Construction method and device of recognition model, recognition method and device, equipment, storage medium and program product

By building an identification model based on sample data and preset risk indicators, and performing dimensionality reduction processing and feature extraction, the problem of low identification accuracy in telecommunications networks is solved, and efficient identification and prevention of telecommunications fraud behavior is achieved.

CN118917651APending Publication Date: 2024-11-08DAWNING CLOUD COMPUTING TECH CO LTD
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202410840816.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-06-26
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The existing telecommunications networks have the problem of low recognition accuracy, making it difficult to effectively identify telecommunications fraud.

Method used

By obtaining sample data and preset risk indicators of various types of businesses, building a target recognition model and verifying its accuracy, using dimensionality reduction processing and feature extraction to improve the training efficiency and accuracy of the recognition model, and formulating identification rules to improve the model identification accuracy.

Benefits of technology

It significantly improves the accuracy of telecom fraud identification, ensures the accuracy and efficiency of the identification model, and effectively prevents telecom fraud.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118917651A_ABST
    Figure CN118917651A_ABST
Patent Text Reader

Abstract

The invention relates to a construction method and device of an identification model, an identification method and device, equipment, a storage medium and a program product. The method comprises the steps of obtaining sample data and a preset risk index of each type of business, training an initial recognition model according to the sample data and the preset risk index of each type of business, constructing a target recognition model, and performing accuracy verification on the target recognition model to obtain a verified target recognition model. According to the method, the initial recognition model is trained to obtain the target recognition model according to the sample data and the preset risk index, and compared with an existing method for constructing a recognition model only based on the sample data, the target recognition model constructed by the method can recognize risks more accurately due to the fact that the preset risk index is considered.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method for constructing a recognition model, a recognition method, a device, a equipment, a storage medium and a program product. Background Art

[0002] With the continuous development of electronic information and network communication technology, telecommunication technology has brought people new functional experiences such as convenient communication, shopping, social interaction, and games. However, with the continuous development and change of telecommunication technology, telecommunication fraud, telecommunication inducement and other bad behaviors have also appeared in the telecommunication network, bringing great network risks to telecommunication network users.

[0003] In the prior art, the telecommunication network platform generally collects the telephone number of the user terminal during the call and identifies the telephone number to determine whether the telephone number is an illegal telephone number. If the telephone number is determined to be an illegal telephone number, the user terminal is reminded that the current call partner is an illegal person, so that the user terminal performs a telecommunication anti-explosion protection operation to improve the telecommunication call security of the user terminal.

[0004] However, the above identification method based on telephone number has the problem of low identification accuracy. Summary of the invention

[0005] Based on this, it is necessary to provide a method for constructing an identification model, an identification method, an apparatus, a device, a storage medium and a program product that can improve the identification accuracy of telecommunications fraud in response to the above-mentioned technical problems.

[0006] In a first aspect, the present application provides a method for constructing a recognition model, comprising:

[0007] Obtain sample data and preset risk indicators for various types of businesses;

[0008] Based on the sample data of various types of businesses and preset risk indicators, the initial recognition model is trained to build a target recognition model;

[0009] The target recognition model is verified for accuracy to obtain a verified target recognition model.

[0010] The above method trains the initial recognition model according to the sample data and the preset risk indicators at the same time to obtain the target recognition model. Compared with the existing method of constructing the recognition model only based on the sample data, the target recognition model constructed in the present application can identify the risks more accurately because it takes into account the preset risk indicators. In addition, after the target recognition model is constructed, the accuracy of the target recognition model is also verified to further ensure the recognition accuracy of the target recognition model.

[0011] In one embodiment, the initial recognition model is trained based on the sample data of each type of business and the preset risk indicators to construct the target recognition model, including:

[0012] According to the preset high-low dimensional conversion relationship, the sample data of each type of business is subjected to dimensionality reduction processing to obtain sample data after dimensionality reduction processing; dimensionality reduction processing refers to using the sample data after dimensionality reduction processing to simulate the similarity between the sample data before dimensionality reduction processing;

[0013] Based on the sample data after dimensionality reduction and the preset risk indicators, the initial recognition model is trained to build a target recognition model.

[0014] The method provided in the embodiment of the present application performs dimensionality reduction processing on sample data, so that the sample data after dimensionality reduction processing can reduce the convergence difficulty of the recognition model during the training process and improve the training efficiency of the recognition model.

[0015] In one embodiment, the above-mentioned sample data of each type of business is subjected to dimensionality reduction processing according to the preset high-low dimensional conversion relationship to obtain the sample data after dimensionality reduction processing, including:

[0016] According to the preset high-low dimensional conversion relationship, the sample data of each type of business is processed with dimension reduction to obtain intermediate processed sample data;

[0017] The sample data after dimensionality reduction processing is obtained according to the first similarity between any two sample data in the sample data of each type of business and the second similarity between any two sample data in the intermediate processed sample data.

[0018] The dimensionality reduction method provided in the embodiment of the present application obtains sample data after dimensionality reduction based on the similarity between data in high-dimensional space and the similarity between data in low-dimensional space, thereby providing a data basis for subsequent training of a recognition model based on the dimensionality reduction data.

[0019] In one embodiment, the sample data after dimensionality reduction processing is obtained according to the first similarity between any two sample data in the sample data of each type of business and the second similarity between any two sample data in the intermediate processed sample data, including:

[0020] Determine a first similarity between any two sample data in the sample data of each type of business;

[0021] Determine a second similarity between any two sample data in the intermediate processed sample data;

[0022] Based on the principle of identical similarity, the preset high-low dimensional conversion relationship is adjusted according to the first similarity and the second similarity, and the sample data of each type of business is subjected to dimensionality reduction processing based on the adjusted preset high-low dimensional conversion relationship to obtain the sample data after dimensionality reduction processing.

[0023] The dimensionality reduction method provided in the embodiment of the present application obtains sample data after dimensionality reduction based on the similarity between data in high-dimensional space and the similarity between data in low-dimensional space, thereby providing a data basis for subsequent training of a recognition model based on the dimensionality reduction data.

[0024] In one embodiment, the preset high-low dimensional conversion relationship is adjusted according to the first similarity and the second similarity, and the sample data of each type of business is subjected to dimensionality reduction processing based on the adjusted preset high-low dimensional conversion relationship to obtain the sample data after dimensionality reduction processing, including:

[0025] Determining whether the first similarity and the second similarity are consistent;

[0026] If the first similarity and the second similarity are consistent, the intermediate processed sample data is used as the sample data after dimensionality reduction processing;

[0027] If the first similarity and the second similarity are inconsistent, the preset high-low dimensional conversion relationship is adjusted, and based on the adjusted preset high-low dimensional conversion relationship, the step of returning to execute dimensionality reduction processing on sample data of each type of business according to the preset high-low dimensional conversion relationship to obtain intermediate processed sample data is performed.

[0028] The dimensionality reduction method provided in the embodiment of the present application obtains sample data after dimensionality reduction based on the similarity between data in high-dimensional space and the similarity between data in low-dimensional space, thereby providing a data basis for subsequent training of a recognition model based on the dimensionality reduction data.

