Data processing method, device and equipment

Abstract

The embodiment of the invention provides a data processing method, device and equipment, and relates to the technical field of computers.The method comprises the steps that a risk detection request for a target merchant is received; in response to the risk detection request, obtaining to-be-detected data corresponding to the target merchant; performing modal alignment processing on data of different modals in the to-be-detected data to obtain target text data corresponding to the to-be-detected data; according to a large language model, performing information fusion processing on the target text data to obtain target information corresponding to the target merchant; obtaining target to-be-detected data corresponding to the risk detection demand of the target merchant in the to-be-detected data, and performing feature recognition processing on the target to-be-detected data to obtain feature information of the target merchant; and determining a risk detection result for the target merchant according to the target information and the feature information.

Description

Technical Field

[0001] This document relates to the field of computer technology, and in particular to a data processing method, device and equipment. Background Art

[0002] As people pay more and more attention to their private data, in order to protect user privacy and ensure data security, it is necessary to detect whether the merchants providing services to users are risky merchants to ensure user data security. For example, security personnel can manually analyze merchant data to detect whether the merchant is a risky merchant.

[0003] However, as the data structure of merchant data becomes more and more complex and the amount of data becomes larger and larger, the detection efficiency and accuracy of risk detection by manual analysis are low. To this end, the embodiments of this specification provide a better technical solution for risk detection of merchants. Summary of the invention

[0004] The purpose of the embodiments of this specification is to provide a better technical solution for risk detection of merchants.

[0005] In order to implement the above technical solution, the embodiments of this specification are implemented as follows: A data processing method provided in an embodiment of the present specification includes: receiving a risk detection request for a target merchant; in response to the risk detection request, obtaining data to be detected corresponding to the target merchant, wherein the data to be detected includes multimodal data related to the business operation behavior of the target merchant; performing modal alignment processing on data of different modalities in the data to be detected to obtain target text data corresponding to the data to be detected; performing information fusion processing on the target text data according to a large language model to obtain target information corresponding to the target merchant; obtaining target data to be detected corresponding to the risk detection requirements of the target merchant in the data to be detected, and performing feature recognition processing on the target data to be detected to obtain feature information of the target merchant; and determining a risk detection result for the target merchant based on the target information and the feature information.

[0006] A data processing device provided in an embodiment of the present specification includes: a request receiving module, which is used to receive a risk detection request for a target merchant; a data acquisition module, which is used to respond to the risk detection request and obtain the data to be detected corresponding to the target merchant, wherein the data to be detected includes multimodal data related to the business operation behavior of the target merchant; a modality alignment module, which is used to perform modality alignment processing on data of different modalities in the data to be detected to obtain target text data corresponding to the data to be detected; an information fusion module, which is used to perform information fusion processing on the target text data according to a large language model to obtain target information corresponding to the target merchant; a feature recognition module, which is used to obtain target data to be detected in the data to be detected that corresponds to the risk detection requirements of the target merchant, and perform feature recognition processing on the target data to be detected to obtain feature information of the target merchant; and a risk detection module, which is used to determine the risk detection result for the target merchant based on the target information and the feature information.

[0007] A data processing device is provided in an embodiment of the present specification, and the data processing device includes: a processor; and a memory arranged to store computer-executable instructions, wherein the executable instructions, when executed, cause the processor to: receive a risk detection request for a target merchant; in response to the risk detection request, obtain data to be detected corresponding to the target merchant, wherein the data to be detected includes multimodal data related to the business operation behavior of the target merchant; perform modality alignment processing on data of different modalities in the data to be detected to obtain target text data corresponding to the data to be detected; perform information fusion processing on the target text data according to a large language model to obtain target information corresponding to the target merchant; obtain target data to be detected in the data to be detected that corresponds to the risk detection requirement of the target merchant, and perform feature recognition processing on the target data to be detected to obtain feature information of the target merchant; determine the risk detection result for the target merchant based on the target information and the feature information.

[0008] The embodiments of the present specification also provide a storage medium, which is used to store computer-executable instructions, and the executable instructions implement the following process when executed by a processor: receiving a risk detection request for a target merchant; in response to the risk detection request, obtaining data to be detected corresponding to the target merchant, wherein the data to be detected includes multimodal data related to the business operation behavior of the target merchant; performing modal alignment processing on data of different modalities in the data to be detected to obtain target text data corresponding to the data to be detected; performing information fusion processing on the target text data according to a large language model to obtain target information corresponding to the target merchant; obtaining target data to be detected in the data to be detected that corresponds to the risk detection requirement of the target merchant, and performing feature recognition processing on the target data to be detected to obtain feature information of the target merchant; determining the risk detection result for the target merchant based on the target information and the feature information.

[0009] The embodiments of the present specification also provide a computer program product, including a computer program, which implements the following process when executed by a processor: receiving a risk detection request for a target merchant; in response to the risk detection request, obtaining data to be detected corresponding to the target merchant, wherein the data to be detected includes multimodal data related to the business operation behavior of the target merchant; performing modal alignment processing on data of different modalities in the data to be detected to obtain target text data corresponding to the data to be detected; performing information fusion processing on the target text data according to a large language model to obtain target information corresponding to the target merchant; obtaining target data to be detected corresponding to the risk detection requirements of the target merchant in the data to be detected, and performing feature recognition processing on the target data to be detected to obtain feature information of the target merchant; determining the risk detection result for the target merchant based on the target information and the feature information. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the drawings required for use in the embodiments or the prior art description will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative labor. Figure 1 This is an embodiment of a data processing method of this specification; Figure 2 This is another data processing method embodiment of the present specification; Figure 3 A schematic diagram of a data processing process of this specification; Figure 4 This is a schematic diagram of another data processing process of this specification; Figure 5 This is an embodiment of a data processing device of the present specification; Figure 6 This is an embodiment of a data processing device in this specification. DETAILED DESCRIPTION

[0011] The embodiments of this specification provide a data processing method, device and equipment.

