Risk object identification method and device, storage medium and electronic equipment

Abstract

The invention discloses a risk object identification method and device, a storage medium and electronic equipment. The method comprises the steps of obtaining transaction data of a target object in a target financial institution; processing the transaction data through a risk identification model in the target financial institution to obtain an identification result, the risk identification model being obtained by training historical transaction data of a plurality of financial institutions based on a federated learning framework, model parameters of the risk identification model in the federated learning process are determined based on at least one of data distribution information of historical transaction data of the plurality of financial institutions and the data volume of the historical transaction data; and determining target information according to the identification result, wherein the target information is used for representing whether the target object is the object with the risk transaction. Through the method and the device, the problem of low recognition accuracy when a user with risk transaction is recognized through a risk recognition model due to the fact that a small amount of transaction data is adopted to train the risk recognition model in the related technology is solved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and more specifically, to a method, device, storage medium, and electronic device for identifying risky objects. Background Art

[0002] In the field of FinTech, the accuracy and reliability of transaction risk identification are crucial for ensuring the security of financial transactions. Related technologies typically use risk identification models to identify users engaging in risky transactions. However, many current risk identification models face data shortages during training, specifically limited transaction data samples. This results in low accuracy when using risk identification models to identify users engaging in risky transactions.

[0003] Currently, no effective solution has been proposed to the above-mentioned problems in related technologies. Summary of the Invention

[0004] The main purpose of this application is to provide a method, device, storage medium and electronic device for identifying risk objects, so as to solve the problem in related technologies of using a small amount of transaction data to train a risk identification model, resulting in low recognition accuracy when the risk identification model is used to identify users who have engaged in risky transactions.

[0005] To achieve the above-mentioned objectives, according to one aspect of the present application, a method for identifying risky objects is provided. The method comprises: obtaining transaction data of a target object in a target financial institution; processing the transaction data using a risk identification model in the target financial institution to obtain an identification result, wherein the risk identification model is based on a federated learning framework and is trained using historical transaction data from multiple financial institutions, and during the federated learning process, model parameters of the risk identification model are determined based on at least one of the following: data distribution information of the historical transaction data of the multiple financial institutions, data volume of the historical transaction data of the multiple financial institutions, the multiple financial institutions including the target financial institution, and the identification result represents a probability value that the target object has a risky transaction; and determining target information based on the identification result, wherein the target information is used to represent whether the target object is an object of risky transactions.

[0006] Furthermore, the federated learning framework includes local models corresponding to the clients of each financial institution, and the local models are initial risk identification models, wherein the method for identifying risk objects also includes: randomly selecting K financial institutions from multiple financial institutions; for each of the K financial institutions, obtaining the parameters of the global model from the target server through the client of the financial institution, and updating the local model based on the parameters of the global model to obtain a first risk identification model, wherein the model structure of the global model and the local model are the same; training the first risk identification model based on the historical transaction data in the financial institution to obtain a second risk identification model; determining the target parameters of the global model based on the model parameters of the second risk identification model in the K financial institutions, and at least one of the data distribution information and the data volume; if the preset iteration condition is not met, repeating the step of randomly selecting K financial institutions from multiple financial institutions until the iteration condition is met, and updating the parameters of the local model in the target financial institution based on the target parameters of the global model to obtain a risk identification model.

[0007] Furthermore, the method for identifying risk objects also includes: in each iterative round, calculating the updated gradient value of the model parameters of the first risk identification model based on the preset loss function and the identification result of the first risk identification model; adding differential privacy noise to the updated gradient value to obtain the target gradient value; and updating the parameters of the first risk identification model based on the target gradient value.

[0008] Furthermore, the method for identifying risk objects also includes: before determining the target parameters of the global model based on the model parameters of the second risk identification model in K financial institutions, and at least one of the data distribution information and the data volume, for each financial institution among the multiple financial institutions, generating a histogram of the historical transaction data of the financial institution in multiple feature dimensions; for each histogram of the financial institution, determining the probability distribution information of the histogram based on the data volume corresponding to each interval in the histogram, wherein the probability distribution information includes the probability distribution value corresponding to each interval; and determining the data distribution information of the financial institution based on the probability distribution information of each histogram of the financial institution.

[0009] Furthermore, the method for identifying risk objects also includes: averaging the probability distribution values ​​of the same interval belonging to the same feature dimension in the data distribution information of K financial institutions to obtain target data distribution information; for each of the K financial institutions, determining a first weight based on the relative entropy between the data distribution information of the financial institution and the target data distribution information; determining a second weight based on the data volume of the historical transaction data of the financial institution and the total data volume of the historical transaction data of the K financial institutions; determining the target weight of the financial institution based on at least one of the first weight and the second weight, and determining the target parameters based on the target weights of the K financial institutions and the model parameters of the second risk identification model.

[0010] Furthermore, the risk object identification method also includes: performing weighted sum calculation on the first weight and the second weight based on a preset weight factor combination to obtain a target weight, or determining the first weight as the target weight, or determining the second weight as the target weight.

[0011] Furthermore, the method for identifying risky objects also includes: when the probability value in the identification result is greater than a preset probability value, determining that the target information represents that there is a risky transaction with the target object; when the probability value in the identification result is less than or equal to the preset probability value, determining that the target information represents that there is no risky transaction with the target object.

[0012] To achieve the above-mentioned objectives, according to another aspect of the present application, a device for identifying risky objects is provided. The device includes: an acquisition module for acquiring transaction data of a target object in a target financial institution; a first processing module for processing the transaction data using a risk identification model in the target financial institution to obtain an identification result, wherein the risk identification model is based on a federated learning framework and is trained using historical transaction data from multiple financial institutions, and during the federated learning process, model parameters of the risk identification model are determined based on at least one of the following: data distribution information of the historical transaction data of the multiple financial institutions, data volume of the historical transaction data of the multiple financial institutions, the multiple financial institutions including the target financial institution, and the identification result represents a probability value that the target object has a risky transaction; and a first determination module for determining target information based on the identification result, wherein the target information is used to represent whether the target object is an object that has a risky transaction.