[0029] In one embodiment, the method further comprises:

[0030] Extract features from sample data of various types of business to obtain data features of the sample data; data features include key features, illegal behavior process features, object behavior features, and illegal resource transfer features;

[0031] According to the preset high-low dimensional conversion relationship, the sample data of each type of business is subjected to dimensionality reduction processing to obtain the sample data after dimensionality reduction processing, including:

[0032] According to the preset high-low dimensional conversion relationship, the data features of the sample data are processed by dimensionality reduction to obtain the sample data after dimensionality reduction.

[0033] The data feature extraction method provided in the embodiment of the present application ensures that the data points of each sample data are close to each other before the data is input into the initial recognition model training, which facilitates the movement of each clustering center and improves the training efficiency of the initial recognition model.

[0034] In one embodiment, the initial recognition model is trained based on the sample data after dimensionality reduction processing and the preset risk indicators to construct the target recognition model, including:

[0035] Formulate identification rules based on preset risk indicators; the identification rules include the risk level corresponding to each indicator range in the preset risk indicators;

[0036] The sample data and recognition rules after dimensionality reduction processing are input into the initial recognition model, so that the initial recognition model is trained based on the recognition rules to build a target recognition model.

[0037] The method for constructing a target recognition model provided in an embodiment of the present application first formulates recognition rules according to preset risk indicators, and then trains the initial recognition model based on the recognition rules and sample data after dimensionality reduction processing, which greatly improves the training efficiency of the model.

[0038] In one embodiment, the accuracy verification of the target recognition model to obtain the verified target recognition model includes:

[0039] Input the verification sample data into the target recognition model for recognition, and obtain the risk value corresponding to the verification sample data;

[0040] By statistically verifying the risk value corresponding to the sample data, the indicator hit rate of the target recognition model is determined; the indicator hit rate includes the indicator false alarm rate and user hit rate of the target recognition model;

[0041] According to the indicator hit rate of the target recognition model, the verified target recognition model is obtained.

[0042] The target recognition model verification method provided in the embodiment of the present application ensures the recognition accuracy of the target recognition model.

[0043] In one embodiment, the method further comprises:

[0044] Standardizing the sample data of each type of business to obtain standardized sample data; the standardization process includes at least one of data cleaning process, data conversion process, data normalization process and data verification process;

[0045] According to the sample data of various types of businesses and the preset risk indicators, the initial recognition model is trained to build the target recognition model, including:

[0046] Based on the standardized sample data and preset risk indicators, the initial recognition model is trained to build a target recognition model.

[0047] The standardized processing method provided in the embodiment of the present application ensures the accuracy and standardization of information, provides a data basis for subsequent training of recognition models based on standardized sample data, and further improves the accuracy of the trained model.

[0048] In a second aspect, the present application provides an identification method, comprising:

[0049] Obtaining business interaction information of the user to be identified;

[0050] Extracting features from the business interaction information to obtain business features corresponding to the business interaction information;

[0051] The risk of illegal behavior is identified based on business characteristics and target recognition models to obtain the risk value corresponding to the business interaction information; the target recognition model is obtained by training the initial recognition model based on sample data of various types of business and preset risk indicators, and its accuracy is verified.

[0052] The identification method provided in the embodiment of the present application realizes effective identification of risky behaviors in business interaction information by extracting business features in business interaction information and identifying risk values ​​corresponding to the business interaction information based on an identification model according to the business features.

[0053] In a third aspect, the present application also provides a device for constructing a recognition model, comprising:

[0054] The acquisition module is used to obtain sample data and preset risk indicators of various types of businesses;

[0055] The training module is used to train the initial recognition model and build a target recognition model based on sample data of various types of businesses and preset risk indicators;

[0056] The verification module is used to verify the accuracy of the target recognition model to obtain a verified target recognition model.

[0057] In a fourth aspect, the present application further provides an identification device, comprising:

[0058] An acquisition module is used to acquire the business interaction information of the user to be identified;

[0059] An extraction module is used to extract features from the business interaction information to obtain business features corresponding to the business interaction information;

[0060] The identification module is used to identify the risks of illegal behaviors based on business characteristics and target identification models, and obtain the risk value corresponding to the business interaction information; the target identification model is obtained by training the initial identification model based on sample data of various types of businesses and preset risk indicators, and its accuracy is verified.

[0061] In a fifth aspect, the present application further provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0062] Obtain sample data and preset risk indicators for various types of businesses;

[0063] Based on the sample data of various types of businesses and preset risk indicators, the initial recognition model is trained to build a target recognition model;

[0064] The target recognition model is verified for accuracy to obtain a verified target recognition model.

[0065] In a sixth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0066] Obtain sample data and preset risk indicators for various types of businesses;

[0067] Based on the sample data of various types of businesses and preset risk indicators, the initial recognition model is trained to build a target recognition model;

[0068] The target recognition model is verified for accuracy to obtain a verified target recognition model.

[0069] In a seventh aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0070] Obtain sample data and preset risk indicators for various types of businesses;

[0071] Based on the sample data of various types of businesses and preset risk indicators, the initial recognition model is trained to build a target recognition model;

[0072] The target recognition model is verified for accuracy to obtain a verified target recognition model.

[0073] The above-mentioned recognition model construction method, recognition method, device, equipment, storage medium and program product, the above-mentioned method trains the initial recognition model according to sample data and preset risk indicators at the same time to obtain a target recognition model. Compared with the existing method of constructing a recognition model based only on sample data, the target recognition model constructed in the present application takes into account the preset risk indicators and can identify risks more accurately; in addition, after the target recognition model is constructed, the accuracy of the target recognition model is also verified to further ensure the recognition accuracy of the target recognition model. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0075] FIG1 is an application environment diagram of a method for constructing a recognition model in one embodiment;

[0076] Figure 2 A schematic diagram of a process for constructing a recognition model in an embodiment;

[0077] Figure 3 A schematic diagram of a flow chart of a method for constructing a recognition model in another embodiment;

[0078] Figure 4 A schematic diagram of a flow chart of a method for constructing a recognition model in another embodiment;

[0079] Figure 5 A schematic diagram of a flow chart of a method for constructing a recognition model in another embodiment;

[0080] Figure 6 A schematic diagram of a flow chart of a method for constructing a recognition model in another embodiment;

[0081] Figure 7 A schematic diagram of a flow chart of a method for constructing a recognition model in another embodiment;

[0082] Figure 8 A schematic diagram of a flow chart of a method for constructing a recognition model in another embodiment;

[0083] Fig. 9 A schematic diagram of a flow chart of a method for constructing a recognition model in another embodiment;

[0084] Fig.10 A schematic diagram of a flow chart of a method for constructing a recognition model in another embodiment;

[0085] Fig.11 is a schematic flow chart of an identification method in an embodiment;

[0086] Fig.12 A schematic diagram of a flow chart of a method for constructing a recognition model in another embodiment;

[0087] FIG13 is a structural block diagram of a device for constructing a recognition model in one embodiment;

[0088] FIG. 14 is a structural block diagram of a device for constructing a recognition model in one embodiment. DETAILED DESCRIPTION

[0089] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0090] With the continuous development of electronic information and network communication technology, telecommunication technology has brought people new functional experiences such as convenient communication, shopping, social interaction, and games. However, with the continuous development and change of telecommunication technology, telecommunication fraud, telecommunication inducement and other bad behaviors have also appeared in the telecommunication network, bringing great network risks to telecommunication network users.

