Transaction risk detection model training method, transaction risk detection method and device

Abstract

Embodiments of the present specification provide a transaction risk detection model training method, a transaction risk detection method and device, wherein the training method comprises: obtaining a transaction sample set to be trained; the transaction sample set comprises personal transaction samples, non-target enterprise transaction samples and target enterprise transaction samples, a first quantity of the personal transaction samples is greater than a second quantity of the non-target enterprise transaction samples, and the second quantity is greater than a third quantity of the target enterprise transaction samples; each transaction sample comprises common transaction features for knowledge transfer; the personal transaction samples, the non-target enterprise transaction samples and the target enterprise transaction samples are sequentially used to iteratively train a neural network based on a knowledge transfer training mode to obtain a transaction risk detection model; and the transaction risk detection model is used for risk detection on transaction data of the target enterprise.

Description

Technical Field

[0001] This document relates to the field of computer technology, and in particular to a training method, a method and apparatus for detecting trading risks. Background Technology

[0002] As businesses expand, they often engage in new business models with new partners. To ensure transaction security, companies typically develop risk detection strategies and process transactions based on these strategies. However, existing risk detection strategies often fail to adequately cover new business models with new partners, resulting in poor accuracy. Furthermore, due to the new business model and partners, the amount of transaction data is limited, insufficient for independent model training. Therefore, how to perform transaction risk detection for new business models with limited transaction data is a pressing technical problem that needs to be solved. Summary of the Invention

[0003] This specification provides one or more embodiments of a method for training a transaction risk detection model. The method includes acquiring a transaction sample set to be trained. The transaction sample set includes individual transaction samples, non-target enterprise transaction samples, and target enterprise transaction samples. A first number of individual transaction samples is greater than a second number of non-target enterprise transaction samples. The second number is greater than a third number of target enterprise transaction samples. Each transaction sample includes common transaction features for knowledge transfer. The neural network to be trained is iteratively trained sequentially using the individual transaction samples, the non-target enterprise transaction samples, and the target enterprise transaction samples, based on a knowledge transfer training method, to obtain the transaction risk detection model. The transaction risk detection model is used to perform risk detection on the transaction data of the target enterprise.

[0004] This specification provides one or more embodiments of a transaction risk detection method. The method includes acquiring transaction data of a target transaction of a target enterprise; extracting transaction features from the transaction data; and performing risk detection processing on the transaction features using a pre-trained transaction risk detection model to obtain a risk detection result for the target transaction. The transaction risk detection model is trained using the training method described above.

[0005] This specification provides one or more embodiments of a training apparatus for a transaction risk detection model. The apparatus includes an acquisition module for acquiring a transaction sample set to be trained. The transaction sample set includes individual transaction samples, non-target enterprise transaction samples, and target enterprise transaction samples. The first number of individual transaction samples is greater than the second number of non-target enterprise transaction samples. The second number is greater than the third number of target enterprise transaction samples. Each transaction sample includes common transaction features for knowledge transfer. The apparatus also includes a training module that iteratively trains a neural network to be trained using the individual transaction samples, the non-target enterprise transaction samples, and the target enterprise transaction samples, based on a knowledge transfer training method, to obtain the transaction risk detection model. The transaction risk detection model is used to perform risk detection on the transaction data of the target enterprise.

[0006] This specification provides one or more embodiments of a transaction risk detection device. The device includes an acquisition module for acquiring transaction data of a target transaction of a target enterprise. The device also includes an extraction module for extracting transaction features from the transaction data. Furthermore, the device includes a detection module for performing risk detection processing on the transaction features using a pre-trained transaction risk detection model to obtain a risk detection result for the target transaction. The transaction risk detection model is trained using the aforementioned training method.

[0007] This specification provides one or more embodiments of a training device for a transaction risk detection model. The device includes a processor. The device also includes a memory arranged to store computer-executable instructions. When executed, the computer-executable instructions cause the processor to acquire a set of transaction samples to be trained. The transaction sample set includes individual transaction samples, non-target enterprise transaction samples, and target enterprise transaction samples. A first number of individual transaction samples is greater than a second number of non-target enterprise transaction samples. The second number is greater than a third number of target enterprise transaction samples. Each transaction sample includes common transaction features for knowledge transfer. Using the individual transaction samples, the non-target enterprise transaction samples, and the target enterprise transaction samples sequentially, an iterative training process is performed on the neural network to be trained based on a knowledge transfer training method to obtain a transaction risk detection model. The transaction risk detection model is used to perform risk detection on the transaction data of the target enterprise.

[0008] This specification provides one or more embodiments of a training device for a transaction risk detection model. The device includes a processor. The device also includes a memory arranged to store computer-executable instructions. When executed, the computer-executable instructions cause the processor to acquire transaction data of a target transaction of a target enterprise. Transaction features of the transaction data are extracted. A pre-trained transaction risk detection model is used to perform risk detection processing on the transaction features to obtain a risk detection result for the target transaction. The transaction risk detection model is trained using the training method described above.

[0009] This specification provides one or more embodiments of a storage medium for storing computer-executable instructions. When executed by a processor, the computer-executable instructions acquire a set of transaction samples to be trained. The transaction sample set includes individual transaction samples, non-target enterprise transaction samples, and target enterprise transaction samples. A first number of individual transaction samples is greater than a second number of non-target enterprise transaction samples. The second number is greater than a third number of target enterprise transaction samples. Each transaction sample includes common transaction features for knowledge transfer. Using the individual transaction samples, the non-target enterprise transaction samples, and the target enterprise transaction samples sequentially, an iterative training process is performed on the neural network to be trained based on a knowledge transfer training method to obtain a transaction risk detection model. The transaction risk detection model is used to perform risk detection on the transaction data of the target enterprise.

