A method, device, equipment and medium for extracting credit model features during loan

Abstract

The embodiments of the present invention disclose a method, apparatus, device, and medium for extracting features from a credit model during loan processing. The method comprises: obtaining transaction behavior sequence data of a merchant; converting the transaction behavior sequence data into a merchant feature vector based on a set item vector conversion model; and constructing a user feature vector of the target user based on the transaction behavior data between the target user and at least one merchant and the merchant feature vector corresponding to the merchant. The embodiments of the present invention calculate the user's feature vector through the merchant's feature vector, thereby characterizing the user's implicit transaction features, improving the level of identification of the user's credit risk, and enhancing the level of credit limit increase and interception services for users during loan processing.

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of financial security technology, and in particular to a method, device, equipment, and medium for extracting features of a credit model during lending. Background Art

[0002] The development of internet technology has also brought new opportunities for traditional finance. Within the existing financial sector, with rising consumption levels and continuous technological advancements, various credit scenarios and models are becoming increasingly prevalent. This raises the question of how to better assess users' overdue risk and avoid financial losses.

[0003] The current characteristics of the loan credit model are mostly to quantify user data by analyzing the time interval (recency), consumption frequency (frequency), and consumption amount (monetary) of the user's most recent consumption, characterizing the user's loan behavior. When applied to the loan credit model, it can serve as the user's explicit characteristics to assess the user's loan overdue risk.

[0004] However, this feature extraction method based on merchant classification only depicts the explicit similarities of merchants and cannot capture the implicit similarities of merchants. Therefore, the level of overdue risk identification of loan users is not high, and it is difficult to effectively support the normal development of credit business. Summary of the Invention

[0005] The present invention provides a method, device, equipment and medium for extracting features of a loan credit model, so as to effectively characterize the implicit transaction features of users and improve the level of identification of user credit risks.

[0006] In a first aspect, an embodiment of the present invention provides a method for extracting features of a loan credit model, the method comprising:

[0007] Obtain merchant transaction behavior sequence data;

[0008] Based on a set item vector conversion model, convert the transaction behavior sequence data into a merchant feature vector;

[0009] A user feature vector of the target user is constructed based on transaction behavior data between the target user and at least one merchant, and a merchant feature vector corresponding to the merchant.

[0010] In a second aspect, an embodiment of the present invention further provides a device for extracting features of a loan credit model, the device comprising:

[0011] Transaction data acquisition module, used to obtain merchant transaction behavior sequence data;

[0012] A feature vector conversion module, configured to convert the transaction behavior sequence data into a merchant feature vector based on a set item vector conversion model;

[0013] The feature vector construction module is used to construct a user feature vector of the target user based on transaction behavior data between the target user and at least one merchant and a merchant feature vector corresponding to the merchant.

[0014] In a third aspect, an embodiment of the present invention further provides a device for extracting features of a loan credit model, the device comprising:

[0015] one or more processors;

[0016] a storage device for storing one or more programs;

[0017] When the one or more programs are executed by the one or more processors, the one or more processors implement any of the loan credit model feature extraction methods described in the first aspect.

[0018] In a fourth aspect, an embodiment of the present invention further provides a storage medium comprising computer-executable instructions, on which computer-executable instructions are stored. When the computer-executable instructions are executed by a computer processor, any of the loan credit model feature extraction methods described in the first aspect can be implemented.

[0019] The present invention calculates the user's feature vector through the merchant's feature vector, thereby describing the user's implicit transaction characteristics, improving the level of recognition of the user's credit risk, and enhancing the level of credit limit increase and interception services for users in the loan process. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments made with reference to the following drawings:

[0021] Figure 1 This is a flow chart of a method for extracting features from a loan credit model provided in Example 1 of the present invention;

[0022] Figure 2 Flowchart of a method for extracting features from a loan credit model provided in Example 2 of the present invention

[0023] Figure 3 This is a structural block diagram of a device for extracting features from a loan credit model provided in the third embodiment of the present invention;

[0024] Figure 4 This is a structural diagram of a device for extracting features from a loan credit model provided in the fourth embodiment of the present invention. DETAILED DESCRIPTION

[0025] The present invention will be further described in detail below with reference to the accompanying drawings and examples. It will be understood that the specific embodiments described herein are intended only to illustrate the present invention and are not intended to limit the present invention. It should also be noted that, for ease of description, the accompanying drawings only illustrate portions relevant to the present invention, not all structures.

