessment. Risk assessment refers to determining whether an account is a normal account. If an account is a normal account, the risk of the transactions it conducts is relatively small; if an account is not a normal account (i.e., an abnormal account), the risk of the transactions it conducts is relatively large and is a suspicious transaction. Such suspicious transactions usually need to be stopped. Based on the corresponding relationship between accounts and users, users corresponding to normal accounts are normal users, and such users usually belong to whitelist users. Therefore, identifying whether an account is normal is to identify whether the corresponding user belongs to a whitelist user, so the account identification process is also called the user whitelisting process.

[0083] Step 220: extract features of the second account information that has a management relationship with the account to be identified to obtain second feature information.

[0084] Specifically, the second account that has a management relationship with the account to be identified refers to an account that has a strong correlation with the account to be identified other than transactions. For example, the target user corresponding to the account to be identified is the legal person or other management personnel (such as a super manager) of the second account. The second account information may include information such as the address of the second account, the industry of the second account, and the risk assessment of the second account. This second account information that has a management relationship with the account to be identified can usually also reflect the transaction risk of the account to be identified to a certain extent. Therefore, feature extraction of the second account information and then use of the second feature information to identify whether the account is normal can improve the accuracy of the identification result. The feature information extracted here can be features that can be compared between the account to be identified and the second account, such as their transaction addresses, industries, etc., and the risk assessment of the second account can also be used as extracted feature information.

[0085] Step 230: Perform feature cross-derivation processing on the historical transaction data to generate third feature information based on the feature cross-combination.

[0086] Specifically, historical transaction data includes transaction data of multiple feature dimensions, such as transaction time, transaction address, transaction amount, etc. The feature cross-derivation processing is to cross-combine these feature dimensions to form data information corresponding to the new feature dimension, which is the third feature information. For example, by cross-combining the transaction time and transaction amount, a new feature dimension can be obtained to represent the transaction amount within a certain transaction time interval; by cross-combining the transaction address and the transaction amount, a new feature dimension can be obtained to represent the transaction amount within a certain transaction address; by cross-combining the transaction time, transaction address and transaction amount, a new feature dimension can be obtained to represent the transaction amount of a certain transaction address within a certain transaction time interval; through the above feature cross-derivation processing, the third feature information can be obtained, including the transaction amount data within a certain transaction time interval, the transaction amount data within a certain transaction address, and the transaction amount data of a certain transaction address within a certain transaction time interval. It can be seen that through the feature cross-derivation processing, more feature dimension transaction information can be obtained, thereby expanding the account characteristics, which is conducive to improving the accuracy of account identification.

[0087] Step 240: Determine whether the account to be identified is a normal account based on the first characteristic information, the second characteristic information, and the third characteristic information.

[0088] Specifically, after obtaining the above three types of characteristic information, these three types of characteristic information can be input into the account identification model, which is a model pre-trained and tested based on sample transaction data. The input account identification model outputs probability data, which indicates the probability that the account to be identified is a normal account. The larger the probability data, the greater the probability that the account to be identified is a normal account. Then a threshold can be set. When the probability data is greater than the preset threshold, the account to be identified is determined to be a normal account; when the probability data is less than the preset threshold, the account to be identified is determined to be an abnormal account. It can be understood that the situation where the probability data is equal to the preset threshold can be classified into either the probability data greater than the preset threshold and the probability data less than the preset threshold according to actual needs.

[0089] In the technical solution provided in the embodiment of the present application, first, based on the historical transaction data of the account to be identified, feature extraction is performed on the first account information that has a transaction relationship with the account to be identified to obtain the first feature information; at the same time, feature extraction is performed on the second account information that has a management relationship with the account to be identified to obtain the second feature information; and, feature cross-derivation processing is performed on the historical transaction data to generate third feature information based on the feature cross-combination; finally, based on the first feature information, the second feature information and the third feature information, it is determined whether the account to be identified is a normal account, the first feature information and the second feature information respectively extract account features from the transaction relationship and the management relationship, and the third type of feature information expands the account features through feature cross-derivation. The combination of these three types of feature information greatly enriches the data dimensions referenced by account identification, thereby improving the accuracy of account identification.

[0090] Figure 3 The following schematically shows a flow chart of an account identification method provided by an embodiment of the present application, which can be implemented by a terminal device or a server. Figure 3 As shown, the account identification method provided in this embodiment includes steps 310 to 390, which are as follows:

[0091] Step 310: construct a transaction pair network corresponding to the account to be identified based on the historical transaction data of the account to be identified; the nodes in the transaction pair network are used to represent the account to be identified and the first account that has a transaction relationship with the account to be identified.

[0092] Specifically, if there is a transaction between the account to be identified and the first account, the account to be identified and the first account can form a transaction pair. The first accounts involved in the historical transaction data of the account to be identified and the account to be identified are used as nodes to form a transaction pair network. Two nodes with a transaction relationship in the network are a transaction pair.

[0093] In one embodiment of the present application, when constructing a transaction pair network, the historical transaction data may be preprocessed, and the transaction data that meets the requirements may be extracted to construct the transaction pair network. For example, the transaction amount of each transaction may be detected, and the transaction data with a transaction amount less than a preset value may be eliminated, and the transaction pair network may be constructed based on the transaction data with a transaction amount greater than the preset value. For another example, the number of transactions between the account to be identified and the first account may be detected, and the transaction data corresponding to the first account with a transaction number less than the preset number may be eliminated, and the transaction pair network may be constructed based on the transaction data corresponding to the first account with a transaction number greater than the preset number.

