Business risk assessment method and device, storage medium and electronic equipment

Abstract

The invention discloses a business risk assessment method and device, a storage medium and electronic equipment. The method comprises the steps that multiple sets of modal data, aiming at a target service, of a target account are acquired, and the multiple sets of modal data at least comprise text data, image data and audio data; multiple sets of modal features corresponding to the multiple sets of modal data are determined, and the multiple sets of modal data and the multiple sets of modal features are in one-to-one correspondence; performing feature fusion processing on the multiple groups of modal features to obtain target fusion features; and inputting the target fusion feature into a pre-trained business risk assessment model to obtain a business risk assessment result of the target account. According to the invention, a technical problem of low business risk assessment accuracy caused by risk assessment only considering single modal data in a business risk assessment method in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of data mining technology, and in particular to a business risk assessment method, device, storage medium and electronic device. Background Art

[0002] At present, traditional business risk assessment methods usually only target a certain type of data, such as text data or transaction data, and are unable to integrate multiple types of data for analysis, which increases the possibility of missing abnormal risk behaviors. A large amount of labeled data is required for training, but the risk data in reality is often very small, resulting in insufficient training data, affecting the accuracy and robustness of the model. False positives or negative positives are prone to occur, resulting in low precision.

[0003] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0004] The embodiments of the present invention provide a business risk assessment method, device, storage medium and electronic device to at least solve the technical problem that the business risk assessment method in the related art only considers single modality data for risk assessment, resulting in low accuracy of business risk assessment.

[0005] According to one aspect of an embodiment of the present invention, a business risk assessment method is provided, comprising: obtaining multiple groups of modal data for a target business of a target account, wherein the multiple groups of modal data include at least text data, image data and audio data; determining multiple groups of modal features corresponding to the multiple groups of modal data, wherein there is a one-to-one correspondence between the multiple groups of modal data and the multiple groups of modal features; performing feature fusion processing on the multiple groups of modal features to obtain target fusion features; inputting the target fusion features into a pre-trained business risk assessment model to obtain a business risk assessment result of the target account.

[0006] According to another aspect of an embodiment of the present invention, a business risk assessment device is also provided, including: a first acquisition module, used to obtain multiple groups of modal data of a target account for a target business, wherein the multiple groups of modal data include at least text data, image data and audio data; a first determination module, used to determine multiple groups of modal features corresponding to the multiple groups of modal data, wherein there is a one-to-one correspondence between the multiple groups of modal data and the multiple groups of modal features; a first fusion module, used to perform feature fusion processing on the multiple groups of modal features to obtain target fusion features; and a second determination module, used to input the target fusion features into a pre-trained business risk assessment model to obtain a business risk assessment result of the target account.

[0007] According to another aspect of an embodiment of the present invention, a non-volatile storage medium is provided, wherein the non-volatile storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing any one of the above-mentioned business risk assessment methods.

[0008] According to another aspect of an embodiment of the present invention, there is also provided an electronic device, comprising one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement any one of the above-mentioned business risk assessment methods.

[0009] In an embodiment of the present invention, multiple groups of modal data for a target business of a target account are obtained, wherein the multiple groups of modal data include at least text data, image data and audio data; multiple groups of modal features corresponding to the multiple groups of modal data are determined, wherein there is a one-to-one correspondence between the multiple groups of modal data and the multiple groups of modal features; feature fusion processing is performed on the multiple groups of modal features to obtain target fusion features; the target fusion features are input into a pre-trained business risk assessment model to obtain a business risk assessment result of the target account, thereby achieving the purpose of accurately assessing business risks, thereby realizing the technical effect of improving the accuracy of business risk assessment, and further solving the technical problem of low accuracy of business risk assessment caused by the business risk assessment method in the related technology considering only a single modal data for risk assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0011] Figure 1 is a schematic diagram of a business risk assessment method according to an embodiment of the present invention;

[0012] Figure 2 is a schematic diagram of an optional business risk assessment method according to an embodiment of the present invention;

[0013] Figure 3 is a schematic diagram of a business risk assessment device according to an embodiment of the present invention;

[0014] Figure 4 is a schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0015] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below 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 of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0016] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention 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 that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0017] According to an embodiment of the present invention, a method embodiment of business risk assessment is 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.