[0091] In the prior art, the telecommunication network platform generally collects the telephone number of the user terminal during the call and identifies the telephone number to determine whether the telephone number is an illegal telephone number. If the telephone number is determined to be an illegal telephone number, the user terminal is reminded that the current call object is an illegal person, so that the user terminal performs a telecommunication anti-explosion protection operation to improve the telecommunication call security of the user terminal. However, the above identification method based on the telephone number has the problem of low identification accuracy. The present application aims to solve this problem.

[0092] After the above introduction of the background technology of the construction method of the recognition model provided by the embodiment of the present application, the implementation environment involved in the construction method of the recognition model provided by the embodiment of the present application will be briefly described below. The construction method of the recognition model provided by the embodiment of the present application can be applied to a computer device as shown in Figure 1. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store sample data of various types of business. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a construction method of a recognition model is implemented.

[0093] Those skilled in the art will understand that the structure shown in FIG. 1 is merely a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. A specific terminal may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0094] After introducing the application scenarios of the method for constructing the recognition model provided in the embodiments of the present application, the following focuses on introducing the method for constructing the recognition model described in the present application.

[0095] In one embodiment, Figure 2 As shown, a method for constructing a recognition model is provided, and the method is applied to the computer device in FIG. 1 as an example for explanation, including the following steps:

[0096] S201. Obtain sample data and preset risk indicators for various types of businesses.

[0097] Among them, the sample data of various types of businesses refers to the sample data obtained from various registration platforms of transaction users, including the basic information, asset information, account information and transaction information of transaction users. The basic information includes the basic information, family information, occupational information, and residential address information of transaction users. The asset information includes asset status information. The account information includes account transaction information, account login information, and transaction account status information. The transaction information includes transaction amount information, transaction frequency information, transaction payment and receipt object information, abnormal behavior information, etc.

[0098] Among them, the preset risk indicator refers to the risk range corresponding to each type of business determined based on the sample data of each type of business. For example, for the transaction amount information in the sample data, setting the transaction amount between 0-10,000 yuan is low risk, setting the transaction amount between 1-100,000 yuan is medium risk, and setting the transaction amount greater than 100,000 yuan is high risk.

[0099] In the embodiment of the present application, when it is necessary to identify the risks in the transaction process, an identification model can be constructed to realize that by inputting the business data generated in the transaction process into the constructed identification model, it is possible to obtain whether the business data has risks, and in the case of risks, promptly dissuade business transaction personnel from stopping transactions to avoid causing property losses to the transaction personnel. Then, before constructing the identification model, sample data of various types of business can be obtained from various types of business systems (for example, banking systems, transaction systems, payment systems, etc.), and the preset risk indicators of various types of business can be determined based on the sample data of various types of business; or, sample data of various types of business can be obtained from other devices using an interface (for example, a USB interface), and the preset risk indicators of various types of business can be determined based on the sample data of various types of business.

[0100] S202: Based on the sample data of each type of business and the preset risk indicators, the initial recognition model is trained to construct a target recognition model.

[0101] The initial recognition model refers to a Gaussian distribution probability model.

[0102] In the embodiment of the present application, after the sample data and preset risk indicators of each type of business are obtained as described above, the sample data and preset risk indicators of each type of business can be input into the initial recognition model, and the initial recognition model can be trained to obtain the target recognition model. Optionally, the sample data of each type of business can be labeled based on the preset risk indicators, and the labels include: no risk, low risk, medium risk and high risk, etc., and then the sample data of each type of business with the labels added can be input into the initial recognition model, and the initial recognition model can be trained to obtain the trained initial recognition model, that is, the target recognition model.

[0103] S203: Verify the accuracy of the target recognition model to obtain a verified target recognition model.

[0104] In the embodiment of the present application, after the target recognition model is obtained as described above, it is also necessary to verify whether the target recognition model can accurately identify the risks in the business data. Therefore, it is also necessary to verify the accuracy of the constructed target recognition model. The embodiment of the present application provides a method for verifying the accuracy of the target recognition model. After the target recognition model is constructed as described above, the target recognition model can be verified for accuracy to obtain a verified target recognition model. For example, the data to be recognized is input into the target recognition model to obtain a recognition result, and the data to be recognized is compared with a preset risk indicator to determine the comparison result, and then the similarity between the recognition result and the comparison result is determined. If the similarity is high, it is considered that the recognition accuracy of the target recognition model constructed in S202 is high. If the similarity is low, it is considered that the recognition accuracy of the target recognition model constructed in S202 is low. At this time, it is necessary to adjust the parameters of the target recognition model to make its prediction of the data more accurate.

[0105] The construction of the recognition model provided in the embodiment of the present application obtains sample data and preset risk indicators of various types of businesses, trains the initial recognition model according to the sample data and preset risk indicators of various types of businesses, constructs the target recognition model, verifies the accuracy of the target recognition model, and obtains the verified target recognition model. The above method trains the initial recognition model according to the sample data and the preset risk indicators to obtain the target recognition model. Compared with the existing method of constructing the recognition model based only on sample data, the target recognition model constructed by the present application can more accurately identify risks because it takes into account the preset risk indicators; in addition, after the target recognition model is constructed, the accuracy of the target recognition model is also verified, which further ensures the recognition accuracy of the target recognition model.

[0106] In one embodiment, Figure 2 Based on the embodiment shown, the process of constructing the target recognition model can be described as follows: Figure 3 As shown, the above S202 "training the initial recognition model and building the target recognition model according to the sample data of each type of business and the preset risk indicators" includes:

[0107] S301. Perform dimensionality reduction processing on sample data of each type of business according to a preset high-to-low dimensional conversion relationship to obtain sample data after dimensionality reduction processing.

[0108] The preset high- and low-dimensional conversion relationship refers to a mapping relationship used when converting data from a high-dimensional space to a low-dimensional space. The preset high- and low-dimensional conversion relationship may be a t-SNE (T-distributed Stochastic NeighborEmbedding) algorithm.

[0109] The dimensionality reduction process refers to using the sample data after the dimensionality reduction process to simulate the similarity between the sample data before the dimensionality reduction process.

[0110] In the embodiment of the present application, since the sample data of each type of business obtained are all high-dimensional data, when constructing the recognition model, the high-dimensional data will cause the recognition model to have greater difficulty in converging during the training process. Therefore, it is necessary to convert the high-dimensional data into low-dimensional data to reduce the difficulty of converging the recognition model during the training process. After the sample data of each type of business is obtained as described above, the preset high-to-low dimensional conversion relationship can be used to perform dimensionality reduction processing on the sample data of each type of business in the high-dimensional space to obtain the sample data after dimensionality reduction processing. It should be noted that the sample data of each type of business in the high-dimensional space is in a one-to-one correspondence with the sample data in the low-dimensional space after dimensionality reduction processing, that is, a sample data X in the high-dimensional space is i , that is, there is a corresponding sample data Y in the low-dimensional space i .