[0010] This specification provides one or more embodiments of a storage medium for storing computer-executable instructions. When executed by a processor, the computer-executable instructions acquire transaction data of a target transaction of a target enterprise. Transaction features of the transaction data are extracted. A pre-trained transaction risk detection model is used to perform risk detection processing on the transaction features to obtain a risk detection result for the target transaction. The transaction risk detection model is trained using the training method described above. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in one or more embodiments of this specification or in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a schematic diagram of the first process for training a transaction risk detection model provided in the embodiments of this specification;

[0013] Figure 2A second flowchart illustrating a training method for a transaction risk detection model provided in the embodiments of this specification;

[0014] Figure 3 A schematic diagram illustrating the training process of a transaction risk detection model provided in the embodiments of this specification;

[0015] Figure 4 This is a schematic diagram of a neural network structure provided in an embodiment of this specification;

[0016] Figure 5 A flowchart illustrating a transaction risk detection method provided in the embodiments of this specification;

[0017] Figure 6 A schematic diagram of the module composition of a training device for a transaction risk detection model provided in an embodiment of this specification;

[0018] Figure 7 This is a schematic diagram of the module composition of a transaction risk detection device provided in the embodiments of this specification;

[0019] Figure 8 A schematic diagram of the structure of a training device for a transaction risk detection model provided in an embodiment of this specification;

[0020] Figure 9 This is a schematic diagram of the structure of a transaction risk detection device provided in the embodiments of this specification. Detailed Implementation

[0021] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this document.

[0022] Considering the limited transaction data available for new business models involving new business partners, training a model with only this small amount of data would result in poor robustness and low accuracy in transaction risk detection. Currently, there is a large volume of personal transaction data and a medium volume of transaction data from existing business partners. While the transaction patterns of personal and business transactions differ, and even among different businesses, similarities remain. For example, month-on-month transaction comparisons reflect transaction clustering over a given period, and transaction IPs indicate the location of the transaction. Therefore, this application provides a training method for a transaction risk detection model. This method sequentially utilizes a large volume of personal transaction samples, a medium volume of transaction samples from non-target businesses, and a small volume of transaction samples from target businesses. Iterative training of the neural network to be trained is based on knowledge transfer to obtain a transaction risk detection model for detecting risks in the transaction data of target businesses. This ensures the accuracy of the transaction risk detection model, enables risk detection of the target company's transaction data, detects transaction risks of new business partners and new business models, and allows for model training with a small number of samples.

[0023] Specifically, Figure 1 This is a flowchart illustrating a training method for a transaction risk detection model provided in one or more embodiments of this specification. Figure 1 The method described can be executed by a training device for the transaction risk detection model. This training device can be located on a terminal device or on a server. The terminal device can be a mobile phone, tablet, desktop computer, laptop, etc.; the server can be a standalone server or a server cluster consisting of multiple servers. Figure 1 As shown, the method includes the following steps:

[0024] Step S102: Obtain the transaction sample set to be trained; the transaction sample set includes individual transaction samples, non-target enterprise transaction samples and target enterprise transaction samples, the first number of individual transaction samples is greater than the second number of non-target enterprise transaction samples, and the second number is greater than the third number of target enterprise transaction samples; each transaction sample includes common transaction features for knowledge transfer.

[0025] Optionally, the system can obtain a set of transaction samples to be trained from a specified sample pool; or obtain a set of transaction samples to be trained input by the user; or receive a set of transaction samples to be trained sent by an upper-layer device.

[0026] Each transaction sample in the transaction sample set includes common transaction features for knowledge transfer, and may also include differential transaction features and tags. Common transaction features are those present in individual transaction samples, non-target company transaction samples, and target company transaction samples, and have the same or similar meanings across these three groups, making them suitable for knowledge transfer. Common transaction features may include transaction month-on-month growth, transaction IP, transaction time, and the largest transaction party's share. Differential transaction features are those not present in individual, non-target company, and target company transaction samples, and cannot be used for knowledge transfer. Differential transaction features may include characteristics such as age and gender in individual transaction samples, and characteristics such as registration time, registered capital, and beneficiaries in non-target company and target company transaction samples. Tags are used to characterize whether a specific risk exists in the transaction sample to which it belongs. It is understood that target company transaction samples are a small number of samples, such as the transaction samples of new business models undertaken by new business partners mentioned above.

[0027] Furthermore, the transaction samples included in the transaction sample set can exist in tabular form, for example, each row of the table represents one sample. The transaction samples included in the transaction sample set can also exist in the form of graph data, which includes multiple vertices and multiple edges. Each vertex represents a trading party (i.e., an individual or a company), and the edges between vertices represent the transactions between the trading parties. This specification does not specify a particular form for the existence of the transaction samples; it can be set as needed in practical applications.

[0028] Step S104: Using individual transaction samples, non-target company transaction samples, and target company transaction samples in sequence, the neural network to be trained is iteratively trained using a knowledge transfer training method to obtain a transaction risk detection model; the transaction risk detection model is used to detect risks in the transaction data of the target company.