[0094] Step 320: extract features from the transaction pair network to obtain first feature information.

[0095] Specifically, based on the transaction pair network, data features such as the type of transaction products provided by the first account and the industry in which the first account is located can be extracted as the first feature information.

[0096] In one embodiment of the present application, the process of extracting features from a transaction pair network includes: identifying account types based on account names corresponding to each node in the transaction pair network to obtain account types corresponding to each node; among the nodes representing the first account, counting the number of nodes whose corresponding account types are the specified account types, and the proportion of nodes whose corresponding account types are the specified account types; generating first feature information based on the number of nodes, the proportion of nodes, and the account type corresponding to the nodes representing the account to be identified.

[0097] Specifically, first identify the account type corresponding to the node in the transaction pair network. The account type is also called the merchant type. The account can be classified based on the transaction products or services provided by the account. For example, based on the provided transaction products or services, the account type can include various accounts such as clothing, toys, home appliances, home furnishings, electronic products, and educational products. Optionally, the account can also be classified based on the profit model, such as account types including product accounts, platform accounts, training accounts, community accounts, etc. Product accounts are accounts with products as profit points, such as agents or sales of products. Platform accounts are accounts with platforms as profit points, such as specific product distribution platforms, and development platforms that provide technical support or services. Training accounts are accounts that provide training services with training as profit points, such as various training platforms or institutions and lecturer groups. Community accounts are accounts that obtain income by serving community members with communities as profit points. Optionally, accounts can also be classified based on trading platforms, such as A trading platform accounts, B trading platform accounts, C trading platform accounts, etc.

[0098] In one embodiment of the present application, the account name of an account usually reflects its account type to a certain extent, so the account name corresponding to the node can be identified by account type to obtain the account type corresponding to the node. Specifically, the account name corresponding to each node is first subjected to noise reduction processing to obtain a preprocessed name text. The noise reduction processing may include removing words such as spaces, illegal characters, emoticons, numbers, mobile phone numbers, and removing account name texts whose lengths do not meet the constraints. Then, the preprocessed name text is subjected to a first feature processing through a bidirectional transformation network to obtain a name feature. Finally, the name feature is subjected to a second feature processing through a multi-layer perceptron to obtain the account type corresponding to each node. The data output by the multi-layer perceptron includes the probability that the node belongs to various account types, and finally the account type corresponding to the maximum probability can be taken as the account type corresponding to the node.

[0099] For example, the structure of the bidirectional transformer network is as follows: Figure 4 As shown, it includes an input module 410, an encoding module 420, a decoding module 430 and an output module 440. The encoding module 420 includes a multi-head attention mechanism (Multi-HeadAttention) and a fully connected feedforward (Feed Forward) neural network, and the decoding module 430 includes a masked multi-head attention mechanism (Masked Multi-Head Attention), a multi-head attention mechanism and a fully connected feedforward neural network. The input module 410 converts the input account name into a matrix tuple feature, and then extracts the matrix tuple feature through the encoding module 420 and the decoding module 430, and outputs the name feature through the output module 440.

[0100] For example, the structure of a multilayer perceptron (MLP) is as follows: Figure 5 As shown, a multilayer perceptron generally includes an input layer, multiple hidden layers, and an output layer. In this embodiment, one hidden layer is taken as an example, that is, the multilayer perceptron has a three-layer structure. The name feature is processed by the input layer, the hidden layer, and the output layer to obtain the account type. The activation function of the hidden layer uses the ReLU function, and nonlinear factors can be introduced through this hidden layer. The output of the ReLU function is 0 or a positive number. When the input is positive, there is no gradient saturation problem and the calculation speed is faster. For example, the input-output relationship of the ReLU function is as follows Figure 6 As shown, when the input is greater than 0, the output of the ReLU function grows with the input, and there is no saturation, which can be expressed by the following formula:

[0101]

[0102] Among them, x represents the input data of the ReLU function, and σ(x) represents the output data of the ReLU function.

[0103] After determining the account type corresponding to each node, the number of nodes and the node ratio of the specified account type can be counted. Both parameters are counted in the nodes representing the first account, that is, the account nodes to be identified are not considered. The number of nodes of the specified account type indicates the number of first accounts of the specified account type, and the specified account type may be an account type with a higher transaction risk. The node ratio indicates the ratio of the number of first accounts of the specified account type to all first accounts corresponding to the account to be identified, which can be obtained based on the ratio of the number of nodes of the specified account type to the total number of first account nodes.

[0104] After the above processing, the account type corresponding to each node, the number of nodes of the specified account type and the node ratio can be obtained. Finally, the account type corresponding to the account node to be identified, the number of nodes of the specified account type and the node ratio can be used as the first feature information.

[0105] Step 330: perform consistency comparison between the commonly used transaction address of the account to be identified and the commonly used transaction address of the second account in the second account information to obtain a transaction location consistency feature.