[0018] Figure 1 is a flow chart of a business risk assessment method according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0019] Step S102, obtaining multiple groups of modal data for the target business of the target account, wherein the multiple groups of modal data at least include text data, image data and audio data.

[0020] Optionally, this business risk assessment method integrates multiple data sources, such as text data, image data, and audio data, etc., which solves the problem of a single data source and can improve the accuracy and robustness of the model, thereby solving the problems of low accuracy and insufficient data volume of existing technologies and improving the accuracy of business risk assessment.

[0021] In an optional embodiment, initial modal data of the target account is obtained, wherein the text data in the initial modal data at least includes: business application data, transaction record data, the image data in the initial modal data at least includes: certification material data pre-uploaded by the target account, and the audio data in the initial modal data at least includes: business consulting data in audio form corresponding to the target account; the initial modal data is preprocessed to obtain the multiple groups of modal data, wherein the preprocessing includes at least one of the following: data cleaning processing, data deduplication processing, and data format conversion processing.

[0022] Optionally, obtain the initial modal data of the target account for the target business, including at least text data, image data and audio data, and then clean, deduplicate and convert the initial modal data, where data cleaning refers to cleaning data that has no practical meaning, data deduplication refers to deleting duplicate data, and data format conversion refers to converting the data into a specified format to make the data complete and consistent for the next step of analysis and modeling.

[0023] Step S104, determining multiple groups of modal features corresponding to the multiple groups of modal data, wherein there is a one-to-one correspondence between the multiple groups of modal data and the multiple groups of modal features.

[0024] Optionally, different feature extraction methods are used to extract features from text data, image data, and audio data to obtain corresponding modal feature representations for subsequent feature fusion and business risk prediction. The specific feature extraction method selects appropriate algorithms and models based on the specific tasks and data characteristics. The feature extractor of each modality is usually designed and trained according to the characteristics of the modality to extract the most representative modal features.

[0025] In an optional embodiment, the above-mentioned determination of the multiple groups of modal features corresponding to the above-mentioned multiple groups of modal data includes: performing text feature extraction processing on the text data included in the above-mentioned multiple groups of modal data to obtain text modal features corresponding to the text data included in the above-mentioned multiple groups of modal data; performing image feature extraction processing on the image data included in the above-mentioned multiple groups of modal data to obtain image modal features corresponding to the image data included in the above-mentioned multiple groups of modal data; performing audio feature extraction processing on the audio data included in the above-mentioned multiple groups of modal data to obtain audio modal features corresponding to the audio data included in the above-mentioned multiple groups of modal data.

[0026] Optionally, text modal data uses appropriate text feature extraction methods, such as Bag-of-Words and TF-IDF (Term Frequency-Inverse Document Frequency), to convert text into numerical feature representations. Image modal data uses image feature extraction methods such as Convolutional Neural Networks (CNN) to extract visual feature representations of images. Audio modal data uses acoustic feature extraction methods, such as Mel Frequency Cepstral Coefficients (MFCC), to convert audio into numerical feature representations. Using different feature extraction methods for different types of data can make full use of the characteristics of various data sources, obtain more accurate features, and improve the accuracy of business risk assessment.

[0027] Step S106, performing feature fusion processing on the above multiple groups of modal features to obtain target fusion features.

[0028] Optionally, the multimodal features are fused using the attention mechanism to establish a multimodal feature fusion model, which can more comprehensively consider the feature relationships and impacts between different modalities and within modalities, thereby improving the accuracy of business risk assessment.