[0026] Example 1

[0027] Figure 1 This is a flowchart of a method for extracting features from a loan credit model, provided in Example 1 of the present invention. This embodiment is applicable to situations where user features need to be extracted from existing transaction data when using a credit model in financial services. This method can be performed by the device for extracting features from a loan credit model, provided in an embodiment of the present invention. This device can be implemented using software and / or hardware. The device can be configured in a corresponding device or server. The method specifically includes:

[0028] S110, obtaining merchant transaction behavior sequence data;

[0029] The merchant is the participant in a transaction who provides goods or services. It can be a specific store, such as hot pot restaurant A, or a trading platform. In transactions with the platform, the merchant is the trading platform, not the specific store on the platform. For example, in a takeout transaction with hot pot restaurant A on a food delivery platform, the merchant is the food delivery platform, not hot pot restaurant A.

[0030] Specifically, we can identify the target user for whom we want to obtain feature vectors by analyzing their transaction data. We can also select representative merchants with a high number of transactions over a period of time from the database. After obtaining the merchant name, we can concatenate the merchant and the corresponding transaction participant names to obtain the merchant's transaction behavior sequence data.

[0031] S120: based on a set item vector conversion model, convert the transaction behavior sequence data into a merchant feature vector;

[0032] The item vector conversion model is set as a pre-set model to convert user behavior sequence data into feature vector data that can be processed by existing credit models. The merchant feature vector can be understood as an embedded representation of transaction behavior sequence data in a high-dimensional space.

[0033] Specifically, in an embodiment of the present invention, the item2vec method can be used to construct an item vector conversion model. The vectorized representation of sequence data obtained by the Item2Vec method can better extract the similarity features between data, thereby mining the user's implicit transaction characteristics.

[0034] In an embodiment of the present invention, preferably, based on a set item vector conversion model, the transaction behavior sequence data is converted into a merchant feature vector, including: processing the transaction behavior sequence data based on a serialization feature extraction model to obtain processed transaction behavior sequence data; inputting the processed transaction behavior sequence data into the set item vector conversion model to obtain a merchant feature vector.

[0035] Specifically, since the transaction behavior sequence data itself is obtained from the database, the name of each data may be relatively complex, and directly inputting it into the model is inefficient. Therefore, in the embodiment of the present invention, the transaction behavior sequence data is first input into the serialization feature extraction model for processing to ensure that the data input into the setting item vector conversion model is a simple sequence data in character form. For example, if the merchant user corresponding sequence before input is ['Xiao Ming', 'Xiao Wang', 'Xiao Zhang'], then the transaction behavior sequence obtained after processing is ['0001', '0002', '0003']. The simpler transaction behavior sequence after processing is then input into the setting item vector conversion model, which reduces the processing pressure of the setting item vector conversion model and improves the acquisition speed of the merchant feature vector.

[0036] S120: Construct a user feature vector of the target user based on transaction behavior data between the target user and at least one merchant, and a merchant feature vector corresponding to the merchant.

[0037] Among them, the target user, as a participant in the transaction, can be the buyer of the corresponding goods or services.

[0038] Specifically, transaction behavior data can reveal the names of merchants with whom the target user transacts, as well as the number, time, and frequency of transactions. Based on this transaction information and the corresponding merchant's feature vector, a user feature vector for the target user is constructed, calculated based on the merchant vector.

[0039] This embodiment obtains transaction behavior sequence data of merchants; based on a set project vector conversion model, the transaction behavior sequence data is converted into a merchant feature vector; based on the transaction behavior data between the target user and at least one merchant, and the merchant feature vector corresponding to the merchant, a user feature vector of the target user is constructed, and the user's feature vector is calculated through the merchant's feature vector, thereby characterizing the user's implicit transaction characteristics, improving the level of identification of the user's credit risk, and enhancing the level of loan limit increase and interception services for users in the loan process.

