Model acquisition method and device, network equipment and storage medium

Abstract

The invention relates to the technical field of computers, in particular to a model acquisition method and device, network equipment and a storage medium. The model acquisition method comprises the following steps: establishing an anti-fraud customer relationship domain knowledge graph according to collected first historical user feature data; performing ethnic group classification on a feature set included in the anti-fraud customer relationship domain knowledge graph to obtain ethnic group features; learning the ethnic group features and the individual features in the anti-fraud customer relationship domain knowledge graph by adopting an initial graph neural network model, and obtaining association features between the ethnic group features and the individual features; and under the condition that the associated features meet feature requirements, obtaining a target image neural network model. According to the invention, the accuracy of model training can be improved, and the accuracy of fraud risk prediction can be improved.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a model acquisition method, device, network equipment and storage medium. Background Art

[0002] With the development of science and technology, more and more services are provided to users, and improving service quality has become the focus of users. For example, multiple target features (one or more) can be obtained and the AI ​​model can be trained. However, due to the differences in performance between models, the feature selection is relatively simple, resulting in poor accuracy in the risk prediction model. Summary of the invention

[0003] The present disclosure provides a model acquisition method, device, network device and storage medium to improve the accuracy of model training and the accuracy of fraud risk prediction. The technical solution of the present disclosure is as follows:

[0004] According to a first aspect of an embodiment of the present disclosure, a model acquisition method is provided, including:

[0005] Based on the collected first historical user feature data, a knowledge graph in the field of anti-fraud customer relations is established;

[0006] Classifying the feature set included in the anti-fraud customer relationship field knowledge graph into groups to obtain group features;

[0007] Using an initial graph neural network model to learn the group characteristics and individual characteristics in the anti-fraud customer relationship field knowledge graph, and obtaining correlation characteristics between the group characteristics and the individual characteristics;

[0008] When the associated features meet the feature requirements, the target graph neural network model is obtained.

[0009] According to some embodiments, the method further comprises:

[0010] Obtain user feature data;

[0011] Using the target graph neural network model to identify the user feature data, and obtaining relationship link information corresponding to the user feature data;

[0012] Fraud risk information corresponding to the user feature data is determined based on the relationship link information and the variation prediction link set.

[0013] According to some embodiments, after establishing the anti-fraud customer relationship domain knowledge graph based on the collected first historical user feature data, the method further includes:

[0014] Displaying the anti-fraud customer relationship field knowledge graph through a relationship graph, and obtaining a display image corresponding to the anti-fraud customer relationship field knowledge graph;

[0015] The display image is encrypted using a piecewise linear chaotic encryption algorithm, and the display image is stored.

[0016] According to some embodiments, establishing an anti-fraud customer relationship domain knowledge graph based on the collected first historical user feature data includes:

[0017] Performing data preprocessing on the collected first historical user feature data to obtain second historical user feature data, wherein the data preprocessing includes feature labeling processing, preclassification processing, initialization of boundary numbers and community modularization metric processing;

[0018] The second historical user feature data is processed by using a community discovery algorithm to remove the initialization boundary number and the measurement value of the community modularity to obtain the third historical user feature data, and the third historical user feature data is encrypted, and the encrypted third historical user feature data is stored in the blockchain node.

[0019] According to some embodiments, the initial graph neural network model is used to learn the group characteristics and individual characteristics in the anti-fraud customer relationship field knowledge graph to obtain a target graph neural network model, including:

[0020] Using an initial graph neural network model to learn the group characteristics and individual characteristics in the anti-fraud customer relationship field knowledge graph, and obtaining the weights of each data source corresponding to the first historical user feature data;

[0021] According to the weights of the data sources, a target graph neural network model is obtained.

[0022] According to some embodiments, the method further comprises:

[0023] Using random mutation operations, multiple simulations of user operations are performed to generate a set of mutation prediction links;

[0024] Adding the variation prediction link set to the first historical user data to obtain fourth historical user data;

[0025] Using the fourth historical user data, repeat the step of establishing an anti-fraud customer relationship field knowledge graph.

[0026] According to some embodiments, the method further comprises:

[0027] The variation prediction link set is converted into binary data by using piecewise linear chaotic mapping, and the binary data is stored.