[0029] In an optional embodiment, the above-mentioned feature fusion processing is performed on the above-mentioned multiple groups of modal features to obtain target fusion features, including: feature alignment processing is performed on the above-mentioned multiple groups of modal features to obtain multiple groups of processed modal features, wherein the dimensions of the above-mentioned multiple groups of modal features after processing are the same; feature fusion processing is performed on the above-mentioned multiple groups of processed modal features using a pre-trained multimodal feature fusion model to obtain the above-mentioned target fusion features, wherein the above-mentioned multimodal feature fusion model is trained based on multiple groups of sample modal features corresponding to multiple groups of sample modal data, wherein there is a one-to-one correspondence between the above-mentioned multiple groups of sample modal data and the above-mentioned multiple groups of sample modal features.

[0030] Optionally, for multiple groups of modalities, corresponding feature extraction models are used to extract modality-specific feature representations. Using feature alignment technology, the features extracted from different modalities are dimensionally aligned or transformed to ensure that they have the same dimension or representation. Then, a pre-trained multimodal feature fusion model is used to perform feature fusion processing on the processed multiple groups of modal features to obtain target fusion features, wherein a self-attention mechanism is trained to obtain a multimodal feature fusion model including a trained inter-modal weight parameter matrix and an intra-modal weight parameter matrix.

[0031] In an optional embodiment, the above-mentioned pre-trained multimodal feature fusion model is used to perform feature fusion processing on the above-mentioned processed multiple groups of modal features to obtain the above-mentioned target fusion features, including: based on the above-mentioned processed multiple groups of modal features and the pre-trained inter-modal weight parameter matrix, determining the first attention weights corresponding to the above-mentioned processed multiple groups of modal features respectively; based on the above-mentioned processed multiple groups of modal features and the pre-trained intra-modal weight parameter matrix, determining the second attention weights corresponding to the above-mentioned processed multiple groups of modal features respectively, wherein the above-mentioned multimodal feature fusion model includes the above-mentioned inter-modal weight parameter matrix and the above-mentioned intra-modal weight parameter matrix; based on the above-mentioned processed multiple groups of modal features, the first attention weights corresponding to the above-mentioned processed multiple groups of modal features respectively, and the second attention weights corresponding to the above-mentioned processed multiple groups of modal features respectively, to obtain the above-mentioned target fusion features.

[0032] Optionally, based on the processed multiple groups of modal features and the pre-trained inter-modal weight parameter matrix, the first attention weights corresponding to the processed multiple groups of modal features are obtained; based on the processed multiple groups of modal features and the pre-trained intra-modal weight parameter matrix, the second attention weights corresponding to the processed multiple groups of modal features are obtained. Finally, based on the processed multiple groups of modal features, the first attention weights corresponding to the processed multiple groups of modal features, and the second attention weights corresponding to the processed multiple groups of modal features, the target fusion features are obtained. The introduction of the attention mechanism for multi-modal feature fusion method can automatically learn the correlation and importance between different modalities and within the modalities, and fuse the features of different modalities according to the attention weights, so that the model can pay more attention to the features that contribute most to business risk assessment.

[0033] In an optional embodiment, the pre-trained multimodal feature fusion model is used to perform feature fusion processing on the processed multiple groups of modal features to obtain the target fusion features, including: based on the processed multiple groups of modal features and the inter-modal weight parameter matrix, determining the first attention weights corresponding to the processed multiple groups of modal features respectively in the following manner:

[0034]

[0035] in, represents the first attention weight corresponding to any one of the multiple groups of modal features after the above processing, softmax() represents the activation function, W_a represents the above inter-modal weight parameter matrix, and x_i represents any one of the above modal features; based on the multiple groups of modal features after the above processing and the above intra-modal weight parameter matrix, the second attention weights corresponding to the multiple groups of modal features after the above processing are determined in the following way:

[0036] β_i=softmax(W_b*x_i)

[0037] Wherein, β_i represents the second attention weight corresponding to any one of the above modal features, and W_b represents the weight parameter matrix within the above modality; based on the above processed multiple groups of modal features, the above first attention weight, and the above second attention weight, the above target fusion features corresponding to the above processed multiple groups of modal features are obtained in the following manner:

[0038]

[0039] Among them, h represents the above target fusion feature.