[0012] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0013] The embodiments of this specification provide a detection mechanism for whether a merchant is a risky merchant. As people pay more and more attention to their own privacy data, in order to protect user privacy and ensure data security, it is necessary to detect whether the merchant providing services to users is a risky merchant to ensure user data security. For example, security personnel can manually analyze merchant data to detect whether the merchant is a risky merchant. However, as the data structure of merchant data becomes more and more complex and the amount of data becomes larger and larger, the detection efficiency and accuracy of risk detection through manual analysis are low. To this end, the embodiments of the present specification provide a better technical solution for risk detection of merchants. In this solution, a risk detection request for a target merchant is received, and in response to the risk detection request, data to be detected corresponding to the target merchant is obtained, the data to be detected includes multimodal data related to the business operation behavior of the target merchant, modal alignment processing is performed on data of different modalities in the data to be detected to obtain target text data corresponding to the data to be detected, and information fusion processing is performed on the target text data according to a large language model to obtain target information corresponding to the target merchant, and target data to be detected corresponding to the risk detection requirements of the target merchant in the data to be detected is obtained, and feature recognition processing is performed on the target data to be detected to obtain feature information of the target merchant, and the risk detection result for the target merchant is determined based on the target information and the feature information. In this way, on the one hand, when the data structure of the merchant data becomes more and more complex, the multimodal data contained in the data to be detected can be aligned to the text data through modal alignment processing, and then the aligned data (i.e., the target text data) can be subjected to information fusion processing by utilizing the understanding, summarization and reasoning capabilities of the large language model to obtain the target information corresponding to the target merchant. On the other hand, the target data to be detected corresponding to the risk detection requirements of the target merchant can be subjected to feature recognition processing to obtain feature information that can meet the risk detection requirements. In this way, by being able to comprehensively and accurately describe the target information of the target merchant and the feature information that can meet the risk detection requirements, the risk detection results for the target merchant can be quickly and accurately determined, thereby improving the detection efficiency and accuracy of risk detection for the merchant. For specific processing, please refer to the specific content in the following embodiments.

[0014] like Figure 1 As shown, the embodiment of this specification provides a data processing method, the execution subject of the method can be a server, wherein the server can be an independent server, or a server cluster composed of multiple servers, etc. The server can be a background server for financial services or online shopping services, or a background server for an application, etc. In this embodiment, the execution subject is taken as an example for detailed description, and the method can specifically include the following steps: In step S102, a risk detection request for a target merchant is received.

[0015] The target merchant may be any physical merchant or a virtual merchant on the platform that can provide services to users. For example, the target merchant may be a physical merchant that provides catering services to users, or the target merchant may be a registered merchant on an online shopping platform that provides shopping services to users.

[0016] In implementation, the server may periodically trigger a risk detection request for the target merchant according to a preset detection cycle. For example, the server may trigger a risk detection request for the target merchant every 10 days.

[0017] In addition, since different types of merchants have different detection requirements, such as physical merchants have lower risk detection requirements, while virtual merchants on the platform have higher risk detection requirements, the server can set different detection cycles for different types of merchants. For example, the server can trigger a risk detection request for virtual merchants (i.e., target merchants) on the online shopping platform every 10 days, or the server can also trigger a risk detection request for physical merchants (i.e., target merchants) every half a month.

[0018] In addition, the server can also detect whether the merchant data of the merchant has changed, and trigger a risk detection request for the merchant (i.e., the target merchant) when the merchant data is detected to have changed. For example, when the server detects that the merchant name of a merchant has changed, the server can determine the merchant as a target merchant and trigger a risk detection request for the target merchant.

[0019] The above-mentioned method for obtaining risk detection requests for target merchants is an optional and feasible method for obtaining. In actual application scenarios, there may be a variety of different methods for obtaining. Different methods for obtaining may be selected according to different actual application scenarios. The embodiments of this specification do not specifically limit this.

[0020] In step S104, in response to the risk detection request, the to-be-detected data corresponding to the target merchant is obtained.

[0021] Among them, the data to be detected may include multimodal data related to the business operations of the target merchant. For example, the data to be detected may include but is not limited to text data, image data, video data, and audio data. Specifically, the text data may include the merchant name, address, industry type and other data of the target merchant, the image data may be image data of the merchant logo of the target merchant, etc., the video data may be relevant video data of the service provision process of the target merchant, and the audio data may include interaction data between the target merchant and the user, etc.

[0022] In step S106, modality alignment is performed on data of different modalities in the data to be detected to obtain target text data corresponding to the data to be detected.

[0023] In implementation, the server can perform modality alignment processing on data of different modalities in the data to be detected according to a pre-trained alignment model to obtain target text data corresponding to the data to be detected. The alignment model can be a model built based on a preset machine learning algorithm.

[0024] In addition, the server can also determine the model according to the labels corresponding to the data of different models, process the data of different modes respectively, obtain the label information corresponding to the data of different modes, and then determine the target text data corresponding to the data to be detected according to the label information corresponding to the data of different modes. The label determination model can be a model built based on a preset machine learning algorithm.

[0025] For example, assuming that the data to be detected includes text data and image data, wherein the text data includes the merchant name of the target merchant, and the image data includes the merchant logo of the target merchant, the server can determine the model based on the label corresponding to the text data, determine the label type corresponding to the text data, determine the model based on the label corresponding to the image data, determine the label type corresponding to the image data, and determine the target text data based on the label type corresponding to the text data and the label type corresponding to the image data, wherein the label type of the text data and the label type of the image data can be used to characterize the merchant type of the target merchant.

[0026] In addition, the above-mentioned modal alignment processing method is an optional and feasible processing method. In actual application scenarios, there may be a variety of different modal alignment processing methods. Different modal alignment processing methods can be selected according to different actual application scenarios. The embodiments of this specification do not make specific limitations on this.

[0027] In step S108, information fusion processing is performed on the target text data according to the large language model to obtain target information corresponding to the target merchant.

[0028] Among them, large language models (LLMs) can be advanced natural language processing (NLP) models built on the basis of deep learning technology. Their underlying converters are a set of neural networks, which can be composed of encoders and decoders with self-attention functions. They can understand and generate human language by processing large amounts of text data, thereby performing various natural language processing tasks. Since large language models have a large number of parameters and a wide range of training data sets, they can capture the richness and subtle differences of language expressions, and can display the captured information through powerful computing power of billions to trillions of parameters, and have strong capabilities in language understanding and generation tasks.

[0029] In implementation, after converting the data of different modes in the data to be detected into text description form, the obtained target text data can be input into the large language model, and through the second prompt information, such as "Please help me integrate all the objective information of the merchant, simplify the description of the transaction type, and describe the relationship between different merchants, especially the relationship with the legal person, as extensively as possible", the text description of each mode in the target text data is integrated into the merchant meta description (Meta-Tokens). The large language model can output the merchant meta description (i.e., target information) containing the objective factual information of the target merchant in a more coherent and meaningful summary form.