[0106] Specifically, the second account is an account that has a management relationship with the account to be identified. Generally speaking, under normal circumstances, the commonly used transaction address of the second account should be consistent with the commonly used transaction address of the account to be identified. Therefore, the transaction location consistency feature between the second account and the account to be identified can be calculated. The transaction location consistency feature indicates whether the commonly used transaction addresses between the two are consistent. The transaction location consistency feature can be used to determine whether the account to be identified is a normal account. Here, the distance between the commonly used transaction address of the account to be identified and the commonly used transaction address of the second account can be calculated. If the distance is within a preset range, it can be considered that the addresses between the two are consistent; if the distance exceeds the preset range, it can be considered that the addresses between the two are inconsistent.

[0107] Step 340: Identify the industry category corresponding to the second account, and calculate the industry similarity between the second account and the account to be identified based on the industry category.

[0108] Specifically, the industry category of the second account can be obtained by using an account industry identification model, that is, the second account information is input into the account industry identification model to obtain the industry category of the second account. The industry category of the second account is then compared with the account type of the account to be identified to obtain the industry similarity between the two.

[0109] In one embodiment of the present application, the industry similarity calculation process includes: vectorizing the industry category and the account type of the account to be identified respectively to obtain the industry category vector and the account type vector; calculating the cosine similarity between the industry category vector and the account type vector, and using the cosine similarity as the industry similarity between the second account information and the account to be identified.

[0110] Specifically, the account type of the account to be identified is obtained based on the aforementioned identification of the account type of the nodes in the transaction pair network. The industry category and the account type of the account to be identified are both text data, which may include multiple words. When vectorizing, each word can be vectorized separately to obtain the vector of the entire text data. Optionally, the open source GloVe (Global Vectors for Word Representation) pre-trained word vector can be used to vectorize the industry category and the account type of the account to be identified. The word vector obtained by the GloVe query can effectively find similar words. For example, Figure 7 As shown in the figure, the word vectors of similar words are relatively close. For example, the word vectors of the words man and woman based on gender are similar.

[0111] Assume that the account type of the account to be identified is recorded as A, and its corresponding account type vector is recorded as V Ai ; The industry category of the second account is recorded as B, and its corresponding industry category vector is recorded as V Bj ; The similarity is calculated using the cosine similarity formula, so the industry similarity W ij The calculation formula is as follows:

[0112]

[0113] Step 350: Generate second feature information based on the consistency characteristics of the transaction location, industry category and industry similarity.

[0114] Specifically, the transaction location consistency characteristics, industry categories and industry similarities calculated above are used as the second feature information.

[0115] In one embodiment of the present application, the second characteristic information may also include the risk status and risk investigation data of the second account, the risk status is used to indicate whether the second account is a normal account, and the risk investigation data includes the number of times the second account is identified. Exemplarily, the risk status of the second account may be a normal state, a high transaction risk, or a transaction risk, or may be other forms of representation. The risk investigation data of the second account includes the number of times the second account is identified, that is, the number of times the second account is investigated to see if it is a normal account. Here, the risk status and the risk investigation data may be used as the second characteristic information, or the risk status and the risk investigation data may be used together with the aforementioned transaction location consistency features, industry categories, and industry similarities as the second characteristic information.

[0116] Step 360: Perform numerical anomaly processing on the structured data in the historical transaction data, and perform category mapping processing on the unstructured data in the historical transaction data, to obtain pre-processed feature data corresponding to the historical transaction data.

[0117] Specifically, structured data refers to data in historical transaction data that can be expressed in numerical form, such as transaction amount, number of transactions, transaction method (such as online payment is recorded as 1, offline payment is recorded as 2), etc. Unstructured data refers to data that cannot be expressed in numerical form in historical transaction data, such as text data, such as transaction time, transaction address, IP (Internet Protocol) address of the device where the account is located, etc. Structured data and unstructured data need to be processed differently to obtain pre-processed feature data corresponding to historical transaction data.

[0118] For structured data, we mainly handle exceptions in numerical dimensions, including removing null values ​​and abnormal values ​​with large deviations. For unstructured data, we mainly do category mapping, that is, dividing the data into different category intervals. For example, taking trading time as an example, when performing category mapping, the trading data can be mapped into time intervals such as night trading and morning trading.

[0119] Step 370: Perform feature cross-derivation processing on the pre-processed feature data based on multiple feature dimensions to generate third feature information based on feature cross-combination.

[0120] Specifically, after preprocessing, feature cross-derivation processing can be performed on the preprocessed feature data based on multiple feature dimensions, that is, different feature dimensions are cross-combined to obtain new feature dimensions, and then the third feature information is generated based on the new feature dimensions.

[0121] In one embodiment of the present application, the feature cross-derivation processing process includes: obtaining classification features corresponding to each feature dimension, the classification features are used to classify the pre-processed feature data; cross-combining the classification features corresponding to each feature dimension to obtain multiple combined classification features; performing statistical processing on the pre-processed feature data according to the multiple combined classification features to obtain statistical features corresponding to each combined classification feature; and using the statistical features corresponding to each combined classification feature as the third feature information.

[0122] Specifically, under each feature dimension, the pre-processed feature data can be classified based on the classification feature. In this embodiment, the feature dimension includes the time dimension, the data statistics method dimension and the attribute dimension. Taking the time dimension as an example, its classification feature can include time intervals of different lengths, such as 1 month, 3 months, 6 months, 9 months, 12 months and other time intervals. Taking the data statistics method dimension as an example, its classification feature can include different statistical methods, such as total (sum of feature data), proportion (value ratio of different attributes), maximum (maximum value of feature data), minimum (minimum value of feature data), average (average value of feature data) and other statistical methods. Taking the attribute dimension as an example, each attribute can include different classification features, and the attribute here can be any data information in the transaction data, such as the transaction method attribute, and the corresponding classification features can include online payment, offline payment, etc.; the transaction amount attribute, the corresponding classification features can include different amount intervals, such as less than 10,000, 10,000-50,000, more than 50,000, etc.; the transaction time attribute, the corresponding classification features can include morning transactions, early morning transactions, etc.