[0040] Optionally, first perform inter-modal attention calculation, calculate the inter-modal attention weights, and indicate the relevance and importance between different modalities. Assuming there are M modalities, the features of each modality can be mapped to the attention weights using a fully connected layer or other methods. The attention weights are normalized by the softmax function to ensure that the sum of the attention weights is 1, and the first attention weights corresponding to the processed multiple groups of modal features are obtained. Secondly, perform intra-modal attention calculation, and for each modality, calculate the intra-modal attention weights to indicate the importance of different features within the modality. A similar method can be used to map the features within each modality to the attention weights using a fully connected layer or other methods. Similarly, normalize the attention weights by the softmax function to ensure that the sum of the attention weights is 1, and obtain the second attention weights corresponding to the processed multiple groups of modal features. Finally, perform feature fusion, multiply the attention weights by the corresponding features, and then sum the weighted features of all modalities to obtain the target fused features.

[0041] Step S108, inputting the above target fusion features into a pre-trained business risk assessment model to obtain the business risk assessment result of the above target account.

[0042] Optionally, the target fusion feature is input into a pre-trained business risk assessment model, where the business risk assessment model is a support vector machine (SVM) model, and finally the business risk assessment result of the target account is obtained. This method improves the accuracy and real-time performance of business risk assessment, and at the same time improves the interpretability of the model.

[0043] In an optional embodiment, before inputting the above-mentioned target fusion features into a pre-trained business risk assessment model to obtain the business risk assessment results of the above-mentioned target account, the above-mentioned method also includes: obtaining the above-mentioned multiple groups of sample modal data for the above-mentioned target business; determining the sample fusion features corresponding to the above-mentioned multiple groups of sample modal data, and the risk labels corresponding to the above-mentioned sample fusion features, wherein the above-mentioned risk labels are: risky or risk-free; performing model training based on the above-mentioned sample fusion features and the above-mentioned risk labels to obtain the above-mentioned business risk assessment model.

[0044] Optionally, first obtain multiple groups of sample modal data for the target business, then fuse the multiple groups of modal features corresponding to the multiple groups of sample modal data to obtain sample fusion features, determine the risk labels corresponding to the sample fusion features, where the risk labels are: risky or risk-free, and finally perform model training based on the sample fusion features and risk labels to obtain a business risk assessment model based on a support vector machine.

[0045] Through the above steps S102 to S108, the purpose of accurately assessing business risks can be achieved, thereby realizing the technical effect of improving the accuracy of business risk assessment, and further solving the technical problem of low accuracy of business risk assessment caused by the business risk assessment method in the related technology only considering single modal data for risk assessment.

[0046] Based on the above embodiments and optional embodiments, the present invention proposes an optional implementation mode: Figure 2 is a flow chart of an optional business risk assessment method according to an embodiment of the present invention. Figure 2 As shown, the method includes: a model training phase and a model application phase, wherein the model training phase includes the training of a multimodal feature fusion model and the training of a business risk assessment model, specifically:

[0047] The training of the multimodal feature fusion model is achieved in the following way:

[0048] Step S11, collecting multiple groups of sample modal data corresponding to the target business, including text modal data, image modal data and audio modal data.

[0049] Step S12, performing data cleaning processing, data deduplication processing and data format conversion processing on the multiple groups of sample modal data to obtain the processed multiple groups of sample modal data.