[0028] According to a second aspect of an embodiment of the present disclosure, there is provided a model acquisition device, including:

[0029] A graph building unit, used to build an anti-fraud customer relationship domain knowledge graph based on the collected first historical user feature data;

[0030] A feature acquisition unit, used to classify the feature set included in the anti-fraud customer relationship field knowledge graph into groups to obtain group features;

[0031] The feature acquisition unit is further used to learn the group features and individual features in the anti-fraud customer relationship field knowledge graph using an initial graph neural network model to acquire correlation features between the group features and the individual features;

[0032] The model acquisition unit is used to acquire the target graph neural network model when the associated features meet the feature requirements.

[0033] According to a third aspect of an embodiment of the present disclosure, a network device is provided, including:

[0034] processor;

[0035] a memory for storing instructions executable by the processor;

[0036] The processor is configured to execute the instructions to implement the model acquisition method described in any one of the aforementioned aspects.

[0037] According to a fourth aspect of an embodiment of the present disclosure, a storage medium is provided. When instructions in the storage medium are executed by a processor of a network device, the network device is enabled to execute the model acquisition method described in any one of the preceding aspects.

[0038] According to a fifth aspect of an embodiment of the present disclosure, a computer program product is provided, including a computer program, wherein the computer program implements any one of the methods described in the preceding aspects when executed by a processor.

[0039] The technical solution provided by the embodiments of the present disclosure brings at least the following beneficial effects:

[0040] In some or related embodiments, a knowledge graph in the field of anti-fraud customer relations is established based on the collected first historical user feature data; the feature set included in the anti-fraud customer relationship knowledge graph is classified into groups to obtain group features; the initial graph neural network model is used to learn the group features and individual features in the anti-fraud customer relationship knowledge graph to obtain the correlation features between the group features and the individual features; and when the correlation features meet the feature requirements, the target graph neural network model is obtained. Therefore, the correlation relationship between historical user feature data, the hidden relationship between user feature data, and the tracking of user features through the knowledge graph can reduce the situation where the model training accuracy is poor due to the single feature selection, improve the accuracy of model training, and improve the accuracy of fraud risk prediction.

[0041] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description are used to explain the principles of the present disclosure, and do not constitute improper limitations on the present disclosure.

[0043] Figure 1 is a flow chart of a model acquisition method according to an exemplary embodiment;

[0044] Figure 2 is a flow chart of a model acquisition method according to an exemplary embodiment;

[0045] Figure 3 is a schematic diagram showing an example of a model acquisition method according to an exemplary embodiment;

[0046] Figure 4 is a block diagram of a model acquisition device according to an exemplary embodiment;

[0047] Figure 5 The invention is a block diagram of a network device according to an exemplary embodiment. DETAILED DESCRIPTION

[0048] In order to enable ordinary persons in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings.

[0049] The embodiments of the present disclosure propose a model acquisition method, device, network device and storage medium. In some embodiments, the terms such as model acquisition method, information processing method, communication method, etc. can be replaced with each other, the terms such as model acquisition device, information processing device, communication device, etc. can be replaced with each other, and the terms such as information processing system, communication system, etc. can be replaced with each other.

[0050] The embodiments of the present disclosure are not exhaustive, but are merely illustrative of some embodiments, and are not intended to be a specific limitation on the scope of protection of the present disclosure. In the absence of contradiction, each step in a certain embodiment can be implemented as an independent embodiment, and the steps can be arbitrarily combined. For example, a solution after removing some steps in a certain embodiment can also be implemented as an independent embodiment, and the order of the steps in a certain embodiment can be arbitrarily exchanged. In addition, the optional implementation methods in a certain embodiment can be arbitrarily combined; in addition, the embodiments can be arbitrarily combined, for example, some or all of the steps of different embodiments can be arbitrarily combined, and a certain embodiment can be arbitrarily combined with the optional implementation methods of other embodiments.

[0051] In each embodiment of the present disclosure, unless otherwise specified or there is a logical conflict, the terms and / or descriptions between the embodiments are consistent and can be referenced to each other, and the technical features in different embodiments can be combined to form a new embodiment based on their internal logical relationships.

[0052] The terms used in the embodiments of the present disclosure are only for the purpose of describing specific embodiments and are not intended to limit the present disclosure.

[0053] In the embodiments of the present disclosure, unless otherwise specified, elements expressed in the singular form, such as "a", "an", "the", "above", "said", "aforementioned", "this", etc., may mean "one and only one", or "one or more", "at least one", etc. For example, when using articles such as "a", "an", "the" in English in translation, the noun after the article may be understood as a singular expression or a plural expression.

[0054] In the embodiments of the present disclosure, “plurality” refers to two or more.

[0055] In some embodiments, the terms “at least one,” “one or more,” “a plurality of,” “multiple,” etc. may be used interchangeably.