[0029] Since the first number of individual transaction samples is greater than the second number of non-target enterprise transaction samples, and this second number is greater than the third number of target enterprise transaction samples, in order to ensure the stability and accuracy of the trained transaction risk detection model, in this embodiment of the specification, a preliminary model is first obtained by iteratively training the neural network to be trained using a large number of individual transaction samples. Then, the preliminary model is fine-tuned based on a medium number of non-target enterprise transaction samples to obtain a model that is closer to the target model. Finally, the model is fine-tuned again using a small number of target enterprise transaction samples to obtain the target model, which is the transaction risk detection model.

[0030] In one or more embodiments of this specification, a large number of individual transaction samples, a medium number of non-target enterprise transaction samples, and a small number of target enterprise transaction samples are sequentially used to iteratively train the neural network to be trained based on a knowledge transfer training method, thereby obtaining a transaction risk detection model for detecting risks in the transaction data of target enterprises. This ensures the accuracy of the transaction risk detection model, achieves risk detection of the transaction data of target enterprises (i.e., detects transaction risks of new enterprise partners and new business models), and also enables model training with a small number of samples.

[0031] The knowledge transfer in the embodiments of this specification includes the transfer of model parameters, specifically, such as... Figure 2 As shown, step S104 may include steps S104-2 to S104-6:

[0032] Step S104-2: Iteratively train the first neural network to be trained using personal transaction samples to obtain the first model; the first model includes first model parameters for knowledge transfer.

[0033] Specifically, the first neural network includes a first branch and a second branch. The initial parameters of the first branch and the second branch of the first neural network are randomly initialized. The initialized first branch is iteratively trained based on the common transaction features of individual transaction samples, and the initialized second branch is iteratively trained based on the differential transaction features of individual transaction samples to obtain a first model. The first branch of the first model uses the parameters of the first model.

[0034] More specifically, after randomly initializing the initial parameters of the first and second branches of the first neural network, individual transaction samples to be trained are iteratively obtained from the transaction sample set. The common transaction features of these individual transaction samples are input into the first branch, and the differential transaction features are input into the second branch for training. When the training iterations reach a preset number or the preset loss function converges, training stops, and the first model is obtained. The preset loss function can be the cross-entropy loss function.

[0035] Step S104-4: Use transaction samples from non-target enterprises and the parameters of the first model to iteratively train the second neural network to be trained to obtain the second model; the second model includes the second model parameters for knowledge transfer.

[0036] Specifically, the initial parameters of the first branch of the second neural network are initialized to the parameters of the first model, and the initial parameters of the second branch of the second neural network are randomly initialized; the initialized first branch is iteratively trained based on the common transaction features of the transaction samples of non-target enterprises, and the initialized second branch is iteratively trained based on the differential transaction features of the transaction samples of non-target enterprises to obtain the second model; the first branch of the second model uses the parameters of the second model.

[0037] More specifically, after initializing the initial parameters of the first and second branches of the second neural network, iteratively acquiring transaction samples from non-target companies to be trained from the transaction sample set, inputting the common transaction features of these non-target company transaction samples into the first branch of the second neural network, and inputting the differential transaction features of these non-target company transaction samples into the second branch of the second neural network for training; when the training reaches a preset number of iterations or the preset loss function converges, training stops, and the second model is obtained. Knowledge transfer is achieved by initializing the initial parameters of the first branch of the second neural network to the parameters of the first model.

[0038] Step S104-6: Use the transaction samples of the target company and the parameters of the second model to iteratively train the third neural network to be trained, and obtain the transaction risk detection model.

[0039] Specifically, the initial parameters of the first branch of the third neural network are initialized to the parameters of the second model, and the initial parameters of the second branch of the third neural network are randomly initialized. The first branch after initialization is iteratively trained based on the common transaction features of the transaction samples of the target enterprise, and the second branch after initialization is iteratively trained based on the differential transaction features of the transaction samples of the target enterprise, so as to obtain the transaction risk detection model.

[0040] More specifically, after initializing the initial parameters of the first and second branches of the third neural network, transaction samples of the target company to be trained are iteratively obtained from the transaction sample set. The common transaction features of the target company's transaction samples are input into the first branch of the third neural network, and the differential transaction features of the target company's transaction samples are input into the second branch of the third neural network for training. When the training reaches a preset number of iterations or the preset loss function converges, training stops, and the transaction risk detection model is obtained. Knowledge transfer is achieved by initializing the initial parameters of the first branch of the third neural network to the parameters of the second model.

[0041] Furthermore, the individual transaction sample is denoted as X1, the non-target enterprise transaction sample as X2, and the target enterprise transaction sample as X3. The common transaction characteristics included in each transaction sample are denoted as X. c Let X be the differential transaction characteristics included in the individual transaction sample.d1 The differential transaction characteristics included in the transaction samples of non-target enterprises are denoted as X. d2 Let X be the differential transaction characteristics included in the transaction sample of the target company. d3 Let the output of the first neural network be denoted as Y1 = σ(X1), the output of the second neural network as Y2 = σ(X2), and the output of the third neural network as Y3 = σ(X3). A schematic diagram of the above knowledge transfer-based training process is shown below. Figure 3 As shown, by performing two knowledge transfer operations, the trained trading risk detection model exhibits high stability and accuracy.

[0042] Furthermore, to facilitate understanding of the data processing methods used in the training process described in this manual, the model structure is explained below. Specifically, as follows... Figure 4 As shown, the neural network to be trained includes: a first branch, a second branch, a merging layer, and a first fully connected layer; correspondingly, step S104 involves iterative training of the neural network to be trained, including:

[0043] The first sample features are obtained by performing a first linear transformation on the shared transaction features through the first branch.