[0111] S302: Based on the sample data after dimensionality reduction processing and the preset risk indicators, the initial recognition model is trained to construct a target recognition model.

[0112] In the embodiment of the present application, after the sample data after the dimensionality reduction processing is obtained, the initial recognition model can be trained according to the sample data after the dimensionality reduction processing and the preset risk index to construct the target recognition model. Optionally, the sample data after the dimensionality reduction processing can be first input into the target recognition model to obtain the recognition result, and the sample data after the dimensionality reduction processing can be compared with the preset risk index to determine the comparison result, and then the similarity between the recognition result and the comparison result is determined. If the similarity is high, it is considered that the recognition accuracy of the constructed target recognition model is high. If the similarity is low, it is considered that the recognition accuracy of the constructed target recognition model is low. At this time, it is necessary to adjust the parameters of the target recognition model to make its prediction of the data more accurate.

[0113] The method provided in the embodiment of the present application performs dimensionality reduction processing on sample data, so that the sample data after dimensionality reduction processing can reduce the convergence difficulty of the recognition model during the training process and improve the training efficiency of the recognition model.

[0114] In one embodiment, Figure 3 Based on the embodiment shown, the process of obtaining sample data after dimensionality reduction processing can be described as follows: Figure 4 As shown, the above S301 "perform dimensionality reduction processing on sample data of each type of business according to the preset high-low dimensional conversion relationship to obtain sample data after dimensionality reduction processing" includes:

[0115] S401. Perform dimensionality reduction processing on sample data of each type of business according to a preset high-to-low dimensional conversion relationship to obtain intermediate processed sample data.

[0116] The preset high- and low-dimensional conversion relationship refers to a mapping relationship used when converting data from a high-dimensional space to a low-dimensional space. The preset high- and low-dimensional conversion relationship may be a t-SNE (T-distributed Stochastic NeighborEmbedding) algorithm.

[0117] The dimensionality reduction process refers to using the sample data after the dimensionality reduction process to simulate the similarity between the sample data before the dimensionality reduction process.

[0118] In the embodiment of the present application, after the sample data of each type of business is obtained, the preset high-to-low dimensional conversion relationship can be used to reduce the dimensionality of the sample data of each type of business in the high-dimensional space to obtain intermediate processed sample data. It should be noted that the sample data of each type of business in the high-dimensional space is in a one-to-one correspondence with the intermediate processed sample data after the dimensionality reduction process, that is, a sample data X in the high-dimensional space is i , that is, there is a corresponding intermediate processing sample data Y in the low-dimensional space i .

[0119] S402: Obtain sample data after dimensionality reduction processing according to a first similarity between any two sample data in the sample data of each type of business and a second similarity between any two sample data in the intermediate processed sample data.

[0120] In an embodiment of the present application, after obtaining the sample data of each type of business and the intermediate processed sample data as mentioned above, the first similarity between any two sample data in the sample data of each type of business is determined, and the second similarity between any two sample data in the intermediate processed sample data is determined, and the sample data after dimensionality reduction processing is determined based on the first similarity and the second similarity.

[0121] The dimensionality reduction method provided in the embodiment of the present application obtains sample data after dimensionality reduction based on the similarity between data in high-dimensional space and the similarity between data in low-dimensional space, thereby providing a data basis for subsequent training of a recognition model based on the dimensionality reduction data.

[0122] In one embodiment, Figure 4 Based on the embodiment shown, the process of obtaining sample data after dimensionality reduction processing can be further described, such as Figure 5 As shown, the above S402 "obtaining sample data after dimensionality reduction processing according to the first similarity between any two sample data in the sample data of each type of business and the second similarity between any two sample data in the intermediate processed sample data" includes:

[0123] S501: Determine a first similarity between any two sample data in sample data of each type of business.

[0124] In the embodiment of the present application, for any two sample data in the sample data of each type of business, the first similarity between the two sample data is determined, as shown in the following formula (1):

[0125]

[0126] in, Sample data representing various types of business and sample data The first similarity between them is expressed by conditional probability. Refers to the data points The Gaussian mean square error centered on , The value of is different; it should be noted that .

[0127] S502: Determine a second similarity between any two sample data in the intermediate processed sample data.

[0128] In the embodiment of the present application, for any two sample data in the intermediate processed sample data, the second similarity between the two sample data is determined, as shown in the following formula (2):

[0129]

[0130] in, Represents any two sample data in the intermediate processing sample data and The second similarity between them is expressed by conditional probability; it should be noted that .

[0131] S503. Based on the principle of the same similarity, the preset high-low dimensional conversion relationship is adjusted according to the first similarity and the second similarity, and the sample data of each type of business is subjected to dimensionality reduction processing based on the adjusted preset high-low dimensional conversion relationship to obtain the sample data after dimensionality reduction processing.

[0132] The principle of the same similarity means that the similarity between any two data after the high-low dimensional conversion process is the same as the similarity between any two data before the high-low dimensional conversion process. It should be noted that only when the principle of the same similarity is guaranteed can the dimensionality reduction effect of the data be guaranteed.

[0133] In an embodiment of the present application, after obtaining the first similarity between any two sample data in the sample data of each type of business and the second similarity between any two sample data in the intermediate processed sample data, the preset high-low dimensional conversion relationship is adjusted according to the numerical relationship between the first similarity and the second similarity, and the sample data of each type of business is re-dimensionalized based on the adjusted preset high-low dimensional conversion relationship to obtain the sample data after dimensionality reduction processing.

[0134] Optionally, a specific implementation method of obtaining sample data after dimensionality reduction processing according to the first similarity and the second similarity is provided below, such as Figure 6 As shown, that is, S503 "based on the principle of the same similarity, adjusting the preset high-low dimensional conversion relationship according to the first similarity and the second similarity, and performing dimensionality reduction processing on the sample data of each type of business based on the adjusted preset high-low dimensional conversion relationship to obtain sample data after dimensionality reduction processing", includes:

[0135] S5031: Determine whether the first similarity and the second similarity are consistent. If the first similarity and the second similarity are consistent, execute S5032; if the first similarity and the second similarity are inconsistent, execute S5033.

[0136] In the embodiment of the present application, after the first similarity and the second similarity are obtained, it can be further determined whether the first similarity and the second similarity are consistent, and if the first similarity and the second similarity are consistent, S5032 is executed, and if the first similarity and the second similarity are inconsistent, S5033 is executed.

[0137] S5032. Use the intermediate processed sample data as sample data after dimensionality reduction processing.

[0138] In the embodiment of the present application, when the first similarity and the second similarity are consistent, the intermediate processed sample data is directly used as the sample data after dimensionality reduction processing.