[0030] The server can use the understanding ability of the large language model to perform information fusion processing on the target text data, that is, the server can input the target text data and the second prompt information (ie, promote) into the large language model to obtain target information that can be used to describe the target merchant.

[0031] For example, the target information of the target merchant output by the large language model can be "The merchant name of the merchant is "xxx", and according to the merchant name, the industry in which the merchant is located is the catering industry. According to the image data of the merchant, the service provided by the merchant is catering service. According to the transaction behavior data of the merchant, it can be determined that the business hours of the merchant are concentrated between 3 am and 4 am."

[0032] In step S110, target data to be detected that corresponds to the risk detection requirements of the target merchant is obtained from the data to be detected, and feature recognition processing is performed on the target data to be detected to obtain feature information of the target merchant.

[0033] Among them, risk detection needs may include dynamic information detection needs (such as transaction anomaly detection needs, etc.) and static information detection needs (such as merchant name conflict detection needs, industry conflict detection needs, etc.).

[0034] In implementation, for example, taking the target merchant's risk detection requirement as an industry conflict detection requirement, the server can obtain target data to be detected in the detection requirement that can be used to determine the industry in which the target merchant is located (such as text data containing the merchant name of the target merchant, and image data containing the merchant logo of the target user, etc.).

[0035] In addition, the server can also determine the incremental data in the data to be detected that corresponds to the risk detection requirements of the target merchant as the target data to be detected. For example, the server can obtain the incremental data added to the data to be detected within a preset detection cycle, and determine the data in the incremental data that corresponds to the risk detection requirements of the target merchant as the target data to be detected.

[0036] In this way, the server can use the understanding ability of the large language model to perform information fusion processing on the data to be detected with large data volume and complex data structure, and obtain target information that can comprehensively describe the target merchant. It can also perform feature recognition processing on the newly added target data to be detected that corresponds to the risk detection needs of the target merchant, and obtain the feature information of the target merchant, which can improve data utilization and risk detection efficiency.

[0037] After determining the target data to be detected, the server can perform feature recognition processing on the target data to be detected in a variety of ways. For example, the server can perform keyword recognition processing on the target data to be detected based on preset keywords, and determine the feature information of the target merchant based on the recognized keywords. Specifically, taking the above-mentioned industry conflict detection requirements as an example, the server can perform keyword recognition processing on the key words related to the industry in the target data to be detected, and determine the feature information of the target merchant based on the recognized keywords related to the industry, such as the feature information can be "shopping industry", "catering industry", etc.

[0038] Specifically, for text data, the server can directly perform keyword recognition processing on the text data. For image data, the server can identify the character information contained in the image data based on optical character recognition (OCR) technology, and perform keyword recognition processing on the recognized character information, etc., so as to determine the characteristic information of the target merchant based on the recognized keywords.

[0039] In step S112, the risk detection result for the target merchant is determined according to the target information and the characteristic information.

[0040] In implementation, the server can determine the risk detection result for the target merchant based on the target information and feature information according to the risk detection demand corresponding to the target merchant. For example, taking the risk detection demand corresponding to the target merchant as the industry conflict detection demand as an example, assuming that the target information is "the merchant name of the merchant is "xxx", and according to the merchant name of the merchant, the industry in which the merchant is located is the catering industry, and according to the image data of the merchant, the service provided by the merchant is catering service", if the feature information of the target merchant is "catering industry", then it can be determined that there is no contradiction between the target information and the feature information, that is, the risk detection result for the target merchant can be that there is no risk. If the feature information of the target merchant is "shopping industry", then it can be determined that there is a contradiction between the target information and the feature information, that is, the risk detection result for the target merchant can be that there is a risk.

[0041] The above-mentioned method for determining the risk detection result is an optional and feasible determination method. In addition, there may be a variety of different determination methods. Different determination methods can be selected according to different actual application scenarios. The embodiments of this specification do not make specific limitations on this.

[0042] The embodiments of the present specification provide a data processing method, which receives a risk detection request for a target merchant, obtains data to be detected corresponding to the target merchant in response to the risk detection request, wherein the data to be detected includes multimodal data related to the business operation behavior of the target merchant, performs modal alignment processing on data of different modalities in the data to be detected to obtain target text data corresponding to the data to be detected, performs information fusion processing on the target text data based on a large language model to obtain target information corresponding to the target merchant, obtains target data to be detected in the data to be detected that corresponds to the risk detection requirement of the target merchant, performs feature recognition processing on the target data to be detected to obtain feature information of the target merchant, and determines the risk detection result for the target merchant based on the target information and the feature information. In this way, on the one hand, when the data structure of merchant data becomes more and more complex, the multimodal data contained in the data to be detected can be aligned to the text data through modal alignment processing, and then the aligned data (i.e., target text data) can be subjected to information fusion processing by utilizing the understanding, summarization and reasoning capabilities of the large language model to obtain the target information corresponding to the target merchant. On the other hand, feature recognition processing can be performed on the target data to be detected corresponding to the risk detection needs of the target merchant to obtain feature information that can meet the risk detection needs. In this way, by being able to comprehensively and accurately describe the target information of the target merchant and the feature information that can meet the risk detection needs, the risk detection results for the target merchant can be quickly and accurately determined, thereby improving the detection efficiency and accuracy of risk detection for merchants.

[0043] In practical applications, there are many ways to determine the specific processing of the risk detection result for the target merchant according to the target information and feature information in step S112. An optional processing method is provided below, such as Figure 2 As shown, the processing may specifically include the following steps S1122 to S1124.

[0044] In step S1122, target prompt information corresponding to the risk detection requirements of the target merchant is obtained.

[0045] In actual applications, there may be various specific processing methods for obtaining the target prompt information corresponding to the risk detection requirements of the target merchant in the above step S1122. An optional processing method is provided below, which may specifically include the processing of the following steps A1 to A4.

[0046] In step A1, first prompt information corresponding to the risk detection requirement of the target merchant is obtained.

[0047] During implementation, the server may generate first prompt information corresponding to the risk detection requirements of the target merchant based on a pre-trained generation model, wherein the generation model may be a model constructed based on a preset machine learning algorithm.

[0048] For example, assuming that the risk detection requirement of the target merchant is a merchant name conflict detection requirement, the server can build a generative model based on a neural network algorithm, and train the generative model according to the training data to obtain a trained generative model. The server can input the merchant name conflict detection requirement into the trained generative model to obtain a first prompt information, and the first prompt information may be "Is there a conflict in the merchant name?".