[0044] The second branch is used to perform a second linear transformation on the differential transaction features to obtain the second sample features.

[0045] The first sample features and the second sample features are merged by a merging layer to obtain the third sample features;

[0046] The first fully connected layer maps the features of the third sample to obtain the output result.

[0047] The first linear transformation process and the second linear transformation process can be the same or different, and can be set according to the needs of practical application. Merging the first and second sample features through the merging layer can include sequentially concatenating the first and second sample features to obtain the third sample feature. The first fully connected layer can include a sigmoid activation function, which is used to map the third sample feature to obtain an output representing whether the corresponding transaction sample has a specified risk. The mapping process of the sigmoid activation function can be referred to existing technology, and will not be detailed in this application.

[0048] Furthermore, such as Figure 4 As shown, the first branch includes a first input layer and M second fully connected layers, and the second branch includes a second input layer and N third fully connected layers; where M and N are integers greater than 1.

[0049] Accordingly, the first sample features obtained by performing a first linear transformation on the shared transaction features through the first branch can include:

[0050] The common transaction features are input into the first input layer, and then processed by M second fully connected layers to perform a first linear transformation on the first input features to obtain the first sample features. The first input features of the first second fully connected layer include the common features, and the first input features of the m-th second fully connected layer include the output features of the (m-1)-th second fully connected layer, where 2 ≤ m ≤ M. For example, the first input features of the second second fully connected layer include the output features of the first second fully connected layer.

[0051] Furthermore, by performing a second linear transformation on the differential transaction features through the second branch, the second sample features are obtained, which may include:

[0052] The differential transaction features are input into the second input layer, and then processed by N third fully connected layers using a second linear transformation to obtain the second sample features. The second input features of the first third fully connected layer include the differential transaction features, and the input features of the nth third fully connected layer include the output features of the (n-1)th third fully connected layer, where 2 ≤ n ≤ N. For example, the input features of the third third fully connected layer include the output features of the second third fully connected layer.

[0053] It is understandable that the aforementioned first, second, and third neural networks all have... Figure 4 The network structure is shown. It should be noted that N in the first, second, and third neural networks can be the same or different; that is, the number of layers in the third fully connected layer of the second branch of the first, second, and third neural networks can be the same or different.

[0054] It should be noted that the model training method provided in this specification can be applied not only to the training scenarios of the aforementioned risk detection model, but also to other model training scenarios with a small number of samples. For example, based on the aforementioned knowledge transfer concept, a model can be trained using a large amount of image data in a common format and a small amount of image data in a specific format to obtain an image processing model for processing image data in that specific format.

[0055] In one or more embodiments of this specification, a large number of individual transaction samples, a medium number of non-target enterprise transaction samples, and a small number of target enterprise transaction samples are sequentially used to iteratively train the neural network to be trained based on a knowledge transfer training method, thereby obtaining a transaction risk detection model for detecting risks in the transaction data of target enterprises. This ensures the accuracy of the transaction risk detection model, achieves risk detection of the transaction data of target enterprises (i.e., detects transaction risks of new enterprise partners and new business models), and enables model training with a small number of samples.

[0056] Corresponding to the training method of the aforementioned transaction risk detection model, and based on the same technical concept, one or more embodiments of this specification also provide a transaction risk detection method. Figure 5 This is a flowchart illustrating a transaction risk detection method provided in one or more embodiments of this specification. Figure 5 The method described can be executed by a transaction risk detection device; this device can be installed in a terminal device or on a server; the terminal device can be a mobile phone, tablet, desktop computer, laptop, etc.; the server can be a standalone server or a server cluster consisting of multiple servers. Figure 5 As shown, the method includes the following steps:

[0057] Step S202: Obtain the transaction data of the target company's target transaction;

[0058] Optionally, the system can acquire transaction data of the target company's target transaction input by the user; or, it can receive transaction data of the target company's target transaction sent by the transaction device. This transaction data includes transaction time, transaction IP, and information of both parties to the transaction.

[0059] Step S204: Extract transaction features from the transaction data;

[0060] Specifically, information such as transaction time and transaction IP address can be extracted from transaction data as transaction characteristics of the target transaction. Information such as the target company's registration time, registered capital, and beneficiaries can also be obtained from a specified database and combined with the information extracted from the transaction data as transaction characteristics of the target transaction.

[0061] Step S206: Use a pre-trained transaction risk detection model to perform risk detection processing on the transaction features to obtain the risk detection result of the target transaction.

[0062] Specifically, transaction features are input into a transaction risk detection model for risk detection processing to obtain the risk detection result of the target transaction. The transaction risk detection model is trained using the training method provided in any of the aforementioned embodiments.

[0063] In one or more embodiments of this specification, when performing risk detection processing on target transactions of target enterprises, the transaction risk detection model is obtained by iteratively training the neural network to be trained using a knowledge transfer training method, sequentially utilizing a large number of individual transaction samples, a medium number of non-target enterprise transaction samples, and a small number of target enterprise transaction samples. This training method ensures the accuracy of the transaction risk detection model, thus achieving effective risk detection on the transaction data of target enterprises, that is, achieving transaction risk detection for new enterprise partners and new business models.

[0064] Corresponding to the training method of the transaction risk detection model described above, based on the same technical concept, one or more embodiments of this specification also provide a training device for the transaction risk detection model. Figure 6 This diagram illustrates the module composition of a training apparatus for a transaction risk detection model provided in one or more embodiments of this specification. The apparatus is used to execute the aforementioned training method for the transaction risk detection model, such as... Figure 6 As shown, the device includes:

[0065] The acquisition module 301 acquires a transaction sample set to be trained; the transaction sample set includes individual transaction samples, non-target enterprise transaction samples, and target enterprise transaction samples, wherein the first number of individual transaction samples is greater than the second number of non-target enterprise transaction samples, and the second number is greater than the third number of target enterprise transaction samples; each transaction sample includes common transaction features for knowledge transfer.