[0013] Furthermore, the federated learning framework includes local models corresponding to the clients of each financial institution, and the local models are initial risk identification models, wherein the risk object identification device also includes: a selection module for randomly selecting K financial institutions from multiple financial institutions; a first update module for obtaining the parameters of the global model from the target server through the client of each financial institution in the K financial institutions, and updating the local model based on the parameters of the global model to obtain a first risk identification model, wherein the model structure of the global model is the same as that of the local model; a training module for training the first risk identification model based on the historical transaction data in the financial institution to obtain a second risk identification model; a second determination module for determining the target parameters of the global model based on the model parameters of the second risk identification model in the K financial institutions, and at least one of the data distribution information and the data volume; a second processing module for repeatedly executing the step of randomly selecting K financial institutions from multiple financial institutions until the iteration condition is met, and updating the parameters of the local model in the target financial institution according to the target parameters of the global model to obtain a risk identification model.

[0014] Furthermore, the risk object identification device also includes: a calculation module, which is used to calculate the updated gradient value of the model parameters of the first risk identification model in each iteration round based on the preset loss function and the identification result of the first risk identification model; a third processing module, which is used to add differential privacy noise to the updated gradient value to obtain the target gradient value; and a second update module, which is used to update the parameters of the first risk identification model based on the target gradient value.

[0015] Furthermore, the risk object identification device also includes: a generation module, which is used to generate a histogram of the historical transaction data of a financial institution in multiple feature dimensions for each financial institution among a plurality of financial institutions; a third determination module, which is used to determine the probability distribution information of the histogram for each histogram of the financial institution based on the amount of data corresponding to each interval in the histogram, wherein the probability distribution information includes the probability distribution value corresponding to each interval; and a fourth determination module, which is used to determine the data distribution information of the financial institution based on the probability distribution information of each histogram of the financial institution.

[0016] Furthermore, the second determination module also includes: a calculation submodule, which is used to average the probability distribution values ​​of the same interval belonging to the same feature dimension in the data distribution information of K financial institutions to obtain target data distribution information; a first determination submodule, which is used to determine a first weight for each of the K financial institutions based on the relative entropy between the data distribution information of the financial institution and the target data distribution information; a second determination submodule, which is used to determine a second weight based on the data volume of the historical transaction data of the financial institution and the total data volume of the historical transaction data of the K financial institutions; a third determination submodule, which is used to determine the target weight of the financial institution based on at least one of the first weight and the second weight, and determine the target parameters based on the target weights of the K financial institutions and the model parameters of the second risk identification model.

[0017] Furthermore, the third determination submodule also includes: a first determination unit, used to perform weighted sum calculation on the first weight and the second weight based on a preset weight factor combination to obtain a target weight, or a second determination unit, used to determine the first weight as the target weight, or a third determination unit, used to determine the second weight as the target weight.

[0018] Furthermore, the first determination module also includes: a fourth determination submodule, which is used to determine that the target information represents the existence of a risky transaction for the target object when the probability value in the recognition result is greater than a preset probability value; and a fifth determination submodule, which is used to determine that the target information represents the existence of a risky transaction for the target object when the probability value in the recognition result is less than or equal to a preset probability value.

[0019] In order to achieve the above-mentioned purpose, according to another aspect of the present application, a computer-readable storage medium is provided, which includes a stored executable program, wherein when the executable program is running, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned risk object identification method.

[0020] In order to achieve the above-mentioned purpose, according to another aspect of the present application, an electronic device is provided, which includes a memory storing an executable program; and a processor for running the program, wherein the above-mentioned risk object identification method is executed when the program is running.

[0021] In an embodiment of the present application, a risk identification model for a target financial institution is designed based on a federated learning framework and trained with historical transaction data from multiple financial institutions. This allows for the target financial institution's risk identification model to be trained based on the historical transaction data of multiple financial institutions while effectively protecting the data privacy and security of each financial institution. This overcomes the limitations of training a model based on a single dataset, improves the model's training effectiveness, and increases the accuracy of identifying users with transaction risks. Furthermore, by determining model parameters based on at least one of the data distribution information and the data volume of the historical transaction data of multiple financial institutions, effective consideration is achieved for data differences between multiple financial institutions, thereby improving the accuracy of determining model parameters and the accuracy of identifying users with transaction risks.

[0022] It can be seen that the method provided in this application achieves the purpose of obtaining a model for risk identification based on the federated learning framework training, and determining the model parameters in the federated learning process based on the data information of historical data in the institution, achieving the technical effect of improving the recognition accuracy, and solving the technical problem that a small amount of transaction data is used in related technologies to train the risk identification model, resulting in low recognition accuracy when identifying users with risky transactions through the risk identification model. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0024] Figure 1 This is a hardware structure block diagram of a computer terminal provided according to an embodiment of the present application;

[0025] Figure 2 is a flow chart of a method for identifying risk objects according to an embodiment of the present application;

[0026] Figure 3 This is a schematic diagram of the relationship between the central server and multiple financial institutions provided in an embodiment of the present application;

[0027] Figure 4 is a schematic diagram of a risk object identification device provided according to an embodiment of the present application;

[0028] Figure 5 This is a structural block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0029] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0030] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0031] It should be noted that the collected information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this application are information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data comply with relevant laws, regulations and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation portals for users to choose to authorize or refuse. For example, an interface is set up between this system and relevant users or institutions to provide users with corresponding operation portals for users to choose to agree or refuse the automated decision-making results; if the user chooses to refuse, the expert decision-making process will be entered.

[0032] Example 1

[0033] According to an embodiment of the present application, an embodiment of a method for identifying a risk object is also provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0034] The method embodiment provided in the first embodiment of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1The following is a hardware block diagram of a computer terminal (or mobile device) for implementing a method for identifying risk objects. Figure 1 As shown, the computer terminal 10 (or mobile device) may include one or more (illustrated as 102a, 102b, ..., 102n in the figure) processors 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. It will be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components than shown, or with Figure 1 Different configurations shown.