[0050] Step S13, feature extraction. Perform feature extraction processing on the processed multiple groups of sample modal data to obtain multiple groups of sample modal features, specifically including: using appropriate text feature extraction methods, such as bag-of-words model, TF-IDF, etc., to convert text into numerical feature representations for the text modal data included in the processed multiple groups of sample modal data. Using image feature extraction methods such as convolutional neural networks to extract visual feature representations of images for the image modal data included in the processed multiple groups of sample modal data. Using acoustic feature extraction methods, such as Mel-spectrogram coefficients, etc. to convert audio modal data included in the processed multiple groups of sample modal data into numerical feature representations.

[0051] Step S14, performing feature alignment processing on the multiple groups of sample modal features to obtain the processed multiple groups of sample modal features.

[0052] Step S15, dividing the processed groups of sample modal features into a training set and a test set, using the training set to train the model, and using the test set to evaluate the accuracy and generalization ability of the model.

[0053] Step S16, training the self-attention mechanism based on the training set to obtain a trained model; testing the trained model based on the test set, and when the test result indicates that the trained model has passed the test, using the trained model as a multimodal feature fusion model, wherein the multimodal feature fusion model includes a trained inter-modal weight parameter matrix and an intra-modal weight parameter matrix.

[0054] The training of the business risk assessment model is achieved in the following ways:

[0055] Step S21, obtaining multiple groups of sample modal data for the target business.

[0056] Step S22, fusing the multiple groups of modal features corresponding to the multiple groups of sample modal data to obtain sample fusion features, and determining the risk labels corresponding to the sample fusion features, wherein the risk labels are: risky or non-risky.

[0057] Step S23, performing model training based on sample fusion features and risk labels to obtain a pre-trained business risk assessment model based on a support vector machine.

[0058] Deploy the trained multimodal feature fusion model and business risk assessment model to the production environment to perform real-time customer segmentation operations. The model is applied in the production environment in the following ways:

[0059] Step S1, collecting multiple groups of modal data corresponding to the target business, including text modal data, image modal data and audio modal data.

[0060] Step S2, performing data cleaning processing, data deduplication processing and data format conversion processing on the multiple groups of modal data to obtain the processed multiple groups of modal data.

[0061] Step S3, feature extraction. Perform feature extraction processing on the processed multiple groups of modal data to obtain multiple groups of modal features, specifically including: using appropriate text feature extraction methods, such as bag-of-words model, TF-IDF, etc., to convert text into numerical feature representations for the text modal data included in the processed multiple groups of modal data. Using image feature extraction methods such as convolutional neural networks to extract visual feature representations of images for the image modal data included in the processed multiple groups of modal data. Using acoustic feature extraction methods, such as Mel-spectrogram coefficients, etc. to convert audio modal data included in the processed multiple groups of modal data into numerical feature representations.

[0062] Step S4, performing feature alignment processing on the multiple groups of modal features to obtain multiple groups of processed modal features.

[0063] Step S5, using a pre-trained multimodal feature fusion model to perform feature fusion processing on the processed multiple groups of modal features to obtain target fusion features, specifically includes the following sub-steps:

[0064] Step S51, based on the processed multiple groups of modal features and the inter-modal weight parameter matrix, obtain the first attention weights corresponding to the processed multiple groups of modal features respectively in the following manner:

[0065]

[0066] in, It represents the first attention weight corresponding to any modal feature in the processed multiple groups of modal features, softmax() represents the activation function, W_a represents the inter-modal weight parameter matrix, and x_i represents any modal feature.

[0067] Step S52, based on the processed multiple groups of modal features and the intra-modal weight parameter matrix, obtain the second attention weights corresponding to the processed multiple groups of modal features respectively in the following manner:

[0068] β_i=softmax(W_b*x_i)

[0069] Among them, β_i represents the second attention weight corresponding to any modal feature, and W_b represents the intra-modal weight parameter matrix.

[0070] Step S53, based on the processed multiple groups of modal features, the first attention weight, and the second attention weight, obtain the target fusion features corresponding to the processed multiple groups of modal features in the following manner:

[0071]

[0072] Among them, h represents the target fusion feature.