[0056] In some embodiments, "at least one of A and B", "A and / or B", "A in one case, B in another case", "in response to one case A, in response to another case B", etc., may include the following technical solutions according to the situation: in some embodiments, A (A is executed independently of B); in some embodiments, B (B is executed independently of A); in some embodiments, execution is selected from A and B (A and B are selectively executed); in some embodiments, A and B (both A and B are executed). When there are more branches such as A, B, C, etc., the above is also similar.

[0057] In some embodiments, devices and equipment may be interpreted as physical or virtual, and their names are not limited to the names recorded in the embodiments. In some cases, they may also be understood as "equipment", "device", "circuit", "network element", "node", "function", "unit", "section", "system", "network", "chip", "chip system", "entity", "subject", etc.

[0058] In some embodiments, "terminal" or "terminal device" can be referred to as "user equipment (UE)", "user terminal" "mobile station (MS)", "mobile terminal (MT)", subscriber station, mobile unit, subscriber unit, wireless unit, remote unit, mobile device, wireless device, wireless communication device, remote device, mobile subscriber station, access terminal, mobile terminal, wireless terminal, remote terminal, handset, user agent, mobile client, client, etc.

[0059] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.) and signals involved in this application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0060] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0061] Figure 1 is a flow chart of a model acquisition method according to an exemplary embodiment. Figure 1 As shown, the following steps are included:

[0062] In step S11, an anti-fraud customer relationship domain knowledge graph is established based on the collected first historical user feature data;

[0063] In some embodiments, the execution subject of the embodiments of the present disclosure may be, for example, a network device, and the name of the network device is not limited. For example, the network device may also be called a server, a base station, or a server cluster, etc. The embodiments of the present disclosure are not limited to this. The technical solution of the embodiments of the present disclosure may be applied to, for example, a financial risk quantification control scenario.

[0064] According to some embodiments, the user characteristic data may be, for example, characteristic data related to the user, and the user characteristic data may be, for example, user characteristic data related to credit risk collected from a relationship perspective. The user characteristic data may include, for example, customer registration information, a credit scoring matrix, past risk events and risk levels, user social relationships, etc. The user characteristic data does not specifically refer to a fixed data. For example, when the amount of data included in the user characteristic data changes, the user characteristic data may also change accordingly.

[0065] In some embodiments, the historical user characteristic data may be, for example, user characteristic data that has been generated in history. The historical user characteristic data does not specifically refer to a fixed data. For example, when the historical duration changes, the historical user characteristic data may also change accordingly. For example, when the number of data types included in the historical user characteristic data changes, the historical user characteristic data may also change accordingly.

[0066] According to some embodiments, the first one of the first historical user characteristic data is used to distinguish from the rest of the historical user characteristic data. The first historical user characteristic data does not specifically refer to a fixed data. For example, when the historical duration changes, the first historical user characteristic data may also change accordingly. For example, when the number of data types included in the first historical user characteristic data changes, the first historical user characteristic data may also change accordingly.

[0067] In some embodiments, a knowledge graph refers to a structured semantic knowledge base that is used to quickly describe concepts and their interrelationships in the physical world. By reducing the data granularity from the text level to the data level, a large amount of knowledge can be aggregated to achieve rapid response and reasoning of knowledge.

[0068] In some embodiments, the anti-fraud customer relationship domain knowledge graph may be applicable to a knowledge graph for representing customer relationships in anti-fraud scenarios. The anti-fraud customer relationship domain knowledge graph does not specifically refer to a fixed knowledge graph. For example, when the acquired historical user feature data changes, the anti-fraud customer relationship domain knowledge graph may also change accordingly.

[0069] In some embodiments, for example, an anti-fraud customer relationship domain knowledge graph may be established based on the collected first historical user feature data.

[0070] In step S12, the feature set included in the anti-fraud customer relationship field knowledge graph is classified into groups to obtain group features;

[0071] According to some embodiments, the group feature may be used to indicate type information corresponding to the feature set. The group feature does not specifically refer to a fixed feature. For example, when the feature set changes, the group feature may also change accordingly. For example, when the group classification method changes, the group feature may also change accordingly.

[0072] In some embodiments, for example, the feature set included in the anti-fraud customer relationship field knowledge graph may be classified into groups to obtain group features.

[0073] In step S13, an initial graph neural network model is used to learn the group characteristics and individual characteristics in the anti-fraud customer relationship field knowledge graph to obtain correlation characteristics between the group characteristics and the individual characteristics;

[0074] In some embodiments, the initial graph neural network model may be, for example, a model before training. The initial graph neural network model does not feature a fixed model. For example, when the model type corresponding to the initial graph neural network model changes, the initial graph neural network model may also change accordingly. For example, when the model parameters of the initial graph neural network model change, the initial graph neural network model may also change accordingly.