[0035] It should be noted that the one or more processors 102 and / or other data processing circuits described above may generally be referred to herein as "data processing circuitry". The data processing circuitry may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuitry may be a single independent processing module, or may be incorporated in whole or in part into any of the other components of the computer terminal 10 (or mobile device). As described in the embodiments of the present application, the data processing circuitry serves as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0036] Memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the risk object identification method in the embodiments of the present application. Processor 102 executes the software programs and modules stored in memory 104 to perform various functional applications and data processing, thereby implementing the risk object identification method described above. Memory 104 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, memory 104 may further include memory remotely located relative to processor 102, and such remote memory may be connected to computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0037] The transmission device 106 is configured to receive or transmit data via a network. A specific example of the aforementioned network may include a wireless network provided by the communications provider of the computer terminal 10. In one embodiment, the transmission device 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to enable communication with the Internet. In another embodiment, the transmission device 106 may be a radio frequency (RF) module, which is configured to communicate with the Internet wirelessly.

[0038] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 (or mobile device).

[0039] Under the above operating environment, this application provides Figure 2 The identification method of risk objects is shown. Figure 2 This is a flowchart of a method for identifying risk objects provided in an embodiment of the present application.

[0040] Step S201: Acquire transaction data of a target object in a target financial institution.

[0041] Optionally, electronic devices, application systems, servers and other devices may be used as the execution subject of the present application. In this embodiment, the target processing system is used as the execution subject to execute the above-mentioned risk object identification method.

[0042] Optionally, the target processing system includes at least a target server and clients of each of the multiple financial institutions. The target server may also be referred to as a central server, a host server, etc. The multiple financial institutions include the target financial institution. The multiple financial institutions may not have any relationship with each other, or the multiple financial institutions may be subordinate units of the same parent financial institution. For example, Figure 3 This is a schematic diagram of the relationship between the central server and multiple financial institutions provided in an embodiment of the present application. Figure 3 As shown, the central server can transmit data (such as model parameters, etc.) to multiple financial institutions.

[0043] Optionally, transaction data of target users within the target financial institution can be obtained through the target financial institution's client. Target users may be users whose transactions are to be assessed for risk. Transaction data may include the target user's transaction order's transaction time, location, amount, and counterparty.

[0044] Step S202: Process the transaction data through the risk identification model in the target financial institution to obtain an identification result, wherein the risk identification model is based on a federated learning framework and is trained by historical transaction data of multiple financial institutions, and in the federated learning process, the model parameters of the risk identification model are determined based on at least one of the following: data distribution information of the historical transaction data of multiple financial institutions, data volume of the historical transaction data of multiple financial institutions, the multiple financial institutions include the target financial institution, and the identification result represents the probability value of the target object having a risky transaction.

[0045] Optionally, the risk identification model may be a neural network model. The client of the target financial institution may input transaction data into the risk identification model of the target financial institution to obtain an identification result, wherein the identification result represents a probability value of a risky transaction involving the target object.

[0046] Optionally, the risk identification model is based on a federated learning framework and is trained using historical transaction data from multiple financial institutions. The federated learning framework is a distributed machine learning technology that allows collaborative training among different institutions without the need for direct data exchange. Each participating institution (i.e., financial institution) trains the model on a local dataset (i.e., local historical transaction data), and then shares only its own model update information (e.g., gradients, or model parameters) with the central server, which then aggregates and updates the global model. Historical transaction data in a financial institution may refer to transaction data corresponding to users in the financial institution who have been identified as having engaged in risky transactions. The content and format of the historical transaction data are the same as those of the target object's transaction data, so they will not be described in detail here.

[0047] For example, a target processing system can perform federated learning in the following ways:

[0048] Step 1: Establish a central server to coordinate the learning process, which can be hosted by a trusted third party or one of the federated learning participants.

[0049] Step 2: Determine the clients participating in federated learning, i.e., the clients of the multiple financial institutions mentioned above. Assume that there are M clients, corresponding to D = {D1, D2, ..., D M} datasets.

[0050] Step 3: For multiple financial institutions, select a local model suitable for the risk identification task, such as a deep neural network. Initialize the parameters of the global model on the central server. The global model has the same model structure as the local model.

[0051] Step 4: Each financial institution (client) performs preparatory work such as data cleaning, feature extraction, and labeling on local data.

[0052] Step 5: Federated learning training is performed on each participating financial institution based on the prepared data. During the training process, the parameters of the global model on the central server are continuously updated, and the parameters of the final global model are determined as the parameters required for the local models of each financial institution.

[0053] Optionally, during the federated learning process, model parameters of the risk identification model are determined based on at least one of: data distribution information of historical transaction data of multiple financial institutions, or data volume of historical transaction data of multiple financial institutions. For example, target weights corresponding to financial institutions are determined based on at least one of the aforementioned data distribution information and data volume, and then the parameters of the global model on the central server are updated based on the target weights of each financial institution and the model parameters of the independently trained models of each financial institution.

[0054] In an optional embodiment, the data distribution information of the historical transaction data of a financial institution may refer to the distribution of the historical transaction data in multiple feature dimensions, such as transaction amount distribution, transaction time distribution, etc.

[0055] In an optional embodiment, the data volume of the historical transaction data of a financial institution refers to the amount of historical transaction data provided by the financial institution, and the size of the data volume affects the training effect and generalization ability of the model.

[0056] Step S203: determining target information according to the recognition result, wherein the target information is used to indicate whether the target object is an object of risky transactions.

[0057] Optionally, the target processing system may compare the recognition result with a preset threshold to obtain a comparison result, and thereby determine whether the target object is an object of risky transactions based on the comparison result.

[0058] In an embodiment of the present application, a risk identification model for a target financial institution is designed based on a federated learning framework and trained with historical transaction data from multiple financial institutions. This allows for the target financial institution's risk identification model to be trained based on the historical transaction data of multiple financial institutions while effectively protecting the data privacy and security of each financial institution. This overcomes the limitations of training a model based on a single dataset, improves the model's training effectiveness, and increases the accuracy of identifying users with transaction risks. Furthermore, by determining model parameters based on at least one of the data distribution information and the data volume of the historical transaction data of multiple financial institutions, effective consideration is achieved for data differences between multiple financial institutions, thereby improving the accuracy of determining model parameters and the accuracy of identifying users with transaction risks.

[0059] It can be seen that the method provided in this application achieves the purpose of obtaining a model for risk identification based on the federated learning framework training, and determining the model parameters in the federated learning process based on the data information of historical data in the institution, achieving the technical effect of improving the recognition accuracy, and solving the technical problem that a small amount of transaction data is used in related technologies to train the risk identification model, resulting in low recognition accuracy when identifying users with risky transactions through the risk identification model.