[0066] The training module 302 sequentially uses the individual transaction samples, the non-target enterprise transaction samples, and the target enterprise transaction samples to iteratively train the neural network to be trained based on a knowledge transfer training method, thereby obtaining a transaction risk detection model; the transaction risk detection model is used to detect risks in the transaction data of the target enterprise.

[0067] Optionally, the training module 302 uses the individual transaction samples to iteratively train the first neural network to be trained, obtaining a first model; the first model includes first model parameters for knowledge transfer; and,

[0068] The second neural network to be trained is iteratively trained using the transaction samples of the non-target enterprise and the parameters of the first model to obtain the second model; the second model includes second model parameters for knowledge transfer.

[0069] The transaction risk detection model is obtained by iteratively training the third neural network to be trained using the transaction samples of the target company and the parameters of the second model.

[0070] The training apparatus for the transaction risk detection model provided in one or more embodiments of this specification sequentially utilizes a large number of individual transaction samples, a medium number of transaction samples from non-target enterprises, and a small number of transaction samples from target enterprises. Based on a knowledge transfer training method, iteratively trains the neural network to be trained, resulting in a transaction risk detection model for detecting risks in the transaction data of target enterprises. Thus, it ensures the accuracy of the transaction risk detection model, achieves risk detection of the transaction data of target enterprises (including new business partners and new business models), and enables model training with a small number of samples.

[0071] It should be noted that the embodiments of the training device for the trading risk detection model in this specification and the embodiments of the training method for the trading risk detection model in this specification are based on the same inventive concept. Therefore, the specific implementation of this embodiment can refer to the implementation of the corresponding training method for the trading risk detection model mentioned above, and the repeated parts will not be described again.

[0072] Furthermore, corresponding to the above-mentioned transaction risk detection method, based on the same technical concept, one or more embodiments of this specification also provide a transaction risk detection device. Figure 7 This is a schematic diagram of the module composition of a transaction risk detection device provided in one or more embodiments of this specification. The device is used to execute the aforementioned transaction risk detection method, such as... Figure 7 As shown, the device includes:

[0073] Module 401 retrieves transaction data of the target enterprise's target transaction;

[0074] Extraction module 402 extracts the transaction features of the transaction data;

[0075] The detection module 403 uses a pre-trained transaction risk detection model to perform risk detection processing on the transaction features to obtain the risk detection result of the target transaction; wherein, the transaction risk detection model is trained using the aforementioned training method for transaction risk detection models.

[0076] The transaction risk detection device provided in one or more embodiments of this specification, when performing risk detection processing on target transactions of target enterprises, uses a transaction risk detection model obtained by iteratively training the neural network to be trained based on a knowledge transfer training method, sequentially utilizing a large number of individual transaction samples, a medium number of non-target enterprise transaction samples, and a small number of target enterprise transaction samples. This training method ensures the accuracy of the transaction risk detection model, thus achieving effective risk detection on the transaction data of target enterprises, that is, realizing transaction risk detection for new enterprise partners and new business models.

[0077] It should be noted that the embodiments of the transaction risk detection device in this specification and the embodiments of the transaction risk detection method in this specification are based on the same inventive concept. Therefore, the specific implementation of this embodiment can refer to the implementation of the corresponding transaction risk detection method mentioned above, and the repeated parts will not be described again.

[0078] Furthermore, corresponding to the training method of the trading risk detection model described above, based on the same technical concept, one or more embodiments of this specification also provide a training device for a trading risk detection model, which is used to execute the above-described training method for the trading risk detection model. Figure 8 This is a schematic diagram of the structure of a training device for a transaction risk detection model provided in one or more embodiments of this specification.

[0079] like Figure 8 As shown, the training device for the trading risk detection model can vary significantly due to differences in configuration or performance. It may include one or more processors 501 and memory 502, with memory 502 storing one or more application programs or data. Memory 502 can be temporary or persistent storage. The application programs stored in memory 502 may include one or more modules (not shown), each module including a series of computer-executable instructions from the trading risk detection model training device. Furthermore, processor 501 may be configured to communicate with memory 502, executing the series of computer-executable instructions in memory 502 on the trading risk detection model training device. The trading risk detection model training device may also include one or more power supplies 503, one or more wired or wireless network interfaces 504, one or more input / output interfaces 505, one or more keyboards 506, etc.

[0080] In one specific embodiment, the training device for the trading risk detection model includes a memory and one or more programs, wherein the one or more programs are stored in the memory, and the one or more programs may include one or more modules, and each module may include a series of computer-executable instructions for the training device for the trading risk detection model, and is configured to be executed by one or more processors. The one or more programs include computer-executable instructions for performing the following:

[0081] Obtain a transaction sample set to be trained; the transaction sample set includes individual transaction samples, non-target enterprise transaction samples, and target enterprise transaction samples, wherein the first number of individual transaction samples is greater than the second number of non-target enterprise transaction samples, and the second number is greater than the third number of target enterprise transaction samples; each transaction sample includes common transaction features for knowledge transfer.