[0075] In some embodiments, individual characteristics may be used to refer to characteristics of a certain user, for example. The individual characteristics do not specifically refer to a certain fixed characteristic. For example, when the characteristic data of a certain user changes, the individual characteristics may also change accordingly.

[0076] In some embodiments, the association feature can be used to indicate that there is an association relationship between a group feature and an individual feature. The association feature does not specifically refer to a fixed feature. For example, when the group feature or the individual feature changes, the association feature may also change accordingly.

[0077] In some embodiments, an initial graph neural network model is used to learn the group characteristics and individual characteristics in the anti-fraud customer relationship field knowledge graph to obtain the correlation characteristics between the group characteristics and the individual characteristics.

[0078] In step S14, when the associated features meet the feature requirements, the target graph neural network model is acquired.

[0079] According to some embodiments, when an associated feature is obtained, it can be determined whether the associated feature meets the feature requirement. When the associated feature meets the feature requirement, the target graph neural network model is obtained. The feature requirement does not specifically refer to a fixed requirement. For example, when a modification instruction for the feature requirement is received, the feature requirement may also change accordingly. For example, when the model type of the graph neural network model changes, the feature requirement may also change accordingly.

[0080] In some or related embodiments, a knowledge graph in the field of anti-fraud customer relations is established based on the collected first historical user feature data; the feature set included in the anti-fraud customer relationship knowledge graph is classified into groups to obtain group features; the initial graph neural network model is used to learn the group features and individual features in the anti-fraud customer relationship knowledge graph to obtain the correlation features between the group features and the individual features; and when the correlation features meet the feature requirements, the target graph neural network model is obtained. Therefore, the correlation relationship between historical user feature data, the hidden relationship between user feature data, and the tracking of user features through the knowledge graph can reduce the situation where the model training accuracy is poor due to the single feature selection, improve the accuracy of model training, and improve the accuracy of fraud risk prediction.

[0081] Figure 2 is a flow chart of a model acquisition method according to an exemplary embodiment. Figure 2 As shown, the following steps are included:

[0082] In step S21, an anti-fraud customer relationship domain knowledge graph is established based on the collected first historical user feature data;

[0083] In some embodiments, the execution subject of the embodiments of the present disclosure may be, for example, a network device, and the name of the network device is not limited. For example, the network device may also be called a server, a base station, or a server cluster, etc. The embodiments of the present disclosure are not limited to this.

[0084] The specific process is as described above and will not be repeated here.

[0085] In some embodiments, Figure 3 is an exemplary schematic diagram showing a model acquisition method according to an exemplary embodiment. Figure 3 As shown, it includes collecting user data, data preprocessing, feature engineering construction, relationship graph acquisition, etc. The details are as follows:

[0086] The user characteristic data may include, for example, customer registration information, a credit scoring matrix, past risk events and risk levels, user social relationships, etc. The credit scoring matrix may be used, for example, to score the user identity, wherein the credit scoring matrix does not specifically refer to a fixed matrix. For example, when the number of rows or columns corresponding to the credit scoring matrix changes, the credit scoring matrix may also change accordingly.

[0087] According to some embodiments, after establishing the anti-fraud customer relationship domain knowledge graph based on the collected first historical user feature data, the method further includes:

[0088] Displaying the anti-fraud customer relationship field knowledge graph through a relationship graph, and obtaining a display image corresponding to the anti-fraud customer relationship field knowledge graph;

[0089] The display image is encrypted using a piecewise linear chaotic encryption algorithm and stored. Therefore, the relationship between user-feature and user-group can be reflected in the form of a relationship graph. Since the security of the relationship graph is higher than that of a graph, table or text data, the security and storage efficiency of data storage can be improved.

[0090] The piecewise linear chaotic encryption algorithm can be, for example, as shown in formula (1):

[0091]

[0092] Among them, x∈I=[0,1), p∈(0,1 / 2), p is the control parameter, and F is a piecewise linear chaotic map implemented under L-biit finite precision. The relationship graph is converted into binary data, and the bit-by-bit XOR calculation is performed with the quantized chaotic binary sequence, so that the chaotic signal and the encrypted information signal are intertwined to cover the visibility of the information, which can improve the security of data storage, and binary data storage is simple and efficient.