[0049] In step A2, feature extraction processing is performed on the first prompt information to obtain a target vector corresponding to the first prompt information.

[0050] In step A3, a target risk detection rule corresponding to the first prompt information in the risk detection rules is determined according to the similarity between the target vector and the vector corresponding to the risk detection rule.

[0051] In implementation, the server may perform feature extraction processing on the first prompt information according to a pre-trained vector extraction model to obtain a target vector, wherein the vector extraction model may be a model constructed according to a preset machine learning algorithm.

[0052] The server can build a risk knowledge base based on the risk detection rules. The server can select risk detection rules corresponding to the merchant type of the target merchant from the risk knowledge base based on the merchant type of the target merchant, and perform feature extraction processing on the determined risk detection rules through the above-mentioned pre-trained vector extraction model to obtain the vector corresponding to the risk detection rule.

[0053] Alternatively, when there are a large number of risk detection rules, the server can also perform feature extraction processing on the risk detection rules through the above-mentioned pre-trained vector extraction model to obtain vectors corresponding to the risk detection rules. The server can build a risk knowledge base based on the vectors corresponding to the risk detection rules, and directly obtain the vectors corresponding to the risk detection rules corresponding to the target merchant from the risk knowledge base.

[0054] The server may determine the similarity between the target vector and the vector corresponding to each risk detection rule according to a preset similarity algorithm (such as a cosine algorithm, a Euclidean distance algorithm, etc.), and then select a target risk detection rule according to the similarity.

[0055] In step A4, target prompt information is generated according to the target risk detection rule and the first prompt information.

[0056] During implementation, the server can concatenate the target risk detection rule and the first prompt information to obtain the target prompt information. For example, assuming that the first prompt information is "Is there a conflict in the merchant name?" and the target risk detection rule is "If the industry corresponding to the merchant name does not match the identified industry, then there is a conflict in the merchant name", then the target prompt information can be "Is there a conflict in the merchant name? For example, if the industry corresponding to the merchant name does not match the identified industry, then there is a conflict in the merchant name."

[0057] In this way, Figure 3 As shown, by performing retrieval-augmented generation (RAG) on the first prompt information in the above steps A2 to A4, the target risk detection rules related to the first prompt information can be retrieved from the risk knowledge base, and the target prompt information for input into the large language model can be generated according to the target risk detection rules and the first prompt information, which can enhance the ability of the large language model to handle knowledge-intensive tasks (such as question-answering, text summarization, content generation, etc.). And it can help discover new authenticity risk issues based on taking into account the existing risk detection rules and experience, and at the same time, it also reduces the cost of query and analysis and improves the efficiency of strategy operation.

[0058] In step S1124, based on the target prompt information, the target information and the feature information, a risk detection result for the target merchant is determined by a large language model.

[0059] During implementation, the feature information obtained by the feature recognition model with a smaller model structure and the merchant meta-description obtained based on the large language model can be input into the large language model again, and through the target prompt information, the analysis and decision message (i.e., risk detection result) based on the large model's own understanding, summarization and reasoning capabilities can be output.

[0060] The target prompt information may be "Is there any conflict in the information of this merchant?", and the risk detection result may be: "The name of this merchant is XX Restaurant, which is presumed to be in the catering industry. Its transaction time is concentrated between 1 and 6 in the morning, and the amount of payment is distributed between 800 and 10,000 yuan. Judging from the transaction information, it does not conform to the regular payment collection of catering business, and it is presumed to have authenticity risk." In this way, through two-stage data processing (i.e., the first stage of generating target information through modal alignment processing and information fusion processing, and the second stage of determining risk detection results through feature information obtained through feature recognition processing and target information), we only need to focus on maintaining the processing model of the first stage in the future. We can determine the risk detection results corresponding to different risk detection needs based on the merchant meta description (i.e., target information) it produces and the template prompt information, thereby reducing the operation and maintenance costs of the model.

[0061] In practical applications, the data to be detected includes image data, business operation behavior sequence data, table data, and graph structure data. The graph structure data can be constructed based on the relationship between the target merchant and other merchants.

[0062] Among them, the image data may be image data containing the merchant logo of the target merchant, image data of the business license of the target merchant, etc., the business operation behavior sequence data may include the transaction behavior sequence data, operation behavior sequence data, etc. of the target merchant, the table data may include the table data of the transaction feature data and identity feature data of the target merchant, the graph structure data may include graph structure data constructed according to the association relationship between the target merchant and other merchants, the graph structure data constructed according to the association relationship between the devices of the target merchant, the graph structure data constructed according to the resource flow relationship of the target merchant, and the graph structure data constructed according to the legal person relationship corresponding to the target merchant. In addition, the data to be detected may also include text data of the merchant name, address and other information of the target merchant.

[0063] In practical applications, in step S106, modality alignment is performed on the data of different modalities in the data to be detected, and the specific processing methods for obtaining the target text data corresponding to the data to be detected can be various. The following is an optional processing method, such as Figure 2 As shown, the processing may specifically include the following steps S1062 to S1064.

[0064] In step S1062, according to the pre-trained conversion model, the data of different modes in the data to be detected are encoded respectively to obtain feature vectors corresponding to the data to be detected of each mode.

[0065] In step S1064, the feature vector is decoded according to the pre-trained conversion model to obtain the target text data.

[0066] In implementation, Figure 4 As shown, the conversion model may include multiple encoders (such as text data encoder, image data encoder, table data encoder, etc.) and decoders. The server may encode data of different modalities in the data to be detected according to different encoders in the conversion model to obtain feature vectors corresponding to the data to be detected of each modality, and then decode the feature vectors through different decoders in the conversion model to obtain target text data.

[0067] Since the data of different modes in the data to be detected have great differences in information density, for example, the amount of information contained in text data is relatively rich, while the amount of information contained in table data or business operation behavior sequence data is relatively sparse. For example, the name of a certain merchant is "Ruyi BBQ (xx West Road Store)", and it can be inferred from the name that the merchant is in the catering industry and the business form is an offline physical store. The server can only play a certain role in supplementing information by identifying the industry, business form, etc. based on table data or business behavior sequence data (such as transaction flow table data, merchant operation behavior sequence data, etc.).

[0068] Therefore, by converting multiple encoders in the model, data of different modalities can be converted into text descriptions (tokens) to obtain target text data. On the one hand, data of different information densities can be uniformly converted into text forms with higher information density. On the other hand, modal alignment processing can be performed through text as a medium, which is convenient for subsequent multimodal fusion processing.