[0073] Step S6: input the target fusion features into a pre-trained business risk assessment model to obtain the business risk assessment result of the target account.

[0074] In step S7, the business risk assessment results may be evaluated using standard evaluation indicators, such as accuracy, recall, F1 score, etc.

[0075] The embodiments of the present invention can at least achieve the following technical effects: 1. A multimodal feature fusion method is adopted, which can comprehensively utilize feature information of multiple modalities such as text, image, audio, etc. Compared with the method of only using a single modality data for analysis in the prior art, it can capture business risk features more comprehensively and improve the accuracy of business risk assessment. 2. An attention mechanism is introduced for feature fusion, which can automatically learn and pay attention to the importance of features of different modalities. Through the application of the attention mechanism, features of different modalities can be fused more accurately, so that important features receive more attention, thereby improving the accuracy of business risk assessment.

[0076] In this embodiment, a business risk assessment device is also provided, which is used to implement the above embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the terms "module" and "device" can implement a combination of software and / or hardware for a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0077] According to an embodiment of the present invention, there is also provided an apparatus embodiment for implementing the above-mentioned business risk assessment method. Figure 3 is a schematic diagram of the structure of a business risk assessment device according to an embodiment of the present invention. Figure 3 As shown, the business risk assessment device includes: a first acquisition module 302, a first determination module 304, a first fusion module 306, and a second determination module 308, wherein:

[0078] A first acquisition module 302 is used to acquire multiple sets of modal data of a target account for a target business, wherein the multiple sets of modal data at least include text data, image data, and audio data;

[0079] A first determination module 304, connected to the first acquisition module 302, is used to determine multiple groups of modal features corresponding to the multiple groups of modal data, wherein there is a one-to-one correspondence between the multiple groups of modal data and the multiple groups of modal features;

[0080] A first fusion module 306, connected to the first determination module 304, is used to perform feature fusion processing on the multiple groups of modal features to obtain target fusion features;

[0081] The second determination module 308 is connected to the first fusion module 306 and is used to input the target fusion feature into a pre-trained business risk assessment model to obtain the business risk assessment result of the target account.

[0082] It should be noted that the above modules can be implemented by software or hardware. For example, for the latter, it can be implemented in the following ways: the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0083] It should be noted that the above: first acquisition module 302, first determination module 304, first fusion module 306, second determination module 308 correspond to steps S102 to S108 in the embodiment, and the examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above embodiments. It should be noted that the above modules can be run in a computer terminal as part of the device.

[0084] It should be noted that the optional or preferred implementation of this embodiment can refer to the relevant description in the embodiment, which will not be repeated here.

[0085] The above-mentioned business risk assessment device may also include a processor and a memory. The above-mentioned: the first acquisition module 302, the first determination module 304, the first fusion module 306, the second determination module 308, etc. are all stored in the memory as program modules, and the processor executes the above-mentioned program modules stored in the memory to realize corresponding functions.

[0086] The processor includes a kernel, which retrieves the corresponding program module from the memory. The kernel may be one or more. The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. The memory includes at least one memory chip.

[0087] According to an embodiment of the present application, an embodiment of a non-volatile storage medium is also provided. Optionally, in this embodiment, the non-volatile storage medium includes a stored program, wherein when the program is executed, the device where the non-volatile storage medium is located is controlled to execute any of the above-mentioned business risk assessment methods.

[0088] Optionally, in this embodiment, the non-volatile 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, and the non-volatile storage medium includes a stored program.

[0089] Optionally, when the program is running, the device where the non-volatile storage medium is located is controlled to perform the following functions: obtain multiple groups of modal data for the target business of the target account, wherein the multiple groups of modal data include at least text data, image data and audio data; determine multiple groups of modal features corresponding to the multiple groups of modal data, wherein there is a one-to-one correspondence between the multiple groups of modal data and the multiple groups of modal features; perform feature fusion processing on the multiple groups of modal features to obtain target fusion features; input the target fusion features into a pre-trained business risk assessment model to obtain the business risk assessment results of the target account.