[0060] Optionally, in the risk object identification method provided in the embodiment of the present application, the federated learning framework includes local models corresponding to the clients of each financial institution, and the local model is an initial risk identification model, wherein the risk identification model is obtained in the following manner: randomly selecting K financial institutions from multiple financial institutions; for each of the K financial institutions, obtaining the parameters of the global model from the target server through the client of the financial institution, and updating the local model based on the parameters of the global model to obtain a first risk identification model, wherein the global model and the local model have the same model structure; training the first risk identification model based on the historical transaction data in the financial institution to obtain a second risk identification model; determining the target parameters of the global model based on the model parameters of the second risk identification model in the K financial institutions, and at least one of the data distribution information and the data volume; if the preset iteration condition is not met, repeating the step of randomly selecting K financial institutions from multiple financial institutions until the iteration condition is met, and updating the parameters of the local model in the target financial institution based on the target parameters of the global model to obtain a risk identification model.

[0061] Optionally, the target processing system may randomly select K financial institutions from multiple financial institutions, where K is a positive integer greater than 1. This is done to ensure the randomness and distribution of the federated learning process and to avoid skewness in the global model due to the continued participation of certain specific institutions.

[0062] Optionally, once K financial institutions are selected, the client of each participating financial institution can obtain the parameters of the global model from the central server (i.e., the server side). Among them, the global model is the core of the federated learning framework. It is a model maintained by the central server and learned based on the data of all participating financial institutions. The local model refers to the model maintained locally by each financial institution. The global model and the local model have the same model structure. Both the global model and the local model are used to identify whether the user has risky transactions. By applying the parameters of the global model to the local model, the client of the financial institution can perform a preliminary update of the local model parameters to obtain the first risk identification model in the financial institution. This update process is designed to enable the local model to carry part of the learning results of the global model, so as to better start in local data training.

[0063] After the local model is updated, resulting in the primary risk identification model, participating financial institutions can further train it using their own historical transaction data to generate a secondary risk identification model. By performing multiple rounds of training on local datasets, financial institutions can effectively update model parameters, enabling the model to better learn risky transaction patterns within local data, thereby improving the model's personalized identification capabilities.

[0064] After the K financial institutions have completed training on the first risk identification model, the K financial institutions can upload the current model parameters in their local models to the central server, which then determines the target parameters of the global model based on the model parameters of the second risk identification model in the K financial institutions, as well as at least one of the data distribution information and the data volume. For example, the target weights corresponding to the K financial institutions can be determined based on at least one of the aforementioned data distribution information and the data volume, and then the target parameters of the global model on the central server can be determined based on the target weights of the K financial institutions and the model parameters of the second risk identification model.

[0065] Optionally, if the preset iteration conditions are not met, the federated learning cycle will be repeated until the preset iteration conditions are met, for example, a certain number of iterations is reached or the performance indicator of the global model (such as accuracy) is not lower than a predetermined threshold.

[0066] In each round of iteration, the above steps will be executed: randomly selecting K financial institutions from multiple financial institutions, updating the local models in the K financial institutions based on the parameters in the global model, training the local models based on historical transaction data, and updating the parameters in the global model based on the model parameters of the trained local models.

[0067] Optionally, when preset iteration conditions are met, the central server may use the target parameters of the finalized global model to update the model parameters of the local model of each of the multiple financial institutions, thereby obtaining the risk identification model for the target financial institution and the risk identification models for financial institutions other than the target financial institution. In this case, the model parameters of the risk identification models in each financial institution are the same.

[0068] In an optional embodiment, the target processing system may obtain the risk identification model in the following manner:

[0069] Assume that there are T rounds of global communication and E rounds of local iterations at the client side of each financial institution.

[0070] Step 1: At the beginning of each round of global communication, the central server randomly selects K clients to participate in this training and sends the current global model parameters to each financial institution participating in this round of iteration.

[0071] Step 2: Each financial institution participating in this round of training uses local data to perform E rounds of gradient descent iterations based on the received global model to update the parameters of the local model.

[0072] Step 3: The central server determines the target weight of the financial institution based on at least one of the data distribution information and the data volume, performs weighted aggregation on the model parameters sent by each client based on the target weight, and updates the global model parameters.

[0073] Step 4: Repeat steps 1 to 3 until a predetermined number of global communication rounds T or other stopping conditions are reached.

[0074] Step 5: Evaluate the global model using a subset of independent test datasets retained by the financial institution to assess its performance. This evaluation can include metrics such as accuracy, recall, and precision to quantify the model's performance in the risk identification task.

[0075] Step 6: After all financial institutions confirm that the model performance is satisfactory, the global model will be deployed in practical applications for risk identification.

[0076] It should be noted that through the above method, a risk identification model is effectively trained based on the federated learning framework, which effectively improves the training effect and thus improves the accuracy of identifying risky users.

[0077] Optionally, in the risk object identification method provided in an embodiment of the present application, in the process of training the first risk identification model based on the historical transaction data in the financial institution to obtain the second risk identification model, the method also includes: in each iterative round, calculating the updated gradient value of the model parameters of the first risk identification model according to the preset loss function and the identification result of the first risk identification model; adding differential privacy noise to the updated gradient value to obtain the target gradient value; and updating the parameters of the first risk identification model based on the target gradient value.

[0078] Optionally, during the training process, the target processing system may determine a training sample set based on historical transaction data from the financial institution, and then train the first risk identification model based on the training sample set. A training sample in the training sample set may be a set of historical transaction data corresponding to a user in the financial institution, and the true label of the training sample indicates whether the user has engaged in risky transactions.

[0079] Therefore, in each iteration, the first risk model can calculate the updated gradient values ​​of the model parameters of the first risk identification model based on the preset loss function, the recognition results of the first risk identification model, and the true labels in the training sample set. The recognition results of the first risk identification model represent the probability value of the user in the training sample engaging in risky transactions.

[0080] After obtaining the updated gradient value, in order to protect the privacy of local transaction data, differential privacy noise can be added to the gradient value to obtain the target gradient value. Differential privacy is a statistical method that adds random noise to the statistical results of real data, making it impossible for analysts to reversely infer relevant information from the statistical results, thereby ensuring data privacy.