[0139] S5033: Adjust the preset high-low dimensional conversion relationship, and return to execute S401 based on the adjusted preset high-low dimensional conversion relationship.

[0140] In an embodiment of the present application, when the first similarity and the second similarity are inconsistent, it means that the preset high-low dimensional conversion relationship is unreasonable, and the preset high-low dimensional conversion relationship needs to be adjusted to obtain the adjusted preset high-low dimensional conversion relationship, and the adjusted preset high-low dimensional conversion relationship is used as the new preset high-low dimensional conversion relationship. According to the new preset high-low dimensional conversion relationship, the sample data of each type of business is subjected to dimensionality reduction processing to obtain new intermediate processed sample data, and the first similarity between any two sample data in the sample data of each type of business and the second similarity between any two sample data in the new intermediate processed sample data are determined, and it is determined again whether the first similarity and the second similarity are consistent. If they are consistent, the new intermediate processed sample data is used as the sample data after dimensionality reduction processing. If they are inconsistent, the preset high-low dimensional conversion relationship is adjusted again, ..., until the first similarity and the second similarity are consistent to obtain the final sample data after dimensionality reduction processing.

[0141] The dimensionality reduction method provided in the embodiment of the present application obtains sample data after dimensionality reduction based on the similarity between data in high-dimensional space and the similarity between data in low-dimensional space, thereby providing a data basis for subsequent training of a recognition model based on the dimensionality reduction data.

[0142] In one embodiment, Figure 3 Based on the embodiment shown, Figure 7 As shown, the above method also includes:

[0143] S303: Extract features from sample data of each type of business to obtain data features of the sample data.

[0144] Among them, feature extraction is the process of selecting, transforming and combining relevant information from raw data in order to better express the essential characteristics of the data.

[0145] Among them, the data features include key features, illegal behavior process features, object behavior features, and illegal transfer of resources features. The key features represent the frequency of occurrence of fraud-related keywords during the chat process of transaction users. The illegal behavior process features represent the amount of transaction and the login frequency of transaction accounts. The object behavior features refer to the frequency of occurrence of transaction objects. The illegal transfer of resources features refer to the total transaction amount.

[0146] In an embodiment of the present application, key features are obtained by searching the frequency of occurrence of fraud-related keywords in sample data and then calculating the mean; or, the average value of the transaction amount and the number of transaction account logins in the sample data is calculated to obtain illegal behavior process features; or, the number of occurrences of transaction payment objects in the sample data is merged and calculated, which is regarded as object behavior features; the variance of the transaction inflow and outflow amounts in the sample data is calculated, etc., to eliminate the influence of the positive and negative signs of the inflow and outflow amounts on the mean, which is regarded as an illegal resource transfer feature.

[0147] The above S301 “performing dimensionality reduction processing on sample data of each type of business according to a preset high-to-low dimensional conversion relationship to obtain sample data after dimensionality reduction processing” includes:

[0148] S301. According to a preset high-low dimensional conversion relationship, the data features of the sample data are subjected to dimensionality reduction processing to obtain the sample data after dimensionality reduction processing.

[0149] In the embodiment of the present application, since the sample data of each type of business obtained are all high-dimensional data, when constructing the recognition model, the high-dimensional data will cause the recognition model to have greater difficulty in converging during the training process, so it is necessary to convert the high-dimensional data into low-dimensional data to reduce the difficulty of converging the recognition model during the training process. After the sample data of each type of business is obtained as described above, the data features of the sample data of each type of business can be extracted first, and the data features of the sample data of each type of business in the high-dimensional space can be reduced in dimension by using a preset high-low dimensional conversion relationship to obtain the sample data after the dimensionality reduction processing.

[0150] The data feature extraction method provided in the embodiment of the present application ensures that the data points of each sample data are close to each other before the data is input into the initial recognition model training, which facilitates the movement of each clustering center and improves the training efficiency of the initial recognition model.

[0151] In one embodiment, Figure 3 Based on the embodiment shown, the process of constructing the target recognition model can be described as follows: Figure 8 As shown, the above S302 "training the initial recognition model according to the sample data after dimensionality reduction processing and the preset risk indicators to construct a target recognition model" includes:

[0152] S601. Formulate identification rules based on preset risk indicators.

[0153] Among them, the identification rules include the risk level corresponding to each indicator range in the preset risk indicators.

[0154] In the example of the present application, after the preset risk indicators are obtained as mentioned above, identification rules can be formulated based on the preset risk indicators. For example, the preset risk indicators include two transaction amounts, namely 10,000 yuan and 100,000 yuan. The identification rules formulated based on the preset risk indicators include less than 10,000 yuan as low risk, greater than 10,000 yuan and less than 100,000 yuan as medium risk, and greater than 100,000 yuan as high risk.

[0155] S602: Input the sample data after dimensionality reduction processing and the recognition rules into the initial recognition model, so that the initial recognition model is trained based on the recognition rules to construct a target recognition model.

[0156] In an embodiment of the present application, after the recognition rules are obtained as described above, the sample data after the dimensionality reduction processing and the recognition rules can be respectively input into the initial recognition model, so that the initial recognition model is trained based on the recognition rules to construct a target recognition model. Optionally, during the training process of the target recognition model, a cross entropy loss function can be used to determine whether the target recognition model has been trained. For example, during the training process of the target recognition model, if the value of the cross entropy loss function is less than a preset threshold, the target recognition model is considered to have been trained. If the value of the cross entropy loss function is not less than the preset threshold, the target recognition model is considered to have not been trained. In the case where the target recognition model has not been trained, data embedding is achieved by iteratively updating the data point positions. This process is achieved by using an optimization algorithm called "gradient descent" until the target recognition model has not been trained.

[0157] For example, labels are first added to the sample data after dimensionality reduction processing according to the recognition rules, and the labeled sample data is input into the initial recognition model, the initial recognition model is trained, and the target recognition model is obtained.

[0158] The method for constructing a target recognition model provided in an embodiment of the present application first formulates recognition rules according to preset risk indicators, and then trains the initial recognition model based on the recognition rules and sample data after dimensionality reduction processing, which greatly improves the training efficiency of the model.

[0159] In one embodiment, Figure 2 Based on the embodiment shown, the process of obtaining the verified target recognition model can be described as follows: Fig. 9 As shown, the above S203 "verifying the accuracy of the target recognition model to obtain a verified target recognition model" includes:

[0160] S701. Input the verification sample data into the target recognition model for recognition, and obtain the risk value corresponding to the verification sample data.

[0161] Among them, verification sample data refers to sample data obtained from various registration platforms of transaction users, including basic information, asset information, account information and transaction information of transaction users, among which the basic information includes basic information, family information, occupational information and residential address information of transaction users, the asset information includes asset status information, the account information includes account transaction information, account login information and transaction account status information, and the transaction information includes transaction amount information, transaction frequency information, transaction payment and receipt object information, abnormal behavior information, etc.

[0162] Among them, the risk value represents whether the verification sample data is low risk, medium risk, high risk or no risk.