[0090] According to an embodiment of the present application, an embodiment of a processor is also provided. Optionally, in this embodiment, the processor is used to run a program, wherein the program executes any one of the business risk assessment methods when it is run.

[0091] According to an embodiment of the present application, an embodiment of a computer program product is also provided, which, when executed on a data processing device, is suitable for executing a program that initializes any one of the steps of the above-mentioned business risk assessment method.

[0092] Optionally, the above-mentioned computer program product, when executed on a data processing device, is suitable for executing a program initialized with the following method steps: obtaining multiple groups of modal data for a target account for a target business, wherein the multiple groups of modal data include at least text data, image data and audio data; determining multiple groups of modal features corresponding to the multiple groups of modal data, wherein there is a one-to-one correspondence between the multiple groups of modal data and the multiple groups of modal features; performing feature fusion processing on the multiple groups of modal features to obtain target fusion features; inputting the target fusion features into a pre-trained business risk assessment model to obtain the business risk assessment results of the above-mentioned target account.

[0093] like Figure 4As shown, an embodiment of the present invention provides an electronic device, which includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, the following steps are implemented: obtaining multiple groups of modal data of a target account for a target business, wherein the multiple groups of modal data include at least text data, image data, and audio data; determining multiple groups of modal features corresponding to the multiple groups of modal data, wherein there is a one-to-one correspondence between the multiple groups of modal data and the multiple groups of modal features; performing feature fusion processing on the multiple groups of modal features to obtain target fusion features; inputting the target fusion features into a pre-trained business risk assessment model to obtain a business risk assessment result of the target account.

[0094] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0095] 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 above modules can be a logical function division. There may be other division methods in actual implementation, such as multiple modules 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, modules or indirect coupling or communication connection of modules, which can be electrical or other forms.

[0096] The modules described above as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment.

[0097] In addition, each functional module in each embodiment of the present invention may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module. The above integrated modules may be implemented in the form of hardware or in the form of software functional modules.

[0098] If the above-mentioned integrated module is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable non-volatile storage medium. Based on this understanding, the technical solution of the present invention 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. The computer software product is stored in a non-volatile storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the various embodiments of the present invention. The aforementioned non-volatile storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0099] The above are only preferred embodiments of the present invention. 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 invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A business risk assessment method, characterized in that: include: Acquire multiple sets of modal data for a target business from a target account, wherein the multiple sets of modal data include at least text data, image data, and audio data; Determining multiple groups of modal features corresponding to the multiple groups of modal data, wherein there is a one-to-one correspondence between the multiple groups of modal data and the multiple groups of modal features; Performing feature fusion processing on the multiple groups of modal features to obtain target fusion features; The target fusion feature is input into a pre-trained business risk assessment model to obtain a business risk assessment result of the target account.

2. The method according to claim 1, characterized in that The determining of the multiple groups of modal features corresponding to the multiple groups of modal data includes: Performing text feature extraction processing on the text data included in the multiple groups of modal data to obtain text modal features corresponding to the text data included in the multiple groups of modal data; Performing image feature extraction processing on the image data included in the multiple groups of modal data to obtain image modal features corresponding to the image data included in the multiple groups of modal data; Audio feature extraction processing is performed on the audio data included in the multiple groups of modal data to obtain audio modal features corresponding to the audio data included in the multiple groups of modal data.

3. The method according to claim 1, characterized in that The step of performing feature fusion processing on the multiple groups of modal features to obtain target fusion features includes: Performing feature alignment processing on the multiple groups of modal features to obtain multiple groups of modal features after processing, wherein the multiple groups of modal features after processing have the same dimensions; A pre-trained multimodal feature fusion model is used to perform feature fusion processing on the processed multiple groups of modal features to obtain the target fusion features, wherein the multimodal feature fusion model is trained based on multiple groups of sample modal features corresponding to multiple groups of sample modal data, wherein there is a one-to-one correspondence between the multiple groups of sample modal data and the multiple groups of sample modal features.