[0081] For example, after obtaining the updated gradient value, in order to prevent the gradient explosion problem, the calculated updated gradient value is first clipped according to the gradient norm boundary. The clipping process can be expressed as the following formula:

[0082]

[0083] in, Represents the updated gradient value, and C represents the gradient norm boundary.

[0084] For another example, after obtaining the updated gradient value after clipping, privacy noise is added according to the following formula:

[0085]

[0086] in, represents the target gradient value, represents Laplace distributed noise.

[0087] Optionally, in each iteration round, after obtaining the target gradient value, the parameters of the first risk identification model may be updated based on the target gradient value, and the next iteration round may be performed after the parameters are updated.

[0088] It should be noted that by calculating the updated gradient and adding differential privacy noise, the training effect of the local model can be improved while protecting customer information, thereby improving the reliability of model training.

[0089] Optionally, in the risk object identification method provided in the embodiment of the present application, before determining the target parameters of the global model based on the model parameters of the second risk identification model in K financial institutions, and at least one of the data distribution information and the data volume, the method also includes: for each financial institution among the multiple financial institutions, generating a histogram of the historical transaction data of the financial institution in multiple feature dimensions; for each histogram of the financial institution, determining the probability distribution information of the histogram based on the data volume corresponding to each interval in the histogram, wherein the probability distribution information includes the probability distribution value corresponding to each interval; and determining the data distribution information of the financial institution based on the probability distribution information of each histogram of the financial institution.

[0090] Optionally, multiple feature dimensions include but are not limited to transaction amount, transaction time, transaction frequency, transaction location, and other transaction features related to risk identification.

[0091] A histogram is a statistical table used to visualize the distribution of data (i.e., historical transaction data) along a specific characteristic dimension. It depicts the data distribution by dividing the data into several intervals and calculating the number of data points within each interval. For example, a histogram used to visualize the distribution of historical transaction data along the transaction amount dimension might have multiple intervals such as 1-100 yuan, 100-1000 yuan, and 1000-10000 yuan.

[0092] Optionally, after generating the histogram, for each interval in each histogram, a probability distribution value for that interval can be calculated based on the volume of historical transaction data corresponding to that interval and the total volume of historical transaction data in the current financial institution. For example, the probability distribution value for that interval can be calculated by dividing the volume of historical transaction data corresponding to that interval by the total volume of historical transaction data in the current financial institution. Thus, for each histogram, the probability distribution information for that histogram is composed of the probability distribution values ​​corresponding to each interval in that histogram.

[0093] Optionally, after obtaining the probability distribution information of each histogram, the probability distribution information of all histograms in the financial institution is determined as the data distribution information of the financial institution. The data distribution information of the financial institution can be expressed as H k ={p1, p2, ..., p n}, where n is equal to the sum of the number of intervals corresponding to all feature dimensions, that is, the data distribution information of the financial institution includes the probability distribution value of each interval of each feature dimension.

[0094] In an optional embodiment, for each financial institution, after obtaining the financial institution's data distribution information, the client at the financial institution can add differential privacy noise to the data distribution information to obtain updated data distribution information, thereby preventing the leakage of user sensitive data. The updated data distribution information is then uploaded to the central server for the central server to determine the target weight of the financial institution.

[0095] It should be noted that, through the above method, the data distribution information of financial institutions is determined based on the data distribution of each financial institution in different characteristic dimensions, thereby improving the content richness of the data distribution information and facilitating more accurate determination of target weights in the subsequent period.

[0096] Optionally, in the risk object identification method provided in the embodiment of the present application, the target parameters of the global model are determined based on the model parameters of the second risk identification model in the K financial institutions, and at least one of the data distribution information and the data volume, including: averaging the probability distribution values ​​of the same interval belonging to the same feature dimension in the data distribution information of the K financial institutions to obtain the target data distribution information; for each of the K financial institutions, determining a first weight based on the relative entropy between the data distribution information of the financial institution and the target data distribution information; determining a second weight based on the data volume of the historical transaction data of the financial institution and the total data volume of the historical transaction data of the K financial institutions; determining the target weight of the financial institution based on at least one of the first weight and the second weight, and determining the target parameters based on the target weights of the K financial institutions and the model parameters of the second risk identification model.

[0097] Optionally, the central server can average the probability distribution values ​​of the same interval in the data distribution information of K financial institutions and obtain the target data distribution information. The target distribution information can be expressed as H avg ={q1, q2, ..., q n}, n is equal to the sum of the number of intervals corresponding to all feature dimensions.

[0098] Optionally, for each of the K financial institutions, a first weight is determined based on the relative entropy between the data distribution information of the financial institution and the target data distribution information. The first weight can be calculated based on the following method:

[0099]

[0100] Among them, α k1 represents the first weight, k represents the kth financial institution, Represents the relative entropy between the data distribution information of the k-th financial institution and the target data distribution information.

[0101] Optionally, for each of the K financial institutions, a second weight is determined based on the amount of historical transaction data of the financial institution and the total amount of historical transaction data of the K financial institutions. The second weight can be calculated based on the following method:

[0102]

[0103] Among them, α k2 represents the second weight, |D k | represents the amount of historical transaction data of the kth financial institution, ∑ k∈K |D k | represents the total amount of historical transaction data of K financial institutions.

[0104] After obtaining the first weight and the second weight, the target processing system determines the target weight α of the financial institution according to at least one of the first weight and the second weight. k , then determine the target parameters based on the target weights of the K financial institutions and the model parameters of the second risk identification model. For example, for each of the K financial institutions, the target weight of the financial institution is multiplied by the model parameters of the second risk identification model of the financial institution to obtain the first model parameters corresponding to the financial institution. The first model parameters of the K financial institutions are then summed to obtain the target parameters.

[0105] It should be noted that, through the above method, it is achieved to measure the importance of the model parameters of the financial institution based on at least one of the data distribution information and the data volume, thereby effectively improving the accuracy of determining the target parameters.

[0106] Optionally, in the risk object identification method provided in the embodiment of the present application, the target weight of the financial institution is determined based on at least one of the first weight and the second weight, including: performing weighted sum calculation on the first weight and the second weight based on a preset weight factor combination to obtain the target weight, or determining the first weight as the target weight, or determining the second weight as the target weight.