[0163] In an embodiment of the present application, after the target recognition model is constructed as described above, verification sample data is obtained, and the verification sample data is input into the target recognition model for recognition to obtain a risk value corresponding to the verification sample data.

[0164] S702: Determine the index hit rate of the target recognition model by statistically verifying the risk value corresponding to the sample data.

[0165] The indicator hit rate includes the indicator false alarm rate and user hit rate of the target recognition model. The indicator false alarm rate refers to the target recognition model identifying non-risk data as risky data, or identifying low-risk data as high-risk data, or identifying high-risk data as medium-risk data, etc. The user hit rate refers to the probability of the target recognition model identifying accurately.

[0166] In the embodiment of the present application, all the acquired verification sample data are sequentially or simultaneously input into the target recognition model for recognition, and the risk value corresponding to each verification sample data is obtained, and the number of risk values ​​corresponding to each verification sample data that are low risk, medium risk or high risk is determined, and the ratio of the number of risk values ​​that are low risk, medium risk or high risk to the total number of risk values ​​is determined as the indicator hit rate of the target recognition model. Optionally, the indicator hit rate, false alarm rate and user hit rate are calculated as shown in the following formulas (3)-(5):

[0167]

[0168]

[0169]

[0170] in, is the indicator hit rate, is the number of people at risk, is the total number of risks, is the false alarm rate, a is the number of people who are at risk of false alarms, is the user hit rate, Normal risk hits the number of people.

[0171] S703: Obtain a verified target recognition model according to the indicator hit rate of the target recognition model.

[0172] In an embodiment of the present application, after the indicator hit rate of the target recognition model is determined as above, if the indicator hit rate meets the preset hit rate threshold, it means that the target recognition model is qualified, that is, the target recognition model is used as a verified target recognition model; if the indicator hit rate does not meet the preset hit rate threshold, it means that the target recognition model is unqualified, and it is necessary to adjust the parameters of the target recognition model until the indicator hit rate output by the target recognition model after the adjustment of the parameters meets the preset hit rate threshold, so as to obtain a verified target recognition model.

[0173] The target recognition model verification method provided in the embodiment of the present application ensures the recognition accuracy of the target recognition model.

[0174] In one embodiment, Figure 2 Based on the embodiment shown, Fig.10 As shown, the above method also includes:

[0175] S204: Standardize the sample data of each type of business to obtain standardized sample data.

[0176] Among them, standardization processing includes at least one of data cleaning processing, data conversion processing, data normalization processing and data verification processing. Data cleaning processing refers to removing duplication, spaces, formats, etc.; data conversion processing refers to converting dates, currencies, units of measurement, etc. to make them more uniform; data normalization processing refers to converting text into unified terms and formats, such as setting rules, dictionaries and data formats; data verification processing refers to verifying data to ensure compliance with requirements.

[0177] In an embodiment of the present application, after the sample data of each type of business is obtained as described above, the sample data of each type of business may be subjected to standardization processing such as data cleaning processing, data conversion processing, data normalization processing and data verification processing to obtain the sample data after standardization processing.

[0178] The above S202 “training the initial recognition model and building the target recognition model based on the sample data of each type of business and the preset risk indicators” includes:

[0179] S202. Based on the standardized sample data and the preset risk indicators, the initial recognition model is trained to construct a target recognition model.

[0180] In the embodiment of the present application, after the sample data after the standardization processing and the preset risk index are obtained, the sample data after the standardization processing and the preset risk index can be input into the initial recognition model, and the initial recognition model can be trained to obtain the target recognition model. Optionally, the sample data after the standardization processing can be labeled based on the preset risk index, and the labels include: no risk, low risk, medium risk and high risk, etc., and then the sample data after the label is added is input into the initial recognition model, and the initial recognition model is trained to obtain the trained initial recognition model, that is, the target recognition model.

[0181] The standardized processing method provided in the embodiment of the present application ensures the accuracy and standardization of information, provides a data basis for subsequent training of recognition models based on standardized sample data, and further improves the accuracy of the trained model.

[0182] In one embodiment, Fig.11 As shown, an identification method is provided, which is described by taking the method applied to the computer device in FIG. 1 as an example, and includes the following steps:

[0183] S801: Obtain service interaction information of a user to be identified.

[0184] Among them, business interaction information refers to sample data obtained from various registration platforms of transaction users, including basic information, asset information, account information and transaction information of transaction users. The basic information includes basic information, family information, occupational information, and residential address information of transaction users. The asset information includes asset status information. The account information includes account transaction information, account login information, and transaction account status information. The transaction information includes transaction amount information, transaction frequency information, transaction payment and receipt object information, abnormal behavior information, etc.

[0185] In the embodiment of the present application, when it is necessary to identify the risks in the transaction process, the data in the transaction process can be obtained first, and the data in the transaction process can be input into a pre-trained recognition model for training to obtain the risk identification result. Optionally, the business interaction information of the user to be identified can be obtained from various types of business systems (for example, banking systems, transaction systems, payment systems, etc.); or, the business interaction information of the user to be identified can be obtained from other devices using an interface (for example, a USB interface).

[0186] S802: Extract features from the service interaction information to obtain service features corresponding to the service interaction information.

[0187] Among them, the features include key features, illegal behavior process features, object behavior features, and illegal transfer of resources features. The key features represent the frequency of occurrence of fraud-related keywords during the chat process of transaction users. The illegal behavior process features represent the amount of transaction and the login frequency of transaction accounts. The object behavior features refer to the frequency of occurrence of transaction objects. The illegal transfer of resources features refer to the total transaction amount.

[0188] In an embodiment of the present application, the key features are obtained by searching the frequency of occurrence of fraud-related keywords in the business interaction information and then calculating the mean; or, the transaction amount and the number of transaction account logins in the business interaction information are averaged to obtain the illegal behavior process characteristics; or, the number of occurrences of the transaction payment and receipt objects in the business interaction information are merged and calculated, which is regarded as the object behavior characteristics; the variance of the transaction input and output amounts in the business interaction information is calculated, etc., to eliminate the influence of the positive and negative signs of the input and output amounts on the mean, which is regarded as the illegal resource transfer characteristics.

[0189] S803: Perform risk identification of illegal behaviors according to business characteristics and target identification models, and obtain risk values ​​corresponding to business interaction information.

[0190] Among them, the target recognition model is obtained by training the initial recognition model based on sample data of various types of businesses and preset risk indicators, and its accuracy is verified.

[0191] In the embodiment of the present application, after obtaining the business features corresponding to the business interaction information of the user to be identified, the business features are input into the target recognition model to identify the risk of illegal behavior, and the risk value corresponding to the business interaction information is obtained. The risk value corresponding to the business interaction information includes low risk, medium risk, high risk and no risk.

[0192] The identification method provided in the embodiment of the present application realizes effective identification of risky behaviors in business interaction information by extracting business features in business interaction information and identifying risk values ​​corresponding to the business interaction information based on an identification model according to the business features.