[0069] For example, taking table data as an example, the target text data obtained by text conversion processing of the target merchant's transaction flow table data within the preset detection period can be: "In the past month, 10% of the merchant's transactions occurred between 4 and 5 o'clock, 15% occurred between 5 and 6 o'clock, 20% occurred between 6 and 7 o'clock, 25% occurred between 7 and 8 o'clock, and the rest occurred after 8 o'clock. The proportion of transactions after 1 o'clock in the middle of the night is less than 1%. The proportion of transactions with a transaction amount of less than 100 is 20%, 100-200 is 30%, 200-300 is 35%, and the proportion of transactions above 500 is less than 2%. The maximum transaction amount is xxx yuan, and the minimum transaction amount is x yuan. In the past week, the number of transactions decreased by 20% compared with the previous week, and the amount decreased by 10%, which is the least in one week in a month." In practical applications, the specific processing methods for performing feature recognition processing on the target data to be detected in step S110 to obtain the feature information of the target merchant can be varied. An optional processing method is provided below, such as Figure 2 As shown, the process may specifically include the following step S1102.

[0070] In step S1102, target data to be detected that corresponds to the risk detection requirements of the target merchant is obtained from the data to be detected, and behavioral feature recognition processing is performed on the target data to be detected based on a pre-trained feature recognition model to obtain feature information of the target merchant.

[0071] Among them, the feature recognition model can be a model built based on a preset machine learning algorithm.

[0072] In practical applications, the feature recognition model may include a first sub-model and a second sub-model. The first sub-model may be used to identify the industry in which the target merchant is located, and the second sub-model may be used to identify whether the transaction behavior of the target merchant is risky. Accordingly, in the above step S1102, the target data to be detected is subjected to behavioral feature recognition processing according to the pre-trained feature recognition model, and the specific processing method for obtaining the feature information of the target merchant may also include the processing of the following steps B1 to B3.

[0073] In step B1, according to the pre-trained first sub-model, based on the image data and business operation behavior sequence data in the target data to be detected, the industry characteristic information of the industry in which the target merchant is located is determined.

[0074] Among them, the first sub-model can be a model constructed based on a preset deep learning algorithm.

[0075] In step B2, according to the pre-trained second sub-model, based on the business operation behavior sequence data and graph structure data, the behavior risk characteristic information of the target merchant is determined.

[0076] In implementation, the second sub-model may be a model constructed based on a preset machine learning algorithm.

[0077] In step B3, characteristic information of the target merchant is determined based on the industry characteristic information and the behavior risk characteristic information.

[0078] In practical applications, there are many ways to determine the specific processing of the risk detection result for the target merchant according to the target information and feature information in step S112. An optional processing method is provided below, such as Figure 2 As shown, the processing may specifically include the following step S1126.

[0079] In step S1126, according to the pre-trained risk detection model, based on the target information and feature information, the risk type corresponding to the target merchant is determined, and according to the risk type corresponding to the target merchant, the risk detection result for the target merchant is determined.

[0080] Among them, the risk detection model can be a model built based on a preset machine learning algorithm, and the risk types can include high risk, medium risk, low risk and other types.

[0081] The embodiments of the present specification provide a data processing method, which receives a risk detection request for a target merchant, obtains data to be detected corresponding to the target merchant in response to the risk detection request, wherein the data to be detected includes multimodal data related to the business operation behavior of the target merchant, performs modal alignment processing on data of different modalities in the data to be detected to obtain target text data corresponding to the data to be detected, performs information fusion processing on the target text data based on a large language model to obtain target information corresponding to the target merchant, obtains target data to be detected in the data to be detected that corresponds to the risk detection requirement of the target merchant, performs feature recognition processing on the target data to be detected to obtain feature information of the target merchant, and determines the risk detection result for the target merchant based on the target information and the feature information. In this way, on the one hand, when the data structure of merchant data becomes more and more complex, the multimodal data contained in the data to be detected can be aligned to the text data through modal alignment processing, and then the aligned data (i.e., target text data) can be subjected to information fusion processing by utilizing the understanding, summarization and reasoning capabilities of the large language model to obtain the target information corresponding to the target merchant. On the other hand, feature recognition processing can be performed on the target data to be detected corresponding to the risk detection needs of the target merchant to obtain feature information that can meet the risk detection needs. In this way, by being able to comprehensively and accurately describe the target information of the target merchant and the feature information that can meet the risk detection needs, the risk detection results for the target merchant can be quickly and accurately determined, thereby improving the detection efficiency and accuracy of risk detection for merchants.

[0082] The above is a data processing method provided in the embodiment of this specification. Based on the same idea, the embodiment of this specification also provides a data processing device, such as Figure 5 shown.

[0083] The data processing device includes: a request receiving module 501, a data acquisition module 502, a modality alignment module 503, an information fusion module 504, a feature recognition module 505 and a risk detection module 506, wherein: A request receiving module 501 is used to receive a risk detection request for a target merchant; A data acquisition module 502 is used to obtain the to-be-detected data corresponding to the target merchant in response to the risk detection request, wherein the to-be-detected data includes multimodal data related to the business operation behavior of the target merchant; A modality alignment module 503 is used to perform modality alignment processing on data of different modalities in the data to be detected to obtain target text data corresponding to the data to be detected; An information fusion module 504 is used to perform information fusion processing on the target text data according to a large language model to obtain target information corresponding to the target merchant; The feature recognition module 505 is used to obtain target data to be detected that corresponds to the risk detection requirements of the target merchant from the data to be detected, and perform feature recognition processing on the target data to be detected to obtain feature information of the target merchant; The risk detection module 506 is used to determine the risk detection result for the target merchant according to the target information and the characteristic information.

[0084] In the embodiment of this specification, the risk detection module 506 is used to: Obtain target prompt information corresponding to the risk detection requirements of the target merchant; Based on the target prompt information, the target information and the feature information, a risk detection result for the target merchant is determined by the large language model.

[0085] In the embodiment of this specification, the risk detection module 506 is used to: Acquire first prompt information corresponding to the risk detection requirement of the target merchant; Performing feature extraction processing on the first prompt information to obtain a target vector corresponding to the first prompt information; Determining a target risk detection rule corresponding to the first prompt information in the risk detection rules according to the similarity between the target vector and the vector corresponding to the risk detection rule; The target prompt information is generated according to the target risk detection rule and the first prompt information.