[0107] In an optional embodiment, the weight factor combination includes a first weight factor corresponding to the first weight and a second weight factor corresponding to the second weight. The central server may multiply the first weight factor by the first weight to obtain a first value, multiply the second weight factor by the second weight to obtain a second value, and then add the first value and the second value to obtain the target weight.

[0108] In an optional embodiment, the first weight is directly used as the target weight.

[0109] In an optional embodiment, the second weight is directly used as the target weight.

[0110] It should be noted that, through the above method, the target weight is accurately determined, thereby effectively improving the accuracy of determining the target parameters.

[0111] Optionally, in the risk object identification method provided in the embodiment of the present application, the target information is determined based on the identification result, including: when the probability value in the identification result is greater than the preset probability value, determining that the target information represents the existence of a risky transaction for the target object; when the probability value in the identification result is less than or equal to the preset probability value, determining that the target information represents the absence of a risky transaction for the target object.

[0112] Optionally, the preset probability value may be 0.5. For example, assuming the preset probability value is 0.5, if the probability value in the recognition result is 0.6, it is determined that the target information indicates that the target object has a risky transaction; if the probability value in the recognition result is 0.4, it is determined that the target information indicates that the target object has not a risky transaction.

[0113] It should be noted that the above method can accurately identify whether the user has engaged in risky transactions.

[0114] It can be seen that the method provided in this application achieves the purpose of obtaining a model for risk identification based on the federated learning framework training, and determining the model parameters in the federated learning process based on the data information of historical data in the institution, achieving the technical effect of improving the recognition accuracy, and solving the technical problem that a small amount of transaction data is used in related technologies to train the risk identification model, resulting in low recognition accuracy when identifying users with risky transactions through the risk identification model.

[0115] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0116] Example 2

[0117] The present application also provides a device for identifying a risk object. It should be noted that the device for identifying a risk object in the present application can be used to execute the method for identifying a risk object provided in the present application. The device for identifying a risk object provided in the present application is described below.

[0118] According to an embodiment of the present application, a device for identifying a risk object for implementing the above-mentioned method for identifying a risk object is also provided. Figure 4 As shown, the device includes:

[0119] An acquisition module 401 is used to acquire transaction data of a target object in a target financial institution;

[0120] A first processing module 402 is configured to process the transaction data using a risk identification model in a target financial institution to obtain an identification result, wherein the risk identification model is based on a federated learning framework and is trained using historical transaction data from multiple financial institutions, and wherein model parameters of the risk identification model are determined during the federated learning process based on at least one of: data distribution information of the historical transaction data from the multiple financial institutions, and data volume of the historical transaction data from the multiple financial institutions, wherein the multiple financial institutions include the target financial institution, and the identification result represents a probability value of the target object having a risky transaction;

[0121] The first determination module 403 is configured to determine target information according to the recognition result, wherein the target information is used to indicate whether the target object is an object of risky transactions.

[0122] In an embodiment of the present application, a risk identification model for a target financial institution is designed based on a federated learning framework and trained with historical transaction data from multiple financial institutions. This allows for the target financial institution's risk identification model to be trained based on the historical transaction data of multiple financial institutions while effectively protecting the data privacy and security of each financial institution. This overcomes the limitations of training a model based on a single dataset, improves the model's training effectiveness, and increases the accuracy of identifying users with transaction risks. Furthermore, by determining model parameters based on at least one of the data distribution information and the data volume of the historical transaction data of multiple financial institutions, effective consideration is achieved for data differences between multiple financial institutions, thereby improving the accuracy of determining model parameters and the accuracy of identifying users with transaction risks.

[0123] It can be seen that the method provided in this application achieves the purpose of obtaining a model for risk identification based on the federated learning framework training, and determining the model parameters in the federated learning process based on the data information of historical data in the institution, achieving the technical effect of improving the recognition accuracy, and solving the technical problem that a small amount of transaction data is used in related technologies to train the risk identification model, resulting in low recognition accuracy when identifying users with risky transactions through the risk identification model.

[0124] Optionally, in the risk object identification device provided in an embodiment of the present application, the federated learning framework includes a local model corresponding to the client of each financial institution, and the local model is an initial risk identification model, wherein the risk object identification device also includes: a selection module for randomly selecting K financial institutions from multiple financial institutions; a first update module for obtaining the parameters of the global model from the target server through the client of the financial institution for each of the K financial institutions, and updating the local model based on the parameters of the global model to obtain a first risk identification model, wherein the model structure of the global model is the same as that of the local model; a training module for training the first risk identification model based on the historical transaction data in the financial institution to obtain a second risk identification model; a second determination module for determining the target parameters of the global model based on the model parameters of the second risk identification model in the K financial institutions, and at least one of data distribution information and data volume; a second processing module for repeatedly executing the step of randomly selecting K financial institutions from multiple financial institutions until the iteration condition is met, and updating the parameters of the local model in the target financial institution according to the target parameters of the global model to obtain a risk identification model.

[0125] Optionally, in the risk object identification device provided in the embodiment of the present application, the risk object identification device also includes: a calculation module, used to calculate the updated gradient value of the model parameters of the first risk identification model in each iteration round according to the preset loss function and the identification result of the first risk identification model; a third processing module, used to add differential privacy noise to the updated gradient value to obtain the target gradient value; and a second update module, used to update the parameters of the first risk identification model based on the target gradient value.

[0126] Optionally, in the risk object identification device provided in the embodiment of the present application, the risk object identification device also includes: a generation module, used to generate a histogram of the historical transaction data of a financial institution in multiple feature dimensions for each financial institution among a plurality of financial institutions; a third determination module, used to determine the probability distribution information of the histogram for each histogram of the financial institution based on the amount of data corresponding to each interval in the histogram, wherein the probability distribution information includes the probability distribution value corresponding to each interval; and a fourth determination module, used to determine the data distribution information of the financial institution based on the probability distribution information of each histogram of the financial institution.

[0127] Optionally, in the risk object identification device provided in the embodiment of the present application, the second determination module also includes: a calculation submodule, which is used to average the probability distribution values ​​of the same interval belonging to the same feature dimension in the data distribution information of K financial institutions to obtain target data distribution information; a first determination submodule, which is used to determine the first weight for each of the K financial institutions based on the relative entropy between the data distribution information of the financial institution and the target data distribution information; a second determination submodule, which is used to determine the second weight based on the data volume of the historical transaction data of the financial institution and the total data volume of the historical transaction data of the K financial institutions; a third determination submodule, which is used to determine the target weight of the financial institution based on at least one of the first weight and the second weight, and determine the target parameters based on the target weights of the K financial institutions and the model parameters of the second risk identification model.