[0193] In one embodiment, Fig.12 As shown, a complete method for building a recognition model is also provided, including:

[0194] S10. Obtain sample data and preset risk indicators for various types of businesses;

[0195] S11, performing standardization processing on sample data of each type of business to obtain standardized sample data;

[0196] S12, extracting features from the sample data after standardization to obtain data features of the sample data;

[0197] S13, according to the preset high-low dimensional conversion relationship, the data features of the sample data are subjected to dimensionality reduction processing to obtain intermediate processed sample data;

[0198] S14, determining a first similarity between any two sample data in the sample data of each type of business;

[0199] S15, determining a second similarity between any two sample data in the intermediate processed sample data;

[0200] S16, based on the principle of the same similarity, adjusting the preset high-low dimensional conversion relationship according to the first similarity and the second similarity, and performing dimensionality reduction processing on the sample data of each type of business based on the adjusted preset high-low dimensional conversion relationship to obtain sample data after dimensionality reduction processing;

[0201] S17. Formulate identification rules based on preset risk indicators;

[0202] S18, inputting the sample data after dimensionality reduction processing and the recognition rules into the initial recognition model, so that the initial recognition model is trained based on the recognition rules to construct a target recognition model;

[0203] S19, input the verification sample data into the target recognition model for recognition, and obtain the risk value corresponding to the verification sample data;

[0204] S20, determining the index hit rate of the target recognition model by statistically verifying the risk value corresponding to the sample data;

[0205] S21. Obtain a verified target recognition model according to the indicator hit rate of the target recognition model.

[0206] The above method trains the initial recognition model according to the sample data and the preset risk indicators at the same time to obtain the target recognition model. Compared with the existing method of constructing the recognition model only based on the sample data, the target recognition model constructed in the present application can identify the risks more accurately because it takes into account the preset risk indicators. In addition, after the target recognition model is constructed, the accuracy of the target recognition model is also verified to further ensure the recognition accuracy of the target recognition model.

[0207] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0208] Based on the same inventive concept, the embodiment of the present application also provides a device for constructing a recognition model for implementing the above-mentioned method for constructing a recognition model. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in the embodiments of the device for constructing one or more recognition models provided below can refer to the limitations of the method for constructing a recognition model in the above text, and will not be repeated here.

[0209] In an exemplary embodiment, as shown in FIG13 , a device for constructing a recognition model is provided, comprising: an acquisition module 10, a construction module 11 and a verification module 12, wherein:

[0210] The acquisition module 10 is used to obtain sample data and preset risk indicators of various types of businesses.

[0211] The construction module 11 is used to train the initial recognition model according to the sample data of each type of business and the preset risk indicators to construct the target recognition model.

[0212] The verification module 12 is used to verify the accuracy of the target recognition model to obtain a verified target recognition model.

[0213] In an exemplary embodiment, the construction module 11 includes: a dimensionality reduction processing unit and a training unit, wherein:

[0214] The dimension reduction processing unit is specifically used to perform dimension reduction processing on the sample data of each type of business according to the preset high-low dimensional conversion relationship to obtain the sample data after dimension reduction processing; the dimension reduction processing refers to using the sample data after dimension reduction processing to simulate the similarity between the sample data before dimension reduction processing;

[0215] The training unit is specifically used to train the initial recognition model based on the sample data after dimensionality reduction processing and the preset risk indicators, and to build a target recognition model.

[0216] In an exemplary embodiment, the above-mentioned dimensionality reduction processing unit is specifically used to perform dimensionality reduction processing on sample data of each type of business according to a preset high-low dimensional conversion relationship to obtain intermediate processed sample data; and obtain the sample data after dimensionality reduction processing according to a first similarity between any two sample data in the sample data of each type of business and a second similarity between any two sample data in the intermediate processed sample data.

[0217] In an exemplary embodiment, the above-mentioned dimensionality reduction processing unit is specifically used to determine the first similarity between any two sample data in the sample data of each type of business; determine the second similarity between any two sample data in the intermediate processed sample data; based on the principle of the same similarity, adjust the preset high-low dimensional conversion relationship according to the first similarity and the second similarity, and perform dimensionality reduction processing on the sample data of each type of business based on the adjusted preset high-low dimensional conversion relationship to obtain the sample data after dimensionality reduction processing.

[0218] In an exemplary embodiment, the above-mentioned dimensionality reduction processing unit is specifically used to determine whether the first similarity and the second similarity are consistent; if the first similarity and the second similarity are consistent, the intermediate processed sample data is used as the sample data after dimensionality reduction processing; if the first similarity and the second similarity are inconsistent, the preset high-low dimensional conversion relationship is adjusted, and based on the adjusted preset high-low dimensional conversion relationship, the step of returning to execute dimensionality reduction processing on the sample data of each type of business according to the preset high-low dimensional conversion relationship to obtain the intermediate processed sample data.

[0219] In an exemplary embodiment, the construction module 11 further includes: a feature extraction unit, which is specifically used to extract features from sample data of various types of services to obtain data features of the sample data; the data features include key features, illegal behavior process features, object behavior features, and illegal resource transfer features;

[0220] The above-mentioned dimensionality reduction processing unit is specifically used to perform dimensionality reduction processing on the data features of the sample data according to a preset high-low dimensional conversion relationship to obtain the sample data after dimensionality reduction processing.

[0221] In an exemplary embodiment, the above-mentioned training unit is specifically used to formulate identification rules based on preset risk indicators; the identification rules include the risk level corresponding to each indicator range in the preset risk indicators; the sample data after dimensionality reduction processing and the identification rules are input into the initial recognition model, so that the initial recognition model is trained based on the recognition rules to construct a target recognition model.

[0222] In an exemplary embodiment, the verification module 12 includes: an identification unit, a determination unit and an acquisition unit, wherein:

[0223] The identification unit is specifically used to input the verification sample data into the target identification model for identification, and obtain the risk value corresponding to the verification sample data;

[0224] The determination unit is specifically used to determine the indicator hit rate of the target recognition model by statistically verifying the risk value corresponding to the sample data; the indicator hit rate includes the indicator false alarm rate and user hit rate of the target recognition model;

[0225] The acquisition unit is specifically used to obtain a verified target recognition model according to the indicator hit rate of the target recognition model.

[0226] In an exemplary embodiment, the apparatus further comprises: a standardization processing module, configured to perform standardization processing on sample data of various types of services to obtain standardized sample data; the standardization processing comprises at least one of data cleaning processing, data conversion processing, data normalization processing and data verification processing;

[0227] The above-mentioned construction module 11 is also used to train the initial recognition model according to the standardized sample data and the preset risk indicators to construct the target recognition model.

[0228] In an exemplary embodiment, as shown in FIG14 , a recognition device is provided, comprising: an acquisition module 20, a feature extraction module 21 and a recognition module 22, wherein:

[0229] An acquisition module 20 is used to acquire the service interaction information of the user to be identified;

[0230] The feature extraction module 21 is used to extract features from the business interaction information to obtain business features corresponding to the business interaction information; the features include key features, illegal behavior process features, object behavior features, and illegal resource transfer features;

[0231] The identification module 22 is used to identify the risk of illegal behavior based on business characteristics and target identification models, and obtain the risk value corresponding to the business interaction information; the target identification model is obtained by training the initial identification model based on sample data of various types of business and preset risk indicators, and the accuracy is verified.