[0086] In the embodiments of the present specification, the data to be detected includes image data, business operation behavior sequence data, table data, and graph structure data, and the graph structure data is constructed according to the association relationship between the target merchant and other merchants.

[0087] In the embodiment of this specification, the modality alignment module 503 is used to: According to the pre-trained conversion model, encoding processing is performed on the data of different modes in the data to be detected respectively to obtain feature vectors corresponding to the data to be detected of each mode; According to the pre-trained conversion model, the feature vector is decoded to obtain the target text data.

[0088] In the embodiment of this specification, the feature recognition module 505 is used to: According to a pre-trained feature recognition model, the target data to be detected is processed for behavioral feature recognition to obtain feature information of the target merchant, wherein the feature recognition model is a model constructed based on a preset machine learning algorithm.

[0089] In the embodiment of this specification, the feature recognition model includes a first sub-model and a second sub-model. The first sub-model is used to identify the industry of the target merchant, and the second sub-model is used to identify whether the transaction behavior of the target merchant has risks. The feature recognition module 505 is used to: Determine, according to the pre-trained first sub-model, industry characteristic information of the industry in which the target merchant is located based on the image data and the business operation behavior sequence data in the target data to be detected; Determining the behavior risk characteristic information of the target merchant according to the pre-trained second sub-model and based on the business operation behavior sequence data and the graph structure data; Based on the industry characteristic information and the behavior risk characteristic information, characteristic information of the target merchant is determined.

[0090] In the embodiment of this specification, the risk detection module is used to: According to a pre-trained risk detection model, based on the target information and the feature information, the risk type corresponding to the target merchant is determined, and according to the risk type corresponding to the target merchant, the risk detection result for the target merchant is determined, and the risk detection model is a model constructed based on a preset machine learning algorithm.

[0091] The embodiments of the present specification provide a data processing device, which receives a risk detection request for a target merchant, and in response to the risk detection request, obtains data to be detected corresponding to the target merchant, wherein the data to be detected includes multimodal data related to the business operation behavior of the target merchant, performs modal alignment processing on data of different modalities in the data to be detected to obtain target text data corresponding to the data to be detected, performs information fusion processing on the target text data according to a large language model to obtain target information corresponding to the target merchant, obtains target data to be detected in the data to be detected that corresponds to the risk detection requirement of the target merchant, performs feature recognition processing on the target data to be detected to obtain feature information of the target merchant, and determines the risk detection result for the target merchant based on the target information and the feature information. In this way, on the one hand, when the data structure of merchant data becomes more and more complex, the multimodal data contained in the data to be detected can be aligned to the text data through modal alignment processing, and then the aligned data (i.e., target text data) can be subjected to information fusion processing by utilizing the understanding, summarization and reasoning capabilities of the large language model to obtain the target information corresponding to the target merchant. On the other hand, feature recognition processing can be performed on the target data to be detected corresponding to the risk detection needs of the target merchant to obtain feature information that can meet the risk detection needs. In this way, by being able to comprehensively and accurately describe the target information of the target merchant and the feature information that can meet the risk detection needs, the risk detection results for the target merchant can be quickly and accurately determined, thereby improving the detection efficiency and accuracy of risk detection for merchants.

[0092] The above is a data processing device provided in the embodiment of this specification. Based on the same idea, the embodiment of this specification also provides a data processing device, such as Figure 6 shown.

[0093] The data processing device may provide a terminal device or a server, etc. for the above-mentioned embodiments.

[0094] The data processing device may have relatively large differences due to different configurations or performances, and may include one or more processors 601 and memory 602, and the memory 602 may store one or more storage applications or data. Among them, the memory 602 may be a short-term storage or a persistent storage. The application stored in the memory 602 may include one or more modules (not shown in the figure), and each module may include a series of computer executable instructions in the data processing device. Furthermore, the processor 601 may be configured to communicate with the memory 602 and execute a series of computer executable instructions in the memory 602 on the data processing device. The data processing device may also include one or more power supplies 603, one or more wired or wireless network interfaces 604, one or more input and output interfaces 605, and one or more keyboards 606.

[0095] Specifically in this embodiment, the data processing 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 in the data processing device, and the one or more programs are configured to be executed by one or more processors, including computer executable instructions for performing the following: Receive risk detection requests for target merchants; In response to the risk detection request, obtaining the to-be-detected data corresponding to the target merchant, wherein the to-be-detected data includes multimodal data related to the business operation behavior of the target merchant; Performing modality alignment processing on data of different modalities in the data to be detected to obtain target text data corresponding to the data to be detected; According to the large language model, information fusion processing is performed on the target text data to obtain target information corresponding to the target merchant; Acquire target data to be detected that corresponds to the risk detection requirements of the target merchant from the data to be detected, and perform feature recognition processing on the target data to be detected to obtain feature information of the target merchant; A risk detection result for the target merchant is determined according to the target information and the characteristic information.

[0096] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the data processing device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0097] The embodiments of the present specification provide a data processing device, which receives a risk detection request for a target merchant, and in response to the risk detection request, obtains data to be detected corresponding to the target merchant, wherein the data to be detected includes multimodal data related to the business operation behavior of the target merchant, performs modal alignment processing on data of different modalities in the data to be detected to obtain target text data corresponding to the data to be detected, performs information fusion processing on the target text data according to a large language model to obtain target information corresponding to the target merchant, obtains target data to be detected in the data to be detected that corresponds to the risk detection requirement of the target merchant, performs feature recognition processing on the target data to be detected to obtain feature information of the target merchant, and determines the risk detection result for the target merchant based on the target information and the feature information. In this way, on the one hand, when the data structure of merchant data becomes more and more complex, the multimodal data contained in the data to be detected can be aligned to the text data through modal alignment processing, and then the aligned data (i.e., target text data) can be subjected to information fusion processing by utilizing the understanding, summarization and reasoning capabilities of the large language model to obtain the target information corresponding to the target merchant. On the other hand, feature recognition processing can be performed on the target data to be detected corresponding to the risk detection needs of the target merchant to obtain feature information that can meet the risk detection needs. In this way, by being able to comprehensively and accurately describe the target information of the target merchant and the feature information that can meet the risk detection needs, the risk detection results for the target merchant can be quickly and accurately determined, thereby improving the detection efficiency and accuracy of risk detection for merchants.