[0093] According to some embodiments, when decrypting, the same initial value and control parameter can be used to generate the same chaotic sequence, which is bit-by-bit XORed with the received encrypted signal, and the result output is restored to obtain the original data. Among them, for the chaotic sequence iteratively generated by formula (1), any two point pairs (x(t), x(t+1)) and (x(t′), x(t′+1)) located on the same segment can be obtained, and the parameter p=(x(t′)-x(t)) / (x(t′+1)-x(t+1)) can be determined. In order to prevent the parameter p from being cracked, the general one-time one-key method for stream cipher encryption can be changed, and the method of periodically changing the control parameter p during the encryption process can be adopted to enhance the encryption effect.

[0094] According to some embodiments, data preprocessing and feature engineering construction may include, for example: establishing an anti-fraud customer relationship domain knowledge graph based on the collected first historical user feature data, including:

[0095] Performing data preprocessing on the collected first historical user feature data to obtain second historical user feature data, wherein the data preprocessing includes feature labeling processing, preclassification processing, initialization of boundary numbers and community modularization metric value Q processing;

[0096] The second historical user characteristic data is processed by using a community discovery algorithm, the initialization boundary number and the community modularity measurement value are removed, and the third historical user characteristic data is obtained, and the third historical user characteristic data is encrypted, and the encrypted third historical user characteristic data is stored in the blockchain node. Therefore, the historical user characteristic data can be processed to improve the accuracy of data acquisition, improve the accuracy of model training, and improve the accuracy of fraud risk information prediction.

[0097] Among them, when the third user data is obtained, the data can be stored in the blockchain node after being hashed and encrypted, and the data quality can be improved through classification, and the data can be stored in the blockchain node for encrypted storage to improve the security of data storage. Among them, the blockchain can be, for example, a distributed database technology for recording transactions and events of digital information. It is a decentralized, public, secure and traceable database without a single control center. Each node has a copy and can be verified and updated. The blockchain can be used in application scenarios such as storing data, verifying identity, executing smart contracts and managing digital assets. Among them, the blockchain can be, for example, to store data in a decentralized manner on multiple nodes, and to ensure the security and credibility of the data through encryption algorithms and consensus mechanisms. Each node has its own account book, and all account books are the same, and the consistency and reliability of the data are guaranteed through a consensus mechanism. The data structure of the blockchain is a chain composed of multiple blocks, each of which contains multiple transaction information and the hash value of the previous block, forming an unalterable transaction record chain.

[0098] In some embodiments, the community discovery algorithm may be, for example, an algorithm that uses various statistical indicators of the network to mine communities with strong internal relationships in the network. Among them, the community discovery algorithm may include, for example, the Girvan-Newman (GN) algorithm. Taking the GN algorithm as an example, the number of boundaries in the initial complex relationship network (the number of shortest paths passing through the edge in the shortest paths between all nodes) and the Q value of the community modularity metric (modularity, for example, can be used to indicate a method for measuring the quality of network community division), remove the edge with the highest number of boundaries, and recalculate the Q value of the current network. If the Q value is larger than the original one, the current Q value and network segmentation are updated, otherwise, the next network segmentation is performed, and repeated until the network segmentation is completed. The nodes in each community after segmentation have a high similarity. By finding the distribution of fraudulent nodes in the segmented community, other suspected fraudulent people with high similarity are mined.

[0099] According to some embodiments, the method further comprises:

[0100] Using random mutation operations, multiple simulations of user operations are performed to generate a set of mutation prediction links;

[0101] Adding the variation prediction link set to the first historical user data to obtain fourth historical user data;

[0102] The fourth historical user data is used to repeatedly execute the steps of establishing the knowledge graph in the field of anti-fraud customer relationships. Therefore, the random mutation operation is used to randomly inject features into the knowledge graph, which can continuously fill the mutation prediction link set, continuously train the model, improve the model's rapid response capability, and continuously self-improve the integrity of the knowledge graph, which can improve the model's prediction accuracy and robustness.

[0103] According to some embodiments, the method further comprises:

[0104] The variation prediction link set is converted into binary data by using piecewise linear chaotic mapping, and the binary data is stored.

[0105] Specifically, for example, in order to determine whether the randomly injected features are effective, after injecting a series of risk features, a series of future risk user portraits will be generated in a spring-like manner, which is called a mutation prediction link set. The mutation operations marked as high-risk by the AM-GCN model are called "killed mutants"; after mutation injection, the undiscovered mutation operations are called "survived mutants", among which the group where the "killed mutants" are located is usually a high-risk group, while the group where the "survived mutants" are located is often a low-risk group.