[0040] Based on the above embodiment, preferably, constructing a user feature vector of the target user based on transaction behavior data between the target user and at least one merchant and a merchant feature vector corresponding to the merchant includes:

[0041] Based on the transaction behavior data of at least one merchant, the number of transactions between the target user and each merchant is determined; the intermediate feature vector of each merchant is obtained by multiplying the feature vector of each merchant by the corresponding number of transactions; and the vector obtained by adding the intermediate feature vectors of all merchants is used as the user feature vector of the target user.

[0042] For example, let's take the target user Xiao Ming as an example. If Xiao Ming consumes twice at McDonald's, once at a takeaway store on Meituan, and once at a store on Pinduoduo, the vector representation of the target user Xiao Ming is:

[0043] Target user Xiao Ming's feature vector = 2*McDonald's merchant feature vector + 1*Meituan merchant feature vector + 1*Pinduoduo merchant feature vector.

[0044] The advantage of the embodiment of the present invention is that by multiplying the merchant vector by the number of transactions and then further accumulating the sum, the efficiency of implicit transaction feature characterization is further improved on the basis of using the merchant feature vector to obtain the user feature vector of the target user to characterize the implicit transaction characteristics of the target user, and no excessive transaction information in the transaction data is introduced, thereby ensuring the stability and effectiveness of the implicit transaction characteristics.

[0045] Example 2

[0046] Figure 2 This is a flowchart of a method for extracting credit model features in a loan provided by the second embodiment of the present invention. Based on the above embodiments, this embodiment preferably explains the transaction behavior sequence data obtained from the merchant. Figure 2 The method may include:

[0047] S210: Acquire transaction data of the target user within a preset time period, and read the merchant in the transaction data; and use the transaction data as a data source for transaction behavior sequence data of the merchant.

[0048] The preset time can be selected by the system based on the real-time operating load or pre-set by staff. The length of the preset time can be six months, one year, etc., which is not further explained in this embodiment of the present invention. The longer the preset time, the higher the reference value and practicality of the corresponding user feature vector. In the implementation of the present invention, the minimum unit granularity of the preset time length is preferably a complete trading day to ensure that data can be acquired in an orderly manner.

[0049] Specifically, the target user is first determined, and then the merchant name for which vector features need to be obtained is determined based on the transaction data of the corresponding target user. Compared with the solution of directly determining the merchant, the embodiment of the present invention reduces the number of merchants that need to calculate feature vectors, reduces the consumption of computing resources, and effectively improves the processing capability of the device.

[0050] In the embodiment of the present invention, preferably, before using the transaction data as a source of transaction behavior sequence data of the merchant, the method further includes:

[0051] For a trading day in which the number of identified transaction merchants is less than a first threshold, the data of the identified trading day will be removed from the transaction data; for a trading day in which the number of identified transactions is greater than a second threshold, the transaction data exceeding the second threshold on the identified trading day will be truncated.

[0052] Among them, the first quantity threshold and the second quantity threshold are set by the staff according to the needs, or the system can make statistics based on the existing historical data, and use the daily transaction merchant number and daily transaction frequency in the scenario with the highest historical profit as the first quantity threshold and the second quantity threshold respectively to maximize economic benefits. The first quantity threshold and the second quantity threshold can also be set separately according to the daily consumption habits of each target user.

[0053] Specifically, due to the instability of the target user's transactions, for example, if the target user is sick or busy at a certain time, the target user may transact with too few merchants on a certain day, and the data for that trading day may lack representativeness. Alternatively, if the user makes an unusually high frequency of purchases on a trading day, there may be fraudulent transactions or theft, and the authenticity of the corresponding transaction data cannot be guaranteed. This embodiment of the present invention removes data from trading days where the number of merchants transacting on a given day is less than a first threshold, and truncates transaction data that exceeds a second threshold on trading days where the number of daily transactions exceeds a second threshold, thereby achieving efficient screening of transaction behavior sequences and ensuring the representativeness and authenticity of the data used to generate the vector.