[0098] Furthermore, based on the above Figures 1 to 5 In one embodiment, the present specification further provides a storage medium for storing computer executable instruction information. In a specific embodiment, the storage medium may be a USB flash drive, an optical disk, a hard disk, etc. When the computer executable instruction information stored in the storage medium is executed by the processor, the following process can be implemented: Receive risk detection requests for target merchants; In response to the risk detection request, obtaining the to-be-detected data corresponding to the target merchant, wherein the to-be-detected data includes multimodal data related to the business operation behavior of the target merchant; Performing modality alignment processing on data of different modalities in the data to be detected to obtain target text data corresponding to the data to be detected; According to the large language model, information fusion processing is performed on the target text data to obtain target information corresponding to the target merchant; Acquire target data to be detected that corresponds to the risk detection requirements of the target merchant from the data to be detected, and perform feature recognition processing on the target data to be detected to obtain feature information of the target merchant; A risk detection result for the target merchant is determined according to the target information and the characteristic information.

[0099] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the above-mentioned storage medium embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0100] The embodiments of the present specification provide a storage medium, which receives a risk detection request for a target merchant, obtains data to be detected corresponding to the target merchant in response to the risk detection request, wherein the data to be detected includes multimodal data related to the business operation behavior of the target merchant, performs modal alignment processing on data of different modalities in the data to be detected to obtain target text data corresponding to the data to be detected, performs information fusion processing on the target text data according to a large language model to obtain target information corresponding to the target merchant, obtains target data to be detected in the data to be detected that corresponds to the risk detection requirement of the target merchant, performs feature recognition processing on the target data to be detected to obtain feature information of the target merchant, and determines the risk detection result for the target merchant based on the target information and the feature information. In this way, on the one hand, when the data structure of merchant data becomes more and more complex, the multimodal data contained in the data to be detected can be aligned to the text data through modal alignment processing, and then the aligned data (i.e., target text data) can be subjected to information fusion processing by utilizing the understanding, summarization and reasoning capabilities of the large language model to obtain the target information corresponding to the target merchant. On the other hand, feature recognition processing can be performed on the target data to be detected corresponding to the risk detection needs of the target merchant to obtain feature information that can meet the risk detection needs. In this way, by being able to comprehensively and accurately describe the target information of the target merchant and the feature information that can meet the risk detection needs, the risk detection results for the target merchant can be quickly and accurately determined, thereby improving the detection efficiency and accuracy of risk detection for merchants.

[0101] Furthermore, based on the above Figures 1 to 5 In one or more embodiments of the present specification, a computer program product is provided, including a computer program. When the computer program in the computer program product is executed by a processor, the following process can be implemented: Receive risk detection requests for target merchants; In response to the risk detection request, obtaining the to-be-detected data corresponding to the target merchant, wherein the to-be-detected data includes multimodal data related to the business operation behavior of the target merchant; Performing modality alignment processing on data of different modalities in the data to be detected to obtain target text data corresponding to the data to be detected; According to the large language model, information fusion processing is performed on the target text data to obtain target information corresponding to the target merchant; Acquire target data to be detected that corresponds to the risk detection requirements of the target merchant from the data to be detected, and perform feature recognition processing on the target data to be detected to obtain feature information of the target merchant; A risk detection result for the target merchant is determined according to the target information and the characteristic information.

[0102] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the above-mentioned computer program product embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0103] The embodiments of the present specification provide a computer program product, which receives a risk detection request for a target merchant, obtains data to be detected corresponding to the target merchant in response to the risk detection request, wherein the data to be detected includes multimodal data related to the business operation behavior of the target merchant, performs modal alignment processing on data of different modalities in the data to be detected to obtain target text data corresponding to the data to be detected, performs information fusion processing on the target text data based on a large language model to obtain target information corresponding to the target merchant, obtains target data to be detected in the data to be detected that corresponds to the risk detection requirement of the target merchant, performs feature recognition processing on the target data to be detected to obtain feature information of the target merchant, and determines the risk detection result for the target merchant based on the target information and the feature information. In this way, on the one hand, when the data structure of merchant data becomes more and more complex, the multimodal data contained in the data to be detected can be aligned to the text data through modal alignment processing, and then the aligned data (i.e., target text data) can be subjected to information fusion processing by utilizing the understanding, summarization and reasoning capabilities of the large language model to obtain the target information corresponding to the target merchant. On the other hand, feature recognition processing can be performed on the target data to be detected corresponding to the risk detection needs of the target merchant to obtain feature information that can meet the risk detection needs. In this way, by being able to comprehensively and accurately describe the target information of the target merchant and the feature information that can meet the risk detection needs, the risk detection results for the target merchant can be quickly and accurately determined, thereby improving the detection efficiency and accuracy of risk detection for merchants.

[0104] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0105] In the 1990s, it was very clear whether the improvement of a technology was hardware improvement (for example, improvement of the circuit structure of diodes, transistors, switches, etc.) or software improvement (improvement of the method flow). However, with the development of technology, many improvements of the method flow today can be regarded as direct improvements of the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved method flow into the hardware circuit. Therefore, it cannot be said that the improvement of a method flow cannot be implemented with a hardware entity module. For example, a programmable logic device (PLD) (such as a field programmable gate array (FPGA)) is such an integrated circuit whose logical function is determined by the user's programming of the device. Designers can "integrate" a digital system on a PLD by programming themselves, without having to ask chip manufacturers to design and make dedicated integrated circuit chips. Moreover, nowadays, instead of manually making integrated circuit chips, this kind of programming is mostly implemented by "logic compiler" software, which is similar to the software compiler used when developing and writing programs, and the original code before compilation must also be written in a specific programming language, which is called hardware description language (HDL). There is not only one kind of HDL, but many kinds, 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, RHDL (Ruby Hardware Description Language), etc. The most commonly used ones are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also know that it is only necessary to program the method flow slightly in the above-mentioned hardware description languages ​​and program it into the integrated circuit, and then it is easy to obtain the hardware circuit that implements the logic method flow.

[0106] The controller may be implemented in any suitable manner, for example, the controller may take the form of a microprocessor or processor and a computer-readable medium storing a computer-readable program code (e.g., software or firmware) executable by the (micro)processor, a logic gate, a switch, an application-specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller, examples of which include but are not limited to the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320, and the memory controller may also be implemented as part of the control logic of the memory. It is also known to those skilled in the art that, in addition to implementing the controller in a purely computer-readable program code manner, the controller may be implemented in the form of a logic gate, a switch, an application-specific integrated circuit, a programmable logic controller, and an embedded microcontroller by logically programming the method steps. Therefore, such a controller may be considered as a hardware component, and the devices for implementing various functions included therein may also be considered as structures within the hardware component. Or even, the devices for implementing various functions may be considered as both software modules for implementing the method and structures within the hardware component.