[0082] The transaction risk detection model is obtained by iteratively training the neural network to be trained using the individual transaction samples, the non-target enterprise transaction samples, and the target enterprise transaction samples in sequence, based on a knowledge transfer training method. The transaction risk detection model is used to detect risks in the transaction data of the target enterprise.

[0083] The training device for the transaction risk detection model provided in one or more embodiments of this specification sequentially utilizes a large number of individual transaction samples, a medium number of transaction samples from non-target enterprises, and a small number of transaction samples from target enterprises. Based on a knowledge transfer training method, iterative training is performed on the neural network to be trained, resulting in a transaction risk detection model used for risk detection of transaction data from target enterprises. This ensures the accuracy of the transaction risk detection model, achieves risk detection of transaction data from target enterprises (including new business partners and new business models), and enables model training with a small number of samples.

[0084] Furthermore, corresponding to the transaction risk detection method described above, based on the same technical concept, one or more embodiments of this specification also provide a transaction risk detection device, which is used to perform the above-described transaction risk detection method. Figure 9 This is a schematic diagram of the structure of a transaction risk detection device provided for one or more embodiments of this specification.

[0085] like Figure 9 As shown, transaction risk detection devices can vary significantly due to differences in configuration or performance. They may include one or more processors 601 and memory 602, with memory 602 storing one or more application programs or data. Memory 602 can be temporary or persistent storage. The application programs stored in memory 602 may include one or more modules (not shown), each module including a series of computer-executable instructions from the transaction risk detection device. Furthermore, processor 601 may be configured to communicate with memory 602, executing the series of computer-executable instructions in memory 602 on the transaction risk detection device. The transaction risk detection device may also include one or more power supplies 603, one or more wired or wireless network interfaces 604, one or more input / output interfaces 605, one or more keyboards 606, etc.

[0086] In one specific embodiment, the transaction risk detection device includes a memory and one or more programs, wherein the one or more programs are stored in the memory, and the one or more programs may include one or more modules, and each module may include a series of computer-executable instructions for the transaction risk detection device, and is configured to be executed by one or more processors. The one or more programs include computer-executable instructions for performing the following:

[0087] Obtain transaction data for the target company's target transaction;

[0088] Extract the transaction features from the transaction data;

[0089] The transaction features are processed by a pre-trained transaction risk detection model to obtain the risk detection result of the target transaction; wherein the transaction risk detection model is trained using the training method described in any one of claims 1-7.

[0090] The transaction risk detection device provided in one or more embodiments of this specification, when performing risk detection processing on target transactions of target enterprises, uses a transaction risk detection model obtained by iteratively training the neural network to be trained based on a knowledge transfer training method, sequentially utilizing a large number of individual transaction samples, a medium number of non-target enterprise transaction samples, and a small number of target enterprise transaction samples. This training method ensures the accuracy of the transaction risk detection model, thus achieving effective risk detection on the transaction data of target enterprises, that is, realizing the detection of transaction risks for new enterprise partners and new business models.

[0091] It should be noted that the embodiments of the training device for the trading risk detection model in this specification and the embodiments of the training method for the trading risk detection model in this specification are based on the same inventive concept. Therefore, the specific implementation of this embodiment can be referred to the implementation of the corresponding training method for the trading risk detection model mentioned above, and the repeated parts will not be described again.

[0092] Furthermore, corresponding to the above method and based on the same technical concept, one or more embodiments of this specification also provide a storage medium for storing computer-executable instructions. In a specific embodiment, the storage medium can be a USB flash drive, optical disk, hard disk, etc. When the computer-executable instructions stored in the storage medium are executed by a processor, they can realize the following process:

[0093] Obtain a transaction sample set to be trained; the transaction sample set includes individual transaction samples, non-target enterprise transaction samples, and target enterprise transaction samples, wherein the first number of individual transaction samples is greater than the second number of non-target enterprise transaction samples, and the second number is greater than the third number of target enterprise transaction samples; each transaction sample includes common transaction features for knowledge transfer.

[0094] The transaction risk detection model is obtained by iteratively training the neural network to be trained using the individual transaction samples, the non-target enterprise transaction samples, and the target enterprise transaction samples in sequence, based on a knowledge transfer training method. The transaction risk detection model is used to detect risks in the transaction data of the target enterprise.

[0095] When the computer-executable instructions stored in the storage medium provided in one or more embodiments of this specification are executed by a processor, they sequentially utilize a large number of individual transaction samples, a medium number of transaction samples from non-target enterprises, and a small number of transaction samples from target enterprises. Based on a knowledge transfer training method, the neural network to be trained is iteratively trained to obtain a transaction risk detection model for detecting risks in the transaction data of target enterprises. Thus, the accuracy of the transaction risk detection model is ensured, and risk detection of the transaction data of target enterprises is achieved, including transaction risk detection for new business partners and new business models, while also enabling model training with a small number of samples.

[0096] In another specific embodiment, the storage medium can be a USB flash drive, optical disc, hard disk, etc., and the computer-executable instructions stored in the storage medium, when executed by a processor, can achieve the following process:

[0097] Obtain transaction data for the target company's target transaction;

[0098] Extract the transaction features from the transaction data;

[0099] The transaction features are processed for risk detection using a pre-trained transaction risk detection model to obtain the risk detection result of the target transaction; wherein the transaction risk detection model is trained using the aforementioned training method for transaction risk detection models.