[0123] By cross-combining the classification features corresponding to each feature dimension, multiple combined classification features can be obtained. For example, taking the cross-combination of the classification features of the above data statistics method dimension and the classification features of the transaction method attribute as an example, the following combined classification features can be obtained: total online payment, online payment ratio, maximum online payment, minimum online payment, average online payment, total offline payment, offline payment ratio, maximum offline payment, minimum offline payment and average offline payment.

[0124] Exemplarily, the classification features of the three dimensional features can be cross-combined. Taking the transaction method attribute as an example, there are two classification features: online payment and offline payment. Then the time dimension and the statistical method dimension are cross-combined to generate the following combined classification features: the total amount of offline payments in the past month, the total number of offline payments in the past month, the ratio of the amount of offline payments to the total transaction amount in the past month, the ratio of the number of offline payments to the total number of transactions in the past month, the maximum amount of offline payments in the past month, the minimum amount of offline payments in the past month, the average daily amount of offline payments in the past month, and the average daily number of offline payments in the past month. The same applies to online payments and other months.

[0125] Finally, statistical processing is performed based on the above combined classification features, and the obtained statistical features are the third feature information. It can be seen that after feature cross-derivation processing, the representation dimension of transaction data is greatly expanded, making the dimensions considered for account identification more comprehensive, which is conducive to improving the accuracy of account identification.

[0126] In one embodiment of the present application, the information based on which account identification is based may also include fourth characteristic information, and the calculation process of the fourth characteristic information includes: dividing the historical transaction data according to multiple time intervals to obtain the historical transaction data in each time interval, and performing statistical processing on the historical transaction data in each time interval to obtain the historical transaction statistics for each time interval; for multiple time nodes, respectively calculating the transaction data change between the historical transaction data within a first preset time length before each time node and the historical transaction data within a second preset time length after each time node to obtain the transaction data change corresponding to each time node; using the historical transaction statistics for each time interval and the transaction data change corresponding to each time node as the fourth characteristic information.

[0127] For example, the historical transaction data can be divided into months as time intervals to obtain the historical transaction statistics for each month, such as the average monthly transaction amount. A monthly average transaction amount distribution curve can also be drawn to remove transaction data corresponding to excessively large and small values.

[0128] The selection of time nodes should eliminate the volatility of time factors such as months and seasons as much as possible. For example, four time nodes can be selected in a year: March, June, September, and December. For each time node, the corresponding transaction data change is calculated, and the transaction data change can be expressed as the ratio of the historical transaction data within the second preset time after the time node to the historical transaction data within the first preset time before the time node. Optionally, the first preset time is greater than the second preset time. Exemplarily, the first preset time is 12 months and the second preset time is 6 months. Taking March as an example, the transaction data change of the average monthly transaction volume in the 6 months after March compared with the average monthly transaction volume in the 12 months before March is calculated. Assuming it is March 2022, the transaction data change can be expressed as the ratio between the average monthly transaction volume of the 6 months from April 2022 to September 2022 and the average monthly transaction volume of the 12 months from April 2021 to March 2022. Optionally, the transaction changes can be classified according to the size of the ratio, for example, they can be divided into: an increase of more than 50%, a decrease of more than 50%, and a change between plus or minus 50% (recorded as other).

[0129] Step 380: Input the first feature information, the second feature information, and the third feature information into the account identification model to obtain probability data output by the account identification model, where the probability data represents the probability that the account to be identified is a normal account.

[0130] In one embodiment of the present application, the account identification model is obtained based on training and testing of sample transaction data, and the process specifically includes: for multiple time nodes, obtaining sample transaction data within a first preset time period before each time node, and sample transaction data within a second preset time period after each time node, the sample transaction data including transaction data of at least one sample account; sampling the sample transaction data within the first preset time period to obtain training data corresponding to each time node; generating test data corresponding to each time node based on the sample transaction data within the second preset time period; training the account identification model through the training data corresponding to each time node, and testing the trained account identification model through the test data corresponding to each time node.

[0131] Specifically, the selection of time nodes can refer to the selection method of the fourth characteristic information calculation process to eliminate the volatility of time factors such as months and seasons. For example, the transaction of refrigeration products has more transaction data in summer and less transaction data in winter. If only summer data or only winter data is taken, the data reference will be relatively one-sided, which will easily reduce the accuracy of account identification. Exemplarily, four time nodes are selected in March, June, September and December of the year, and each time node looks back on the transaction data of the past year and the next 6 months. For example: for the node in March 2022, the training set (ie, training data) selects sample transaction data from 2021.04 to 2022.03 for 12 months, and the prediction set (ie, test data) selects sample transaction data from 2022.04 to 2022.09 for 6 months. The data of each time node is randomly sampled and summarized into the total training data. Sampling can also be performed on the test data. Then, the account recognition model is trained using the training data, and the trained account recognition model is tested using the test data. If the test is qualified, the account recognition model can be used for the account recognition processing of the account to be recognized. If the test is unqualified, the above process can be repeated or the sample transaction data can be updated and the above process can be repeated to retrain and test.