[0128] Optionally, in the risk object identification device provided in the embodiment of the present application, the third determination submodule also includes: a first determination unit, used to perform weighted sum calculation on the first weight and the second weight based on a preset weight factor combination to obtain a target weight, or, a second determination unit, used to determine the first weight as the target weight, or, a third determination unit, used to determine the second weight as the target weight.

[0129] Optionally, in the risk object identification device provided in the embodiment of the present application, the first determination module also includes: a fourth determination submodule, used to determine that the target information represents the existence of a risky transaction for the target object when the probability value in the identification result is greater than a preset probability value; and a fifth determination submodule, used to determine that the target information represents the existence of a risky transaction for the target object when the probability value in the identification result is less than or equal to the preset probability value.

[0130] It should be noted that the acquisition module 401, the first processing module 402, and the first determination module 403 correspond to steps S201 to S203 in Example 1. The examples and application scenarios implemented by the three modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned Example 1. It should be noted that the above-mentioned modules or units can be hardware components or software components stored in a memory (e.g., memory 104) and processed by one or more processors (e.g., processors 102a, 102b, ..., 102n). The above-mentioned modules can also be run as part of the device in the computer terminal 10 provided in Example 1.

[0131] Example 3

[0132] An embodiment of the present application may provide an electronic device, Figure 5 This is a structural block diagram of an electronic device according to an embodiment of the present application. Figure 5As shown, the electronic device may include: one or more ( Figure 5 Only one is shown) processor 1002, memory 1004, storage controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0133] Among them, the memory can be used to store software programs and modules, such as program instructions / modules corresponding to the methods and devices in the embodiments of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, implementing the above-mentioned method. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the terminal via a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network and a combination thereof.

[0134] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: obtain transaction data of the target object in the target financial institution; process the transaction data through the risk identification model in the target financial institution to obtain an identification result, wherein the risk identification model is based on a federated learning framework and is trained by historical transaction data of multiple financial institutions, and in the federated learning process, the model parameters of the risk identification model are determined based on at least one of the following: data distribution information of the historical transaction data of multiple financial institutions, data volume of the historical transaction data of multiple financial institutions, multiple financial institutions include the target financial institution, and the identification result represents the probability value of the target object having risky transactions; determine the target information based on the identification result, wherein the target information is used to represent whether the target object is an object with risky transactions.

[0135] The processor can also call the information and application stored in the memory through the transmission device to perform the following steps: randomly select K financial institutions from multiple financial institutions; for each of the K financial institutions, obtain the parameters of the global model from the target server through the client of the financial institution, and update the local model based on the parameters of the global model to obtain a first risk identification model, wherein the model structure of the global model and the local model are the same; train the first risk identification model based on the historical transaction data in the financial institution to obtain a second risk identification model; determine the target parameters of the global model based on the model parameters of the second risk identification model in the K financial institutions, and at least one of the data distribution information and the data volume; if the preset iteration condition is not met, repeat the step of randomly selecting K financial institutions from multiple financial institutions until the iteration condition is met, and update the parameters of the local model in the target financial institution according to the target parameters of the global model to obtain a risk identification model.

[0136] The processor can also call the information and application stored in the memory through the transmission device to perform the following steps: in each iteration round, calculate the updated gradient value of the model parameters of the first risk identification model based on the preset loss function and the identification result of the first risk identification model; add differential privacy noise to the updated gradient value to obtain the target gradient value; and update the parameters of the first risk identification model based on the target gradient value.

[0137] The processor can also call the information and application stored in the memory through the transmission device to perform the following steps: before determining the target parameters of the global model based on the model parameters of the second risk identification model in K financial institutions, and at least one of the data distribution information and the data volume, for each financial institution among the multiple financial institutions, generate a histogram of the historical transaction data of the financial institution in multiple feature dimensions; for each histogram of the financial institution, determine the probability distribution information of the histogram based on the data volume corresponding to each interval in the histogram, wherein the probability distribution information includes the probability distribution value corresponding to each interval; based on the probability distribution information of each histogram of the financial institution, determine the data distribution information of the financial institution.

[0138] The processor can also call the information and application stored in the memory through the transmission device to perform the following steps: averaging the probability distribution values ​​of the same interval belonging to the same feature dimension in the data distribution information of K financial institutions to obtain target data distribution information; for each of the K financial institutions, determining a first weight based on the relative entropy between the data distribution information of the financial institution and the target data distribution information; determining a second weight based on the data volume of the historical transaction data of the financial institution and the total data volume of the historical transaction data of the K financial institutions; determining the target weight of the financial institution based on at least one of the first weight and the second weight, and determining the target parameters based on the target weights of the K financial institutions and the model parameters of the second risk identification model.

[0139] The processor can also call the information and application stored in the memory through the transmission device to perform the following steps: perform weighted sum calculation on the first weight and the second weight based on a preset weight factor combination to obtain the target weight, or determine the first weight as the target weight, or determine the second weight as the target weight.

[0140] The processor can also call the information and application stored in the memory through the transmission device to perform the following steps: when the probability value in the recognition result is greater than the preset probability value, determine that the target information represents that the target object has a risky transaction; when the probability value in the recognition result is less than or equal to the preset probability value, determine that the target information represents that the target object has no risky transaction.

[0141] In an embodiment of the present application, a risk identification model for a target financial institution is designed based on a federated learning framework and trained with historical transaction data from multiple financial institutions. This allows for the target financial institution's risk identification model to be trained based on the historical transaction data of multiple financial institutions while effectively protecting the data privacy and security of each financial institution. This overcomes the limitations of training a model based on a single dataset, improves the model's training effectiveness, and increases the accuracy of identifying users with transaction risks. Furthermore, by determining model parameters based on at least one of the data distribution information and the data volume of the historical transaction data of multiple financial institutions, effective consideration is achieved for data differences between multiple financial institutions, thereby improving the accuracy of determining model parameters and the accuracy of identifying users with transaction risks.