[0106] Among them, after obtaining the variation prediction link set, it is stored in the blockchain node after chaos encryption, which can improve the data storage efficiency. The specific storage method is as above and will not be repeated here.

[0107] In step S22, the feature set included in the anti-fraud customer relationship field knowledge graph is classified into groups to obtain group features;

[0108] The specific process is as described above and will not be repeated here.

[0109] According to some embodiments, group characteristics may include, for example, behavior, transaction, time sequence, and high frequency, which are not limited in the embodiments of the present disclosure.

[0110] In step S23, the initial graph neural network model is used to learn the group characteristics and individual characteristics in the anti-fraud customer relationship field knowledge graph to obtain the correlation characteristics between the group characteristics and the individual characteristics;

[0111] The specific process is as described above and will not be repeated here.

[0112] According to some embodiments, the initial graph neural network model is used to learn the group characteristics and individual characteristics in the anti-fraud customer relationship field knowledge graph to obtain a target graph neural network model, including:

[0113] Using an initial graph neural network model to learn the group characteristics and individual characteristics in the anti-fraud customer relationship field knowledge graph, and obtaining the weights of each data source corresponding to the first historical user feature data;

[0114] According to the weights of the data sources, the target graph neural network model is obtained. Therefore, the features in the knowledge graph can be learned. AM-GCN adaptively learns the weights of different source information through the attention mechanism, greatly improving the ability of the graph convolutional network to integrate node features and topological structures, and enhancing the model's ability to model graph structures. It can increase the weights of features from the same group and improve the accuracy of the graph neural network model in predicting relationship links.

[0115] Among them, the graph neural network model can be, for example, an adaptive multi-channel graph convolutional network (AM-GCN). AM-GCN is based on node embedding of node features, topological structures and their combination, and predicts feature labels and relationship links according to the similarity between features and structures, which can improve the accuracy of prediction.

[0116] In step S24, when the associated features meet the feature requirements, a target graph neural network model is obtained;

[0117] The specific process is as described above and will not be repeated here.

[0118] In step S25, user characteristic data is obtained;

[0119] The specific process is as described above and will not be repeated here.

[0120] According to some embodiments, the user characteristic data may refer to data to be predicted, for example. The user characteristic data does not specifically refer to a fixed data. For example, when the data type included in the user characteristic data changes, the user characteristic data may also change accordingly.

[0121] In step S26, the target graph neural network model is used to identify the user feature data to obtain relationship link information corresponding to the user feature data;

[0122] The specific process is as described above and will not be repeated here.

[0123] In some embodiments, the target graph neural network model may be, for example, a model that has been trained and can be used to predict fraud risk information, and the target graph neural network model does not specifically refer to a fixed model. For example, when the feature requirements change, the target graph neural network model may also change accordingly.

[0124] In some embodiments, the target graph neural network model may be used to identify the user feature data to obtain relationship link information corresponding to the user feature data.

[0125] In step S27, the fraud risk information corresponding to the user feature data is determined based on the relationship link information and the variation prediction link set.

[0126] The specific process is as described above and will not be repeated here.

[0127] In some embodiments, for example, the links in the predicted link set can be compiled based on the acquired relationship link matching, and the fraud risk information corresponding to the user feature data can be determined based on the matched links and the mutated predicted link set.

[0128] The fraud risk information can be used to indicate the risk probability of fraud for the user. The fraud risk information does not specifically refer to a fixed information. For example, when the target graph neural network model changes, the fraud risk information can also change accordingly.

[0129] For example, the target graph neural network model can be updated according to the continuously improved or generated variation prediction link set.

[0130] In some embodiments, for example, based on the acquired fraud risk information, a risk warning may be issued to the user or corresponding risk control measures may be adopted, which is not limited in the embodiments of the present disclosure.

[0131] Examples of fraudulent situations include:

[0132] 1. Risk of white household registration. The borrower’s information is missing and there is not enough data to conduct risk assessment on the borrower.

[0133] 2. Black account risk: The borrower has a record of overdue payment, dishonesty, or fraud.

[0134] 3. Malicious fraud: The borrower deliberately defrauds a loan by falsifying documents.

[0135] 4. False identity: Impersonating another person to commit fraud and obtain loans.

[0136] 5. Loans to pay off loans: Maintaining bad consumption by borrowing from Peter to pay Paul.

[0137] 6. Credit fraud: false identity, gang attack, individual loan fraud, credit risk, intermediary packaging. Different fraud situations may correspond to different risk control measures, which are not limited in the embodiments of the present disclosure.