[0132] During the model training process, the aforementioned first feature information, second feature information, third feature information and fourth feature information can be extracted from the training data, and these feature information corresponding to the training data can be input into the account recognition model. The model loss is calculated based on the output data of the account recognition model and the corresponding label data, and the model parameters are updated.

[0133] Step 390: When the probability data is greater than a preset threshold, the account to be identified is determined to be a normal account.

[0134] Step 380 to step 390 is to input the first characteristic information, the second characteristic information and the third characteristic information into the account identification model, and determine whether the account to be identified is a normal account through the account identification model.

[0135] Optionally, when the fourth characteristic information is also obtained in the above steps, the first characteristic information, the second characteristic information, the third characteristic information and the fourth characteristic information are input into the account identification model, and the account identification model is used to determine whether the account to be identified is a normal account.

[0136] In the technical solution provided in the embodiment of the present application, multiple aspects such as structured data, unstructured data, and management relationship with the account to be identified are considered in the feature extraction process, which is relatively comprehensive; by introducing unstructured data and comparing the account type of the account to be identified with the industry category of the second account for similarity, the characterization ability is stronger; at the same time, the account type of the first account and the risk situation of the second account are introduced to improve the accuracy of account identification.

[0137] The implementation process of the technical solution of this application is described below with a specific embodiment.

[0138] Figure 8 A schematic diagram of an account identification method provided by an embodiment of the present application is schematically shown.

[0139] like Figure 8 As shown, the method includes four parts: transaction pair network construction, related merchant analysis, transaction flow analysis and white model identification.

[0140] 1. Construction of trading pair network

[0141] To build a transaction pair network corresponding to the target user, we can obtain the target user's historical transaction data for the past year, aggregate it by user dimension, count the transaction frequency and transaction amount of each user, and obtain the counterparty users (i.e., the first account) that each user has historically traded with. We can only retain transaction pairs with transaction amounts greater than RMB 100. The target user represents the account to be identified.

[0142] The account type corresponding to the node in the transaction pair network is obtained through the micro-business text model, and the account type in this embodiment is the micro-business type. The target user and the counterparty's account nickname are denoised, mainly removing words such as spaces, illegal characters, emoticons, numbers, mobile phone numbers, and texts whose lengths do not meet the constraints are removed. The nickname after denoising is input into the micro-business text model to obtain the micro-business type of the node in the transaction pair network, including the counterparty's micro-business type (i.e., the counterparty's micro-business result) and the user's micro-business type (i.e., the user's micro-business result). The micro-business text model mainly uses the Transformer structure to superimpose MLP for prediction. After the nickname text is input, it is encoded as a matrix tuple, and feature learning is performed through a bidirectional transformer, and then the hidden state after characterization is input into the MLP layer for prediction output to obtain the micro-business type.

[0143] Based on the WeChat business type of each node, feature processing is performed to extract the WeChat business category, the number of opponent WeChat businesses, and the ratio of opponent WeChat businesses. The WeChat business category refers to the WeChat business type of the target user. The WeChat business label and the maximum prediction probability output by the WeChat business text model for the target user nickname are taken. The number of opponent WeChat businesses refers to the number of WeChat business counterparties of the target user (that is, the number of first account nodes with the specified account type). The ratio of opponent WeChat businesses is the number of WeChat business counterparties of the target user divided by the number of all trading counterparties of the user (that is, the ratio of nodes with the specified account type). The data obtained by feature processing the transaction pair network is the first feature information.

[0144] 2. Analysis of affiliated merchants

[0145] The associated merchant in this embodiment does not mean that the target user has ever conducted transactions on the merchant, but means that the target user is a legal person, super administrator, etc. of the merchant, and such a role indicates that the user is strongly related to the merchant. The associated merchant here is the second account, which has a management relationship with the target user.

[0146] The associated merchants are processed to extract second characteristic information, which includes the following characteristics:

[0147] (1) Merchant status, i.e., the risk status of the associated merchant, directly takes the merchant’s final status, such as normal, with XX risk, etc.

[0148] (2) Merchant screening times, i.e., risk screening data, which counts the number of times the merchant has been audited (i.e., account identification has been performed).

[0149] (3) Transaction location consistency, i.e., transaction location consistency feature, counts the target users’ commonly used transaction locations and the commonly used transaction locations of related merchants, and compares them to see if they are consistent.

[0150] (4) Merchant category, i.e., the industry category of the associated merchant, can be obtained from the existing merchant industry identification model.

[0151] (5) Industry similarity: the text similarity between the target user's WeChat business type and the industry category of the associated merchants is calculated to obtain the industry similarity between the two. The specific calculation process can refer to the calculation process of industry similarity in the above embodiment, which will not be repeated here.

[0152] 3. Transaction flow analysis

[0153] Transaction flow analysis is used to extract the third and fourth characteristic information, which include: monthly deposits and withdrawals, transaction scenarios (i.e., transaction methods), large transaction amounts, transaction fluctuations (i.e., transaction data changes), special transactions, high-frequency transactions, etc. This process also includes the selection and processing of sample data, which can be referred to the description of the third and fourth characteristic information extraction and model training process, and will not be repeated here.