[0100] When the computer-executable instructions stored in the storage medium provided in one or more embodiments of this specification are executed by a processor, in the process of performing risk detection processing on the target transaction of the target enterprise, the transaction risk detection model is obtained by iteratively training the neural network to be trained using a knowledge transfer training method, which sequentially utilizes a large number of individual transaction samples, a medium number of non-target enterprise transaction samples, and a small number of target enterprise transaction samples. This training method ensures the accuracy of the transaction risk detection model, thus realizing effective risk detection of the target enterprise's transaction data, that is, realizing transaction risk detection for new enterprise partners and new business models.

[0101] It should be noted that the embodiments concerning the storage medium in this specification are based on the same inventive concept as the embodiments concerning the training method of the transaction risk detection model and the transaction risk detection method in this specification. Therefore, the specific implementation of this embodiment can refer to the implementation of the corresponding training method of the transaction risk detection model and the transaction risk detection method mentioned above, and the repeated parts will not be described again.

[0102] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0103] In the 1990s, improvements to a technology could be clearly distinguished as either hardware improvements (e.g., improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the methodology). However, with technological advancements, many methodological improvements today can be considered direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved methodology into the hardware circuit. Therefore, it cannot be said that a methodological improvement cannot be implemented using a hardware physical module. For example, a Programmable Logic Device (PLD) (e.g., a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logic function is determined by the user programming the device. Designers can program a digital system themselves to "integrate" it onto a PLD, without needing chip manufacturers to design and manufacture dedicated integrated circuit chips. Furthermore, nowadays, instead of manually manufacturing integrated circuit chips, this programming is mostly implemented using "logic compiler" software. Similar to the software compiler used in program development, the original code before compilation must be written in a specific programming language, called a Hardware Description Language (HDL). There are many HDLs, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, and RHDL (Ruby Hardware Description Language). Currently, the most commonly used are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should understand that by simply performing some logic programming on the method flow using one of these hardware description languages ​​and programming it into an integrated circuit, the hardware circuit implementing the logical method flow can be easily obtained.

[0104] The controller can be implemented in any suitable manner. For example, it can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code form, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, ASICs, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0105] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0106] For ease of description, the above apparatus is described by dividing it into various functional units. Of course, when implementing the embodiments of this specification, the functions of each unit can be implemented in one or more software and / or hardware.

[0107] Those skilled in the art will understand that one or more embodiments of this specification can be provided as a method, system, or computer program product. Therefore, one or more embodiments of this specification may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this specification may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0108] This specification is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0109] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0110] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0111] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0112] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0113] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0114] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0115] One or more embodiments of this specification can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a particular task or implement a particular abstract data type. One or more embodiments of this specification can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0116] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0117] The above description is merely an embodiment of this document and is not intended to limit the scope of this document. Various modifications and variations can be made to this document by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this document should be included within the scope of the claims of this document.

Claims

1. A training method for a trading risk detection model, comprising: Obtain the set of transaction samples to be trained; The transaction sample set includes individual transaction samples, non-target enterprise transaction samples, and target enterprise transaction samples. The first number of individual transaction samples is greater than the second number of non-target enterprise transaction samples, and the second number is greater than the third number of target enterprise transaction samples. Each transaction sample includes common transaction features for knowledge transfer. The neural network to be trained is iteratively trained using the individual transaction samples to obtain a first model; the first model includes first model parameters for knowledge transfer. The second neural network to be trained is iteratively trained using the transaction samples of the non-target enterprise and the parameters of the first model to obtain the second model; the second model includes second model parameters for knowledge transfer. The transaction risk detection model is obtained by iteratively training the third neural network to be trained using the transaction samples of the target company and the parameters of the second model; the transaction risk detection model is used to detect risks in the transaction data of the target company.

2. The method according to claim 1, wherein the transaction sample further includes differential transaction features, and the step of iteratively training the first neural network to be trained using the individual transaction sample to obtain the first model includes: Randomly initialize the initial parameters of the first branch of the first neural network and the initial parameters of the second branch of the first neural network; The first branch after initialization is iteratively trained based on the common transaction features of the individual transaction samples, and the second branch after initialization is iteratively trained based on the differential transaction features of the individual transaction samples to obtain the first model. The first branch of the first model uses the parameters of the first model.

3. The method according to claim 1, wherein the transaction sample further includes differential transaction features, and the step of iteratively training the second neural network to be trained using the transaction samples of the non-target enterprise and the parameters of the first model to obtain the second model includes: The initial parameters of the first branch of the second neural network are initialized to the parameters of the first model, and the initial parameters of the second branch of the second neural network are randomly initialized. The first branch after initialization is iteratively trained based on the common transaction features of the transaction samples of the non-target enterprises, and the second branch after initialization is iteratively trained based on the differential transaction features of the transaction samples of the non-target enterprises to obtain the second model. The first branch of the second model uses the parameters of the second model.

4. The method according to claim 1, wherein the transaction sample further includes differential transaction features, and the step of iteratively training the third neural network to be trained using the transaction sample of the target enterprise and the second model parameters to obtain the transaction risk detection model includes: The initial parameters of the first branch of the third neural network are initialized to the parameters of the second model, and the initial parameters of the second branch of the third neural network are randomly initialized. The first branch after initialization is iteratively trained based on the common transaction features of the transaction samples of the target enterprise, and the second branch after initialization is iteratively trained based on the differential transaction features of the transaction samples of the target enterprise, so as to obtain the transaction risk detection model.

5. The method according to claim 1, wherein the transaction sample further includes differential transaction features, and the neural network to be trained comprises: First branch, second branch, merged layer, and first fully connected layer; The iterative training process for the neural network to be trained includes: The first sample features are obtained by performing a first linear transformation on the shared transaction features through the first branch. The second branch is used to perform a second linear transformation on the differential transaction features to obtain the second sample features; The first sample feature and the second sample feature are merged through the merging layer to obtain the third sample feature; The first fully connected layer maps the features of the third sample to obtain the output result.