[0107] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0108] For the convenience of description, the above devices are described in terms of functions and are divided into various units. Of course, when implementing one or more embodiments of this specification, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0109] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems, or computer program products. Therefore, one or more embodiments of this specification may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, one or more embodiments of this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0110] The embodiments of this specification are described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of this specification. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable fraud case serial and parallel device to produce a machine, so that the instructions executed by the processor of the computer or other programmable fraud case serial and parallel device generate instructions for implementing the processes in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0111] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable fraud case serial and parallel device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0112] These computer program instructions may also be loaded onto a computer or other programmable device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0113] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0114] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0115] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules 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 technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined in this article, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0116] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0117] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems or computer program products. Therefore, one or more embodiments of this specification may be in the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Furthermore, one or more embodiments of this specification may be in the form of a computer program product implemented 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.

[0118] One or more embodiments of the present specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. One or more embodiments of the present specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0119] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0120] The above description is only an embodiment of this specification and is not intended to limit this document. For those skilled in the art, this specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this specification should be included in the scope of the claims of this specification.

Claims

1. A data processing method, comprising: Receive risk detection requests for target merchants; In response to the risk detection request, obtaining the to-be-detected data corresponding to the target merchant, wherein the to-be-detected data includes multimodal data related to the business operation behavior of the target merchant; Performing modality alignment processing on data of different modalities in the data to be detected to obtain target text data corresponding to the data to be detected; According to the large language model, information fusion processing is performed on the target text data to obtain target information corresponding to the target merchant; Acquire target data to be detected that corresponds to the risk detection requirements of the target merchant from the data to be detected, and perform feature recognition processing on the target data to be detected to obtain feature information of the target merchant; A risk detection result for the target merchant is determined according to the target information and the characteristic information.

2. The method according to claim 1, wherein determining the risk detection result for the target merchant according to the target information and the feature information comprises: Obtaining target prompt information corresponding to the risk detection requirements of the target merchant; Based on the target prompt information, the target information and the feature information, a risk detection result for the target merchant is determined by the large language model.

3. According to the method of claim 2, the step of obtaining target prompt information corresponding to the risk detection requirements of the target merchant comprises: Acquire first prompt information corresponding to the risk detection requirement of the target merchant; Performing feature extraction processing on the first prompt information to obtain a target vector corresponding to the first prompt information; Determining a target risk detection rule corresponding to the first prompt information in the risk detection rules according to the similarity between the target vector and the vector corresponding to the risk detection rule; The target prompt information is generated according to the target risk detection rule and the first prompt information.

4. According to the method of claim 3, the data to be detected includes image data, business operation behavior sequence data, table data, and graph structure data, and the graph structure data is constructed according to the association relationship between the target merchant and other merchants.

5. According to the method of claim 4, the step of performing modality alignment processing on the data of different modalities in the data to be detected to obtain target text data corresponding to the data to be detected comprises: According to the pre-trained conversion model, encoding processing is performed on the data of different modes in the data to be detected respectively to obtain feature vectors corresponding to the data to be detected of each mode; According to the pre-trained conversion model, the feature vector is decoded to obtain the target text data.

6. The method according to claim 4, wherein the step of performing feature recognition processing on the target data to be detected to obtain feature information of the target merchant comprises: According to a pre-trained feature recognition model, the target data to be detected is processed for behavioral feature recognition to obtain feature information of the target merchant, wherein the feature recognition model is a model constructed based on a preset machine learning algorithm.

7. The method according to claim 6, wherein the feature recognition model comprises a first sub-model and a second sub-model, wherein the first sub-model is used to identify the industry in which the target merchant is located, and the second sub-model is used to identify whether the transaction behavior of the target merchant is risky, and the target data to be detected is subjected to behavioral feature recognition processing according to the pre-trained feature recognition model to obtain feature information of the target merchant, including: Determine, according to the pre-trained first sub-model, industry characteristic information of the industry in which the target merchant is located based on the image data and the business operation behavior sequence data in the target data to be detected; Determining the behavior risk characteristic information of the target merchant according to the pre-trained second sub-model and based on the business operation behavior sequence data and the graph structure data; Based on the industry characteristic information and the behavior risk characteristic information, characteristic information of the target merchant is determined.

8. The method according to claim 1, determining the risk detection result for the target merchant according to the target information and the characteristic information, comprising: According to a pre-trained risk detection model, based on the target information and the feature information, the risk type corresponding to the target merchant is determined, and according to the risk type corresponding to the target merchant, the risk detection result for the target merchant is determined, and the risk detection model is a model constructed based on a preset machine learning algorithm.

9. A data processing device, comprising: A request receiving module, used to receive a risk detection request for a target merchant; A data acquisition module, configured to acquire, in response to the risk detection request, the to-be-detected data corresponding to the target merchant, wherein the to-be-detected data includes multimodal data related to the business operation behavior of the target merchant; A modality alignment module, used for performing modality alignment processing on data of different modalities in the data to be detected, so as to obtain target text data corresponding to the data to be detected; An information fusion module, used to perform information fusion processing on the target text data according to a large language model to obtain target information corresponding to the target merchant; A feature recognition module is used to obtain target data to be detected that corresponds to the risk detection requirements of the target merchant from the data to be detected, and perform feature recognition processing on the target data to be detected to obtain feature information of the target merchant; The risk detection module is used to determine the risk detection result for the target merchant according to the target information and the characteristic information.

10. A data processing device, comprising: processor; as well as a memory arranged to store computer executable instructions which, when executed, cause the processor to: Receive risk detection requests for target merchants; In response to the risk detection request, obtaining the to-be-detected data corresponding to the target merchant, wherein the to-be-detected data includes multimodal data related to the business operation behavior of the target merchant; Performing modality alignment processing on data of different modalities in the data to be detected to obtain target text data corresponding to the data to be detected; According to the large language model, information fusion processing is performed on the target text data to obtain target information corresponding to the target merchant; Acquire target data to be detected that corresponds to the risk detection requirements of the target merchant from the data to be detected, and perform feature recognition processing on the target data to be detected to obtain feature information of the target merchant; A risk detection result for the target merchant is determined according to the target information and the characteristic information.

Images