[0154] 4. White model recognition

[0155] The white model is an account identification model. All the features generated in the first three steps are input into the white model, and the user's white judgment result is output. The white model is a machine learning model.

[0156] The technical solution provided by the embodiment of the present application combines three aspects: structured, unstructured, and related information, and has a relatively comprehensive dimension. At the same time, it introduces unstructured text information, and compares the user's WeChat business category with the industry category of its associated merchants for similarity, which has a stronger characterization ability. In addition, it introduces the WeChat business category of the counterparty and the risk situation of the associated merchants, thereby improving the effect of the white judgment model.

[0157] It should be noted that although the steps of the method in the present application are described in a specific order in the drawings, this does not require or imply that the steps must be performed in this specific order, or that all the steps shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps, etc.

[0158] The following introduces an apparatus embodiment of the present application, which can be used to execute the account identification method in the above-mentioned embodiment of the present application. Fig. 9 The structure block diagram of the account identification device provided in the embodiment of the present application is schematically shown. Fig. 9 As shown, the account identification device provided in the embodiment of the present application includes:

[0159] A first feature extraction module 910 is used to extract features of first account information that has a transaction relationship with the account to be identified based on historical transaction data of the account to be identified, to obtain first feature information;

[0160] A second feature extraction module 920 is used to extract features from second account information that has a management relationship with the account to be identified, to obtain second feature information;

[0161] A third feature extraction module 930, configured to perform feature cross-derivation processing on the historical transaction data to generate third feature information based on feature cross-combination;

[0162] The account identification module 940 is used to determine whether the account to be identified is a normal account according to the first characteristic information, the second characteristic information and the third characteristic information.

[0163] In one embodiment of the present application, the first feature extraction module 910 includes:

[0164] A network construction unit, configured to construct a transaction pair network corresponding to the account to be identified based on the historical transaction data of the account to be identified; the nodes in the transaction pair network are used to represent the account to be identified and a first account having a transaction relationship with the account to be identified;

[0165] The first feature extraction unit is used to extract features from the transaction pair network to obtain the first feature information.

[0166] In one embodiment of the present application, the first feature extraction unit is specifically used to:

[0167] According to the transaction, the account name corresponding to each node in the network is identified as the account type, and the account type corresponding to each node is obtained;

[0168] Among the nodes representing the first account, count the number of nodes corresponding to a specified account type and the proportion of nodes corresponding to the specified account type;

[0169] The first feature information is generated according to the number of nodes, the node ratio, and the account type corresponding to the node representing the account to be identified.

[0170] In one embodiment of the present application, the first feature extraction unit is specifically used to:

[0171] Performing noise reduction processing on the account names corresponding to the respective nodes to obtain pre-processed name texts;

[0172] Performing first feature processing on the preprocessed name text through a bidirectional transformation network to obtain a name feature;

[0173] The second feature processing is performed on the name feature through a multi-layer perceptron to obtain the account type corresponding to each node.

[0174] In one embodiment of the present application, the second feature extraction module 920 is specifically used for:

[0175] Comparing the common transaction address of the account to be identified with the common transaction address of the second account in the second account information for consistency, to obtain a transaction location consistency feature;

[0176] Identify the industry category corresponding to the second account, and calculate the industry similarity between the second account and the account to be identified based on the industry category;

[0177] The second feature information is generated according to the consistency feature of the transaction location, the industry category and the industry similarity.

[0178] In one embodiment of the present application, the second feature extraction module 920 is specifically used for:

[0179] Performing vectorization processing on the industry category and the account type of the account to be identified respectively to obtain an industry category vector and an account type vector;

[0180] The cosine similarity between the industry category vector and the account type vector is calculated, and the cosine similarity is used as the industry similarity between the second account and the account to be identified.

[0181] In one embodiment of the present application, the second feature extraction module 920 is specifically used for:

[0182] extracting the risk status and risk screening data of the second account from the second account information; the risk status is used to indicate whether the second account is a normal account, and the risk screening data includes the number of times the second account is identified;

[0183] The second characteristic information is generated according to the risk status and the risk screening data.

[0184] In one embodiment of the present application, the third feature extraction module 930 includes:

[0185] A preprocessing unit, configured to perform numerical anomaly processing on structured data in the historical transaction data, and perform category mapping processing on unstructured data in the historical transaction data, so as to obtain preprocessing feature data corresponding to the historical transaction data;

[0186] The third feature extraction unit is used to perform feature cross-derivation processing on the pre-processed feature data based on multiple feature dimensions, so as to generate the third feature information based on feature cross-combination.

[0187] In one embodiment of the present application, the third feature extraction unit is specifically used to:

[0188] Obtaining classification features corresponding to each feature dimension, wherein the classification features are used to classify the preprocessed feature data;

[0189] Cross-combining the classification features corresponding to each feature dimension to obtain multiple combined classification features;

[0190] Performing statistical processing on the preprocessed feature data according to a plurality of combined classification features to obtain statistical features corresponding to each combined classification feature;

[0191] The statistical features corresponding to each combined classification feature are used as the third feature information.