[0232] The above recognition model construction device and each module in the recognition device can be implemented in whole or in part by software, hardware and their combination. The above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0233] In an exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be shown in FIG1. ​​The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store business interaction information data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for constructing a recognition model is implemented.

[0234] Those skilled in the art will understand that the structure shown in FIG. 1 is merely a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0235] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0236] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0237] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0238] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0239] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0240] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0241] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for constructing a recognition model, It is characterized in that The method comprises: Obtain sample data and preset risk indicators for various types of businesses; According to the sample data of each type of business and the preset risk indicators, the initial recognition model is trained to construct a target recognition model; The target recognition model is verified for accuracy to obtain a verified target recognition model.

2. The method according to claim 1, It is characterized in that The training of the initial recognition model based on the sample data of each type of business and the preset risk indicators to construct a target recognition model includes: According to the preset high-low dimensional conversion relationship, the sample data of each type of business is subjected to dimensionality reduction processing to obtain sample data after dimensionality reduction processing; the dimensionality reduction processing refers to using the sample data after dimensionality reduction processing to simulate the similarity between the sample data before dimensionality reduction processing; According to the sample data after the dimensionality reduction process and the preset risk indicators, the initial recognition model is trained to construct a target recognition model.

3. The method according to claim 2, It is characterized in that The step of performing dimensionality reduction processing on the sample data of each type of business according to the preset high-low dimensional conversion relationship to obtain the sample data after dimensionality reduction processing includes: According to the preset high-low dimensional conversion relationship, the sample data of each type of business is subjected to dimensionality reduction processing to obtain intermediate processed sample data; The sample data after the dimensionality reduction processing is obtained according to the first similarity between any two sample data in the sample data of each type of business and the second similarity between any two sample data in the intermediate processed sample data.

4. The method according to claim 3, It is characterized in that The obtaining the sample data after the dimensionality reduction processing according to the first similarity between any two sample data in the sample data of each type of business and the second similarity between any two sample data in the intermediate processed sample data includes: Determine a first similarity between any two sample data in the sample data of each type of business; Determine a second similarity between any two sample data in the intermediate processed sample data; Based on the principle of the same similarity, the preset high-low dimensional conversion relationship is adjusted according to the first similarity and the second similarity, and the sample data of each type of business is subjected to dimensionality reduction processing based on the adjusted preset high-low dimensional conversion relationship to obtain the sample data after dimensionality reduction processing.

5. The method according to claim 4, It is characterized in that The step of adjusting the preset high-low dimensional conversion relationship according to the first similarity and the second similarity, and performing dimensionality reduction processing on the sample data of each type of business based on the adjusted preset high-low dimensional conversion relationship to obtain the sample data after dimensionality reduction processing includes: determining whether the first similarity and the second similarity are consistent; If the first similarity is consistent with the second similarity, the intermediate processed sample data is used as sample data after dimensionality reduction processing; If the first similarity and the second similarity are inconsistent, the preset high-low dimensional conversion relationship is adjusted, and based on the adjusted preset high-low dimensional conversion relationship, the step of returning to execute the step of performing dimensionality reduction processing on the sample data of each type of business according to the preset high-low dimensional conversion relationship to obtain intermediate processed sample data.

6. The method according to claim 2, It is characterized in that The method further comprises: Extracting features from the sample data of each type of business to obtain data features of the sample data; the data features include key features, illegal behavior process features, object behavior features, and illegal resource transfer features; The step of performing dimensionality reduction processing on the sample data of each type of business according to the preset high-low dimensional conversion relationship to obtain the sample data after dimensionality reduction processing includes: According to a preset high-to-low dimensional conversion relationship, dimension reduction processing is performed on the data features of the sample data to obtain sample data after dimension reduction processing.

7. The method according to claim 2, It is characterized in that The training of the initial recognition model based on the sample data after the dimensionality reduction process and the preset risk index to construct the target recognition model includes: Formulate identification rules according to the preset risk indicators; the identification rules include the risk level corresponding to each indicator range in the preset risk indicators; The sample data after the dimensionality reduction process and the recognition rules are input into the initial recognition model, so that the initial recognition model is trained based on the recognition rules to construct the target recognition model.

8. The method according to claim 1, It is characterized in that The step of verifying the accuracy of the target recognition model to obtain a verified target recognition model includes: Inputting the verification sample data into the target recognition model for recognition, and obtaining a risk value corresponding to the verification sample data; By counting the risk values ​​corresponding to the verification sample data, the indicator hit rate of the target recognition model is determined; the indicator hit rate includes the indicator false alarm rate and the user hit rate of the target recognition model; According to the indicator hit rate of the target recognition model, a verified target recognition model is obtained.

9. The method according to claim 1, It is characterized in that The method further comprises: Performing standardization processing on the sample data of each type of business to obtain the sample data after standardization processing; the standardization processing includes at least one of data cleaning processing, data conversion processing, data normalization processing and data verification processing; The training of the initial recognition model based on the sample data of each type of business and the preset risk indicators to construct a target recognition model includes: According to the standardized sample data and the preset risk indicators, the initial recognition model is trained to construct a target recognition model.

10. A method of identifying It is characterized in that The method comprises: Obtaining business interaction information of the user to be identified; Extracting features from the service interaction information to obtain service features corresponding to the service interaction information; The risk of illegal behavior is identified based on the business characteristics and the target identification model to obtain the risk value corresponding to the business interaction information; the target identification model is obtained by training the initial identification model based on sample data of each type of business and preset risk indicators, and the accuracy is verified.

11. A device for constructing a recognition model, It is characterized in that The device comprises: The acquisition module is used to obtain sample data and preset risk indicators of various types of businesses; A training module, used to train the initial recognition model according to the sample data of each type of business and the preset risk indicators, and to construct a target recognition model; The verification module is used to verify the accuracy of the target recognition model to obtain a verified target recognition model.

12. An identification device, It is characterized in that The device also includes: An acquisition module is used to acquire the business interaction information of the user to be identified; An extraction module, used to extract features from the service interaction information to obtain service features corresponding to the service interaction information; The identification module is used to identify the risk of illegal behavior based on the business characteristics and the target identification model, and obtain the risk value corresponding to the business interaction information; the target identification model is obtained by training the initial identification model based on sample data of various types of business and preset risk indicators, and the accuracy is verified.

13. A computer device comprising a memory and a processor, wherein the memory stores a computer program. It is characterized in that When the processor executes the computer program, the steps of the method according to any one of claims 1 to 10 are implemented.

14. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 10 are implemented.

15. A computer program product comprising a computer program, It is characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 10 are implemented.

Citation Information

Patent Citations

  • Transaction risk assessment method and device, computer equipment and storage medium

    CN115293907A

  • Fraud detection model construction method and device, equipment and storage medium

    CN117114689A

  • Credit risk identification method, credit risk identification device, equipment and storage medium

    CN117237083A