6. The method according to claim 5, wherein the first branch comprises a first input layer and A second fully connected layer, the second branch including a second input layer and N third fully connected layers; wherein... And N is an integer greater than 1; The first sample feature is obtained by performing a first linear transformation on the shared transaction features through the first branch, including: The shared transaction features are input into the first input layer, and then... Each second fully connected layer performs a first linear transformation on the first input features to obtain first sample features; wherein, the first input features of the first second fully connected layer include the shared transaction features, and the second second fully connected layer performs a first linear transformation on the first input features. The first input feature of the second fully connected layer includes the first The output features of the second fully connected layer ; The second branch is used to perform a second linear transformation on the differential transaction features to obtain second sample features, including: The differential transaction characteristics are input into the second input layer, and then... The third fully connected layer performs a second linear transformation on the second input features to obtain the second sample features; wherein, the second input features of the first third fully connected layer include the differential transaction features, and the third... The input features of the third fully connected layer include the first The output features of the third fully connected layer .

7. A method for detecting transaction risk, comprising: Obtain transaction data for the target company's target transaction; Extract the transaction features from the transaction data; The transaction features are processed by a pre-trained transaction risk detection model to obtain the risk detection result of the target transaction; wherein the transaction risk detection model is trained using the training method described in any one of claims 1-6.

8. A training device for a transaction risk detection model, comprising: The acquisition module retrieves the transaction sample set to be trained. The transaction sample set includes individual transaction samples, non-target enterprise transaction samples, and target enterprise transaction samples. The first number of individual transaction samples is greater than the second number of non-target enterprise transaction samples, and the second number is greater than the third number of target enterprise transaction samples. Each transaction sample includes common transaction features for knowledge transfer. The training module uses the personal transaction samples to iteratively train the neural network to be trained, thereby obtaining a first model; the first model includes first model parameters for knowledge transfer. as well as, The second neural network to be trained is iteratively trained using the transaction samples of the non-target enterprise and the parameters of the first model to obtain the second model; the second model includes second model parameters for knowledge transfer. The transaction risk detection model is obtained by iteratively training the third neural network to be trained using the transaction samples of the target company and the parameters of the second model; the transaction risk detection model is used to detect risks in the transaction data of the target company.

9. A transaction risk detection device, comprising: The acquisition module retrieves transaction data for the target company's target transactions. The extraction module extracts the transaction features from the transaction data; The detection module uses a pre-trained transaction risk detection model to perform risk detection processing on the transaction features to obtain the risk detection result of the target transaction; wherein the transaction risk detection model is trained using the training method described in any one of claims 1-6.

10. A training device for a transaction risk detection model, comprising: processor; as well as; A memory configured to store computer-executable instructions, which, when executed, cause the processor to: Obtain a transaction sample set to be trained; the transaction sample set includes individual transaction samples, non-target enterprise transaction samples, and target enterprise transaction samples, wherein the first number of individual transaction samples is greater than the second number of non-target enterprise transaction samples, and the second number is greater than the third number of target enterprise transaction samples; each transaction sample includes common transaction features for knowledge transfer. The neural network to be trained is iteratively trained using the individual transaction samples to obtain a first model; the first model includes first model parameters for knowledge transfer. The second neural network to be trained is iteratively trained using the transaction samples of the non-target enterprise and the parameters of the first model to obtain the second model; the second model includes second model parameters for knowledge transfer. The transaction risk detection model is obtained by iteratively training the third neural network to be trained using the transaction samples of the target company and the parameters of the second model; the transaction risk detection model is used to detect risks in the transaction data of the target company.

11. A transaction risk detection device, comprising: processor; as well as; A memory configured to store computer-executable instructions, which, when executed, cause the processor to: Obtain transaction data for the target company's target transaction; Extract the transaction features from the transaction data; The transaction features are processed by a pre-trained transaction risk detection model to obtain the risk detection result of the target transaction; wherein the transaction risk detection model is trained using the training method described in any one of claims 1-6.

12. A storage medium for storing computer-executable instructions, which, when executed by a processor, perform the following process: Obtain a transaction sample set to be trained; the transaction sample set includes individual transaction samples, non-target enterprise transaction samples, and target enterprise transaction samples, wherein the first number of individual transaction samples is greater than the second number of non-target enterprise transaction samples, and the second number is greater than the third number of target enterprise transaction samples; each transaction sample includes common transaction features for knowledge transfer. The neural network to be trained is iteratively trained using the individual transaction samples to obtain a first model; the first model includes first model parameters for knowledge transfer. The second neural network to be trained is iteratively trained using the transaction samples of the non-target enterprise and the parameters of the first model to obtain the second model; the second model includes second model parameters for knowledge transfer. The transaction risk detection model is obtained by iteratively training the third neural network to be trained using the transaction samples of the target company and the parameters of the second model; the transaction risk detection model is used to detect risks in the transaction data of the target company.

13. A storage medium for storing computer-executable instructions, which, when executed by a processor, perform the following process: Obtain transaction data for the target company's target transaction; Extract the transaction features from the transaction data; A pre-trained transaction risk detection model is used to perform risk detection processing on the transaction features to obtain the risk detection result of the target transaction; wherein, The transaction risk detection model is trained using the training method described in any one of claims 1-6.

Images