[0192] In one embodiment of the present application, the device further includes:

[0193] a fourth characteristic information extraction module, configured to divide the historical transaction data according to a plurality of time intervals, obtain the historical transaction data in each time interval, and perform statistical processing on the historical transaction data in each time interval to obtain the historical transaction statistical value of each time interval; for a plurality of time nodes, respectively calculate the transaction data change amount between the historical transaction data within a first preset time length before each time node and the historical transaction data within a second preset time length after each time node, to obtain the transaction data change amount corresponding to each time node; and use the historical transaction statistical value of each time interval and the transaction data change amount corresponding to each time node as the fourth characteristic information;

[0194] The account identification module 940 is specifically used for:

[0195] Determine whether the account to be identified is a normal account according to the first feature information, the second feature information, the third feature information, and the fourth feature information.

[0196] In one embodiment of the present application, the account identification module 940 is specifically used for:

[0197] Inputting the first feature information, the second feature information, and the third feature information into an account identification model to obtain probability data output by the account identification model, wherein the probability data indicates a probability that the account to be identified is a normal account;

[0198] When the probability data is greater than a preset threshold, it is determined that the account to be identified is a normal account.

[0199] In one embodiment of the present application, the device further includes:

[0200] A model building module is used to obtain, for multiple time nodes, sample transaction data within a first preset time period before each time node, and sample transaction data within a second preset time period after each time node, the sample transaction data including transaction data of at least one sample account; sample the sample transaction data within the first preset time period to obtain training data corresponding to each time node; generate test data corresponding to each time node based on the sample transaction data within the second preset time period; train the account recognition model through the training data corresponding to each time node, and test the trained account recognition model through the test data corresponding to each time node.

[0201] The specific details of the account identification device provided in each embodiment of the present application have been described in detail in the corresponding method embodiments and will not be repeated here.

[0202] Fig.10 The structure block diagram of a computer system for implementing an electronic device according to an embodiment of the present application is schematically shown.

[0203] It should be noted that Fig.10 The computer system 1000 of 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 application.

[0204] like Fig.10 As shown, the computer system 1000 includes a central processing unit 1001 (CPU), which can perform various appropriate actions and processes according to the program stored in the read-only memory 1002 (ROM) or the program loaded from the storage part 1008 to the random access memory 1003 (RAM). Various programs and data required for system operation are also stored in the random access memory 1003. The central processing unit 1001, the read-only memory 1002 and the random access memory 1003 are connected to each other through a bus 1004. The input / output interface 1005 (Input / Output interface, i.e., I / O interface) is also connected to the bus 1004.

[0205] The following components are connected to the input / output interface 1005: an input section 1006 including a keyboard, a mouse, etc.; an output section 1007 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 1008 including a hard disk, etc.; and a communication section 1009 including a network interface card such as a LAN card, a modem, etc. The communication section 1009 performs communication processing via a network such as the Internet. A drive 1010 is also connected to the input / output interface 1005 as needed. A removable medium 1011, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 1010 as needed, so that a computer program read therefrom is installed into the storage section 1008 as needed.

[0206] In particular, according to an embodiment of the present application, the process described in each method flow chart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a computer readable medium, and the computer program contains a program code for executing the method shown in the flow chart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 1009, and / or installed from the removable medium 1011. When the computer program is executed by the central processor 1001, various functions defined in the system of the present application are executed.

[0207] It should be noted that the computer-readable medium shown in the embodiment of the present application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by an instruction execution system, device or device or used in combination with it. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, wherein a computer-readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer readable signal media may also be any computer readable medium other than computer readable storage media, which may send, propagate, or transmit programs for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0208] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the above-mentioned module, program segment or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0209] It should be noted that, although several modules or units of the equipment for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present application, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into being embodied by multiple modules or units.

[0210] Through the description of the above implementation methods, it is easy for those skilled in the art to understand that the example implementation methods described here can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the implementation methods of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the implementation methods of the present application.

[0211] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary technical means in the art that are not disclosed in the present application.

[0212] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. An account identification method, characterized in that: include: Extracting features of first account information having a transaction relationship with the account to be identified based on historical transaction data of the account to be identified, to obtain first feature information; Extracting features of second account information that has a management relationship with the account to be identified to obtain second feature information; Performing feature cross-derivation processing on the historical transaction data to generate third feature information based on the feature cross-combination; Determine whether the account to be identified is a normal account according to the first feature information, the second feature information, and the third feature information.

2. The account identification method according to claim 1, characterized in that: Based on the historical transaction data of the account to be identified, feature extraction is performed on the first account information that has a transaction relationship with the account to be identified to obtain first feature information, including: Constructing a transaction pair network corresponding to the account to be identified based on the historical transaction data of the account to be identified; the nodes in the transaction pair network are used to represent the account to be identified and the first account that has a transaction relationship with the account to be identified; Perform feature extraction on the transaction pair network to obtain the first feature information.

3. The account identification method according to claim 2, characterized in that: Extracting features from the transaction pair network to obtain the first feature information includes: According to the transaction, the account name corresponding to each node in the network is identified as the account type, and the account type corresponding to each node is obtained; Among the nodes representing the first account, count the number of nodes corresponding to a specified account type and the proportion of nodes corresponding to the specified account type; The first feature information is generated according to the number of nodes, the node ratio, and the account type corresponding to the node representing the account to be identified.