[0138] In some or related embodiments, user feature data is obtained; the target graph neural network model is used to identify the user feature data to obtain relationship link information corresponding to the user feature data; and fraud risk information corresponding to the user feature data is determined based on the relationship link information and the variation prediction link set. Therefore, the target graph neural network model can be used for data identification to improve the accuracy of fraud risk information acquisition, improve the accuracy of risk control, and thus improve service quality.

[0139] Figure 4 is a block diagram of a model acquisition device according to an exemplary embodiment. Figure 4 , the model acquisition device 400 includes:

[0140] A graph building unit 401 is used to build an anti-fraud customer relationship domain knowledge graph based on the collected first historical user feature data;

[0141] A feature acquisition unit 402 is used to classify the feature set included in the anti-fraud customer relationship field knowledge graph into groups to obtain group features;

[0142] The feature acquisition unit 402 is further used to learn the group features and individual features in the anti-fraud customer relationship domain knowledge graph using an initial graph neural network model to acquire correlation features between the group features and the individual features;

[0143] The model acquisition unit 403 is used to acquire the target graph neural network model when the associated features meet the feature requirements.

[0144] According to some embodiments, the map creation unit 401 is further configured to:

[0145] Obtain user feature data;

[0146] Using the target graph neural network model to identify the user feature data, and obtaining relationship link information corresponding to the user feature data;

[0147] Fraud risk information corresponding to the user feature data is determined based on the relationship link information and the variation prediction link set.

[0148] According to some embodiments, the graph establishing unit 401 is further configured to, after establishing the anti-fraud customer relationship domain knowledge graph based on the collected first historical user feature data:

[0149] Displaying the anti-fraud customer relationship field knowledge graph through a relationship graph, and obtaining a display image corresponding to the anti-fraud customer relationship field knowledge graph;

[0150] The display image is encrypted using a piecewise linear chaotic encryption algorithm, and the display image is stored.

[0151] According to some embodiments, the graph establishing unit 401 is used to establish the anti-fraud customer relationship domain knowledge graph based on the collected first historical user feature data, specifically for:

[0152] Performing data preprocessing on the collected first historical user feature data to obtain second historical user feature data, wherein the data preprocessing includes feature labeling processing, preclassification processing, initialization of boundary numbers and community modularization metric processing;

[0153] The second historical user feature data is processed by using a community discovery algorithm to remove the initialization boundary number and the measurement value of the community modularity to obtain the third historical user feature data, and the third historical user feature data is encrypted, and the encrypted third historical user feature data is stored in the blockchain node.

[0154] According to some embodiments, the model acquisition unit 403 is used to learn the group characteristics and individual characteristics in the anti-fraud customer relationship field knowledge graph using the initial graph neural network model, and to obtain the target graph neural network model, specifically for:

[0155] Using an initial graph neural network model to learn the group characteristics and individual characteristics in the anti-fraud customer relationship field knowledge graph, and obtaining the weights of each data source corresponding to the first historical user feature data;

[0156] According to the weights of the data sources, a target graph neural network model is obtained.

[0157] According to some embodiments, the map creation unit 401 is further configured to:

[0158] Using random mutation operations, multiple simulations of user operations are performed to generate a set of mutation prediction links;

[0159] Adding the variation prediction link set to the first historical user data to obtain fourth historical user data;

[0160] Using the fourth historical user data, repeat the step of establishing an anti-fraud customer relationship field knowledge graph.

[0161] According to some embodiments, the map creation unit 401 is further configured to:

[0162] The variation prediction link set is converted into binary data by using piecewise linear chaotic mapping, and the binary data is stored.

[0163] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0164] In some or related embodiments, a graph establishment unit is used to establish an anti-fraud customer relationship domain knowledge graph based on the collected first historical user feature data; a feature acquisition unit is used to classify the feature set included in the anti-fraud customer relationship domain knowledge graph into groups and obtain group features; the feature acquisition unit is also used to learn the group features and individual features in the anti-fraud customer relationship domain knowledge graph using an initial graph neural network model to obtain the correlation features between the group features and the individual features; a model acquisition unit is used to obtain a target graph neural network model when the correlation features meet the feature requirements. Therefore, the correlation relationship between historical user feature data, the hidden relationship between user feature data, and the tracking of user features through the knowledge graph can reduce the situation where the model training accuracy is poor due to the single feature selection, improve the accuracy of model training, and improve the accuracy of fraud risk prediction.