[0142] It can be seen that the method provided in this application achieves the purpose of obtaining a model for risk identification based on the federated learning framework training, and determining the model parameters in the federated learning process based on the data information of historical data in the institution, achieving the technical effect of improving the recognition accuracy, and solving the technical problem that a small amount of transaction data is used in related technologies to train the risk identification model, resulting in low recognition accuracy when identifying users with risky transactions through the risk identification model.

[0143] It can be understood by those skilled in the art that Figure 5 The structure shown is for illustration only, and the electronic device may also be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a PDA, a mobile Internet device (Mobile Internet Devices, MID), a PAD, or other terminal devices. Figure 5 It does not limit the structure of the above electronic device. For example, the electronic device may also include Figure 5 More or fewer components (such as network interfaces, display devices, etc.) shown in, or with Figure 5 Different configurations shown.

[0144] A person skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0145] Example 4

[0146] The embodiment of the present application further provides a storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the risk object identification method provided in the first embodiment.

[0147] Optionally, in this embodiment, the above-mentioned storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group.

[0148] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing the steps of the method for identifying risk objects.

[0149] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0150] In the above embodiments of the present application, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0151] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0152] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0153] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0154] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0155] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for identifying risk objects, characterized in that: include: Obtain transaction data of the target person in the target financial institution; Processing the transaction data using a risk identification model in the target financial institution to obtain an identification result, wherein the risk identification model is based on a federated learning framework and is trained using historical transaction data of multiple financial institutions, and model parameters of the risk identification model are determined during the federated learning process based on at least one of: data distribution information of the historical transaction data of the multiple financial institutions and data volume of the historical transaction data of the multiple financial institutions, the multiple financial institutions including the target financial institution, and the identification result represents a probability value of the target object having a risky transaction; Target information is determined according to the recognition result, wherein the target information is used to indicate whether the target object is an object of a risky transaction.

2. The method according to claim 1, characterized in that The federated learning framework includes a local model corresponding to the client of each financial institution. The local model is an initial risk identification model, wherein the risk identification model is obtained by: randomly selecting K financial institutions from the plurality of financial institutions; For each of the K financial institutions, obtaining parameters of a global model from a target server through a client of the financial institution, and updating the local model based on the parameters of the global model to obtain a first risk identification model, wherein the global model and the local model have the same model structure; Training the first risk identification model based on historical transaction data in the financial institution to obtain a second risk identification model; determining target parameters of the global model based on model parameters of the second risk identification model in the K financial institutions, and at least one of the data distribution information and the data volume; If the preset iteration condition is not met, the step of randomly selecting K financial institutions from the multiple financial institutions is repeated until the iteration condition is met, and the parameters of the local model in the target financial institution are updated according to the target parameters of the global model to obtain the risk identification model.

3. The method according to claim 2, characterized in that In the process of training the first risk identification model based on the historical transaction data of the financial institution to obtain the second risk identification model, the method further includes: In each iteration round, an updated gradient value of a model parameter of the first risk identification model is calculated according to a preset loss function and an identification result of the first risk identification model; Adding differential privacy noise to the updated gradient value to obtain a target gradient value; Parameters of the first risk identification model are updated based on the target gradient value.

4. The method according to claim 2, characterized in that Before determining the target parameters of the global model based on the model parameters of the second risk identification model in the K financial institutions and at least one of the data distribution information and the data volume, the method further includes: For each of the plurality of financial institutions, generating a histogram of the historical transaction data of the financial institution in a plurality of feature dimensions; For each histogram of the financial institution, determine probability distribution information of the histogram based on the amount of data corresponding to each interval in the histogram, wherein the probability distribution information includes a probability distribution value corresponding to each interval; The data distribution information of the financial institution is determined based on the probability distribution information of each histogram of the financial institution.

5. The method according to claim 4, characterized in that Determining target parameters of the global model based on model parameters of the second risk identification model in the K financial institutions and at least one of the data distribution information and the data volume includes: averaging the probability distribution values ​​of the same interval in the same feature dimension in the data distribution information of the K financial institutions to obtain target data distribution information; For each of the K financial institutions, determining a first weight according to a relative entropy between the data distribution information of the financial institution and the target data distribution information; determining a second weight based on the amount of historical transaction data of the financial institution and the total amount of historical transaction data of the K financial institutions; The target weight of the financial institution is determined based on at least one of the first weight and the second weight, and the target parameter is determined based on the target weights of the K financial institutions and the model parameters of the second risk identification model.

6. The method according to claim 5, characterized in that Determining a target weight of the financial institution according to at least one of the first weight and the second weight includes: Performing a weighted sum calculation on the first weight and the second weight according to a preset weight factor combination to obtain the target weight, or, Determine the first weight as the target weight, or, The second weight is determined as the target weight.

7. The method according to claim 1, characterized in that Determining target information according to the recognition result includes: If the probability value in the recognition result is greater than a preset probability value, determining that the target information indicates that the target object has a risky transaction; When the probability value in the recognition result is less than or equal to the preset probability value, it is determined that the target information indicates that there is no risky transaction for the target object.

8. A device for identifying risky objects, characterized in that: include: An acquisition module, used to acquire transaction data of a target object in a target financial institution; a first processing module, configured to process the transaction data using a risk identification model in the target financial institution to obtain an identification result, wherein the risk identification model is based on a federated learning framework and is trained using historical transaction data of multiple financial institutions, and wherein model parameters of the risk identification model are determined during the federated learning process based on at least one of: data distribution information of the historical transaction data of the multiple financial institutions and data volume of the historical transaction data of the multiple financial institutions, wherein the multiple financial institutions include the target financial institution, and the identification result represents a probability value of the target object having a risky transaction; The first determination module is configured to determine target information according to the recognition result, wherein the target information is used to indicate whether the target object is an object of a risky transaction.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored executable program, wherein when the executable program is run, the device where the computer-readable storage medium is located is controlled to execute the risk object identification method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: include: a memory storing an executable program; A processor is configured to run the program, wherein the program, when running, executes the risk object identification method according to any one of claims 1 to 7.

11. A computer program product comprising computer instructions, characterized in that When the computer instructions are executed by a processor, the steps of the risk object identification method according to any one of claims 1 to 7 are implemented.

Images