4. The account identification method according to claim 3, characterized in that: According to the transaction, the account type of the account name corresponding to each node in the network is identified to obtain the account type corresponding to each node, including: Performing noise reduction processing on the account names corresponding to the respective nodes to obtain pre-processed name texts; Performing first feature processing on the preprocessed name text through a bidirectional transformation network to obtain a name feature; The second feature processing is performed on the name feature through a multi-layer perceptron to obtain the account type corresponding to each node.

5. The account identification method according to claim 1, characterized in that: Feature extraction is performed on the second account information that has a management relationship with the account to be identified to obtain second feature information, including: Comparing the common transaction address of the account to be identified with the common transaction address of the second account in the second account information for consistency, to obtain a transaction location consistency feature; Identify the industry category corresponding to the second account, and calculate the industry similarity between the second account and the account to be identified based on the industry category; The second feature information is generated according to the consistency feature of the transaction location, the industry category and the industry similarity.

6. The account identification method according to claim 5, characterized in that: Calculating the industry similarity between the second account and the account to be identified according to the industry category includes: Performing vectorization processing on the industry category and the account type of the account to be identified respectively to obtain an industry category vector and an account type vector; The cosine similarity between the industry category vector and the account type vector is calculated, and the cosine similarity is used as the industry similarity between the second account and the account to be identified.

7. The account identification method according to claim 1, characterized in that: Feature extraction is performed on the second account information that has a management relationship with the account to be identified to obtain second feature information, including: extracting the risk status and risk screening data of the second account from the second account information; the risk status is used to indicate whether the second account is a normal account, and the risk screening data includes the number of times the second account is identified; The second characteristic information is generated according to the risk status and the risk screening data.

8. The account identification method according to claim 1, characterized in that: Performing feature cross-derivation processing on the historical transaction data to generate third feature information based on the feature cross-combination, including: Performing numerical anomaly processing on structured data in the historical transaction data, and performing category mapping processing on unstructured data in the historical transaction data, to obtain preprocessed feature data corresponding to the historical transaction data; The preprocessed feature data is subjected to feature cross-derivation processing based on a plurality of feature dimensions, so as to generate the third feature information based on feature cross-combination.

9. The account identification method according to claim 8, characterized in that: Performing feature cross-derivation processing on the pre-processed feature data based on multiple feature dimensions to generate the third feature information based on feature cross-combination, including: Obtaining classification features corresponding to each feature dimension, wherein the classification features are used to classify the preprocessed feature data; Cross-combining the classification features corresponding to each feature dimension to obtain multiple combined classification features; Performing statistical processing on the preprocessed feature data according to a plurality of combined classification features to obtain statistical features corresponding to each combined classification feature; The statistical features corresponding to each combined classification feature are used as the third feature information.

10. The account identification method according to claim 1, characterized in that: The method further comprises: Dividing the historical transaction data according to multiple time intervals to obtain historical transaction data in each time interval, and performing statistical processing on the historical transaction data in each time interval to obtain historical transaction statistics for each time interval; For multiple time nodes, respectively calculate the transaction data change amount between the historical transaction data within a first preset time period before each time node and the historical transaction data within a second preset time period after each time node, to obtain the transaction data change amount corresponding to each time node; The historical transaction statistics of each time interval and the transaction data changes corresponding to each time node are used as fourth feature information; Wherein, determining whether the account to be identified is a normal account according to the first characteristic information, the second characteristic information, and the third characteristic information includes: Determine whether the account to be identified is a normal account according to the first feature information, the second feature information, the third feature information, and the fourth feature information.

11. The account identification method according to claim 1, characterized in that: Determining whether the account to be identified is a normal account according to the first feature information, the second feature information, and the third feature information includes: Inputting the first feature information, the second feature information, and the third feature information into an account identification model to obtain probability data output by the account identification model, wherein the probability data indicates a probability that the account to be identified is a normal account; When the probability data is greater than a preset threshold, it is determined that the account to be identified is a normal account.

12. The account identification method according to claim 11, characterized in that: The method further comprises: For multiple time nodes, sample transaction data within a first preset time period before each time node and sample transaction data within a second preset time period after each time node are obtained, wherein the sample transaction data includes transaction data of at least one sample account; Sampling the sample transaction data within the first preset time period to obtain training data corresponding to each time node; Generating test data corresponding to each of the time nodes according to the sample transaction data within the second preset time period; The account identification model is trained using the training data corresponding to each time node, and the trained account identification model is tested using the test data corresponding to each time node.

13. An account identification device, characterized in that: include: A first feature extraction module is used to extract features of first account information that has a transaction relationship with the account to be identified based on historical transaction data of the account to be identified, so as to obtain first feature information; A second feature extraction module is used to extract features from second account information that has a management relationship with the account to be identified, to obtain second feature information; A third feature extraction module, used for performing feature cross-derivation processing on the historical transaction data to generate third feature information based on feature cross-combination; An account identification module is used to determine whether the account to be identified is a normal account based on the first characteristic information, the second characteristic information and the third characteristic information.

14. A computer readable medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the account identification method described in any one of claims 1 to 12 is implemented.

15. An electronic device, characterized in that: include: processor; as well as A memory, configured to store executable instructions of the processor; The processor executes the executable instructions so that the electronic device executes the account identification method described in any one of claims 1 to 12.

16. A computer program product, characterized in that The computer program product includes computer instructions stored in a computer-readable storage medium; The processor of the computer device reads and executes the computer instructions from the computer-readable storage medium, so that the computer device executes the account identification method according to any one of claims 1 to 12.

Images