[0165] Figure 5 is a block diagram of a network device 500 provided in an embodiment of the present disclosure. For example, the network device 500 may be provided as a network side device. Figure 5 The network device 500 includes a processing component 522, which further includes at least one processor, and a memory resource represented by a memory 532 for storing instructions executable by the processing component 522, such as an application. The application stored in the memory 532 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 522 is configured to execute instructions to perform any method of the aforementioned method applied to the network side device.

[0166] The network device 500 may also include a power supply component 527 configured to perform power management of the network device 500, a wired or wireless network interface 550 configured to connect the network device 500 to a network, and an input / output (I / O) interface 558. The network device 500 may operate based on an operating system stored in the memory 532, such as Windows Server TM, Mac OS X TM, Unix TM, Linux TM, Free BSD TM or the like.

[0167] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0168] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable model acquisition device, so that the program code, when executed by the processor or controller, implements the functions / operations specified in the flow chart and / or block diagram. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0169] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or 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 foregoing.

[0170] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0171] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain network.

[0172] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server may also be a server of a distributed system, or a server combined with a blockchain.

[0173] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0174] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A model acquisition method, characterized in that: include: Based on the collected first historical user feature data, a knowledge graph in the field of anti-fraud customer relations is established; Classifying the feature set included in the anti-fraud customer relationship field knowledge graph into groups to obtain group features; Using an initial graph neural network model to learn the group characteristics and individual characteristics in the anti-fraud customer relationship field knowledge graph, and obtaining correlation characteristics between the group characteristics and the individual characteristics; When the associated features meet the feature requirements, the target graph neural network model is obtained.

2. The method according to claim 1, characterized in that The method further comprises: Obtain user feature data; Using the target graph neural network model to identify the user feature data, and obtaining relationship link information corresponding to the user feature data; Fraud risk information corresponding to the user feature data is determined based on the relationship link information and the variation prediction link set.

3. The method according to claim 1, characterized in that: After establishing the anti-fraud customer relationship domain knowledge graph based on the collected first historical user feature data, the method further includes: Displaying the anti-fraud customer relationship field knowledge graph through a relationship graph, and obtaining a display image corresponding to the anti-fraud customer relationship field knowledge graph; The display image is encrypted using a piecewise linear chaotic encryption algorithm, and the display image is stored.

4. The method according to claim 1, characterized in that: The method of establishing an anti-fraud customer relationship domain knowledge graph based on the collected first historical user feature data includes: Performing data preprocessing on the collected first historical user feature data to obtain second historical user feature data, wherein the data preprocessing includes feature labeling processing, preclassification processing, initialization of boundary numbers and community modularization metric processing; The second historical user feature data is processed by using a community discovery algorithm to remove the initialization boundary number and the measurement value of the community modularity to obtain the third historical user feature data, and the third historical user feature data is encrypted, and the encrypted third historical user feature data is stored in the blockchain node.

5. The method according to claim 1, characterized in that The initial graph neural network model is used to learn the group characteristics and individual characteristics in the anti-fraud customer relationship field knowledge graph to obtain a target graph neural network model, including: Using an initial graph neural network model to learn the group characteristics and individual characteristics in the anti-fraud customer relationship field knowledge graph, and obtaining the weights of each data source corresponding to the first historical user feature data; According to the weights of the data sources, a target graph neural network model is obtained.

6. The method according to claim 1, characterized in that The method further comprises: Using random mutation operations, multiple simulations of user operations are performed to generate a set of mutation prediction links; Adding the variation prediction link set to the first historical user data to obtain fourth historical user data; Using the fourth historical user data, repeat the step of establishing an anti-fraud customer relationship field knowledge graph.

7. The method according to claim 6, characterized in that The method further comprises: The variation prediction link set is converted into binary data by using piecewise linear chaotic mapping, and the binary data is stored.

8. A model acquisition device, characterized in that: include: A graph building unit, used to build an anti-fraud customer relationship domain knowledge graph based on the collected first historical user feature data; A feature acquisition unit, used to classify the feature set included in the anti-fraud customer relationship field knowledge graph into groups to obtain group features; The feature acquisition unit is further used to learn the group features and individual features in the anti-fraud customer relationship field knowledge graph using an initial graph neural network model to acquire correlation features between the group features and the individual features; The model acquisition unit is used to acquire the target graph neural network model when the associated features meet the feature requirements.

9. A network device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the model acquisition method as described in any one of claims 1 to 7.

10. A storage medium, when instructions in the storage medium are executed by a processor of a network device, enables the network device to execute the model acquisition method according to any one of claims 1 to 7.

Images