4. The method according to claim 3, characterized in that The method of using a pre-trained multimodal feature fusion model to perform feature fusion processing on the processed multiple groups of modal features to obtain the target fusion feature includes: Based on the processed multiple groups of modal features and the pre-trained inter-modal weight parameter matrix, determine the first attention weights corresponding to the processed multiple groups of modal features respectively; Based on the processed multiple groups of modal features and the pre-trained intra-modal weight parameter matrix, determine the second attention weights corresponding to the processed multiple groups of modal features respectively, wherein the multimodal feature fusion model includes the inter-modal weight parameter matrix and the intra-modal weight parameter matrix; Based on the processed multiple groups of modal features, the first attention weights respectively corresponding to the processed multiple groups of modal features, and the second attention weights respectively corresponding to the processed multiple groups of modal features, the target fusion feature is obtained.

5. The method according to claim 4, characterized in that The method of using a pre-trained multimodal feature fusion model to perform feature fusion processing on the processed multiple groups of modal features to obtain the target fusion feature includes: Based on the processed multiple groups of modal features and the inter-modal weight parameter matrix, the first attention weights corresponding to the processed multiple groups of modal features are determined in the following manner: in, represents the first attention weight corresponding to any one of the processed multiple groups of modal features, softmax() represents the activation function, W_a represents the inter-modal weight parameter matrix, and x_i represents any one of the modal features; Based on the processed multiple groups of modal features and the intra-modal weight parameter matrix, the second attention weights corresponding to the processed multiple groups of modal features are determined in the following manner: β_i=softmax(W_b*x_i) Wherein, β_i represents the second attention weight corresponding to any one of the modal features, and W_b represents the weight parameter matrix within the modality; Based on the processed multiple groups of modal features, the first attention weight, and the second attention weight, the target fusion features corresponding to the processed multiple groups of modal features are obtained in the following manner: Wherein, h represents the target fusion feature.

6. The method according to claim 3, characterized in that Before inputting the target fusion feature into a pre-trained business risk assessment model to obtain the business risk assessment result of the target account, the method further includes: Acquire the multiple groups of sample modal data for the target business; Determine sample fusion features corresponding to the multiple groups of sample modal data, and risk labels corresponding to the sample fusion features, wherein the risk label is: risky or non-risky; Model training is performed based on the sample fusion features and the risk labels to obtain the business risk assessment model.

7. The method according to any one of claims 1 to 6, characterized in that: The obtaining of multiple groups of modal data of the target account for the target business includes: Acquire the initial modal data of the target account, wherein the text data in the initial modal data at least includes: business application data, transaction record data, the image data in the initial modal data at least includes: certification material data pre-uploaded by the target account, and the audio data in the initial modal data at least includes: business consultation data in audio form corresponding to the target account; The initial modal data is preprocessed to obtain the multiple groups of modal data, wherein the preprocessing includes at least one of the following: data cleaning processing, data deduplication processing, and data format conversion processing.

8. A business risk assessment device, characterized in that: include: A first acquisition module is used to acquire multiple sets of modal data of a target account for a target business, wherein the multiple sets of modal data include at least text data, image data, and audio data; A first determining module, configured to determine a plurality of groups of modal features corresponding to the plurality of groups of modal data, wherein there is a one-to-one correspondence between the plurality of groups of modal data and the plurality of groups of modal features; A first fusion module is used to perform feature fusion processing on the multiple groups of modal features to obtain target fusion features; The second determination module is used to input the target fusion feature into a pre-trained business risk assessment model to obtain a business risk assessment result of the target account.

9. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the business risk assessment method described in any one of claims 1 to 7.

10. An electronic device, characterized in that: It includes one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the business risk assessment method described in any one of claims 1 to 7.

Images