[0054] S220: Obtain transaction behavior sequence data of the merchant.

[0055] S230: Based on a set item vector conversion model, convert the transaction behavior sequence data into a merchant feature vector.

[0056] S240: Construct a user feature vector of the target user based on transaction behavior data between the target user and at least one merchant and a merchant feature vector corresponding to the merchant.

[0057] S250: Train the original credit model according to the user feature vector of the target user; and use the trained original credit model as a new credit model.

[0058] Among them, the original credit model is a constructed credit model. Regardless of whether it has been used or improved, as long as it is a constructed credit model used to judge the user's credit ability, it falls within the protection scope of the original credit model of this application.

[0059] Specifically, to effectively train the original credit model, a large number of target users can be selected and corresponding target user feature vector samples can be obtained. The specific method for obtaining target user feature vectors can be referenced in the above embodiment and will not be further described in detail in this embodiment of the present invention. While obtaining the target user feature vectors, the target user's overdue payment history can also be collected. The original credit model can be trained and further improved based on the target user's user feature vectors and past overdue payment history.

[0060] In an embodiment of the present invention, preferably, the original credit model is trained according to the feature vector of the target user, including: combining the user feature vector of the target user with the original feature vector of the target user to obtain a combined feature vector of the target user; and inputting the combined feature vector of the target user into the original credit model for training.

[0061] Among them, the original feature vector is a feature vector that only describes the user's explicit features but does not consider the user's implicit transaction features.

[0062] Specifically, the present invention combines the original feature vector obtained using existing techniques with the user feature vector used in the present invention to generate a combined feature vector for the target user. This combined feature vector simultaneously captures both the explicit and implicit transaction characteristics of the target user. The new credit model trained using this combined training feature vector can more accurately reflect the user's actual consumption habits and credit risk.

[0063] Based on the above embodiments, the embodiments of the present invention further reduce the consumption of computing resources, improve the processing capability of the device, and realize the pre-screening of transaction behavior sequences, thereby ensuring the representativeness and authenticity of the data used to generate vectors. At the same time, the new credit model obtained by training the combined training feature vectors can more accurately reflect the user's actual consumption habits and credit risks.

[0064] Example 3

[0065] Figure 3 This is a structural block diagram of a device for extracting features of a loan credit model provided by the third embodiment of the present invention. The device can execute the method for extracting features of a loan credit model provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method. Figure 3 As shown, the device may include:

[0066] Transaction data acquisition module 310, used to obtain merchant transaction behavior sequence data;

[0067] A feature vector conversion module 320 is configured to convert the transaction behavior sequence data into a merchant feature vector based on a set item vector conversion model;

[0068] The feature vector construction module 330 is configured to construct a user feature vector of the target user based on transaction behavior data between the target user and at least one merchant and a merchant feature vector corresponding to the merchant.

[0069] The above-mentioned product can execute the feature extraction method based on the loan credit model provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0070] Optionally, the feature vector conversion module 320 is specifically used to process the transaction behavior sequence data based on a serialized feature extraction model to obtain processed transaction behavior sequence data; and input the processed transaction behavior sequence data into the set item vector conversion model to obtain a merchant feature vector.

[0071] Optionally, the feature vector construction module 330 is specifically used to determine the number of transactions between the target user and each merchant based on the transaction behavior data of at least one merchant; multiply the feature vector of each merchant by the corresponding number of transactions to obtain the intermediate feature vector of each merchant; and use the vector obtained by adding the intermediate feature vectors of all merchants as the user feature vector of the target user.

[0072] Optionally, the device for extracting features from a loan credit model further includes: a target user determination module 340 .

[0073] The target user determination module 340 is specifically configured to obtain the transaction data of the target user within a preset time period, read the merchant in the transaction data, and use the transaction data as a data source for the merchant's transaction behavior sequence data.

[0074] Optionally, the target user determination module 340 is further used to identify trading days on which the number of transaction merchants is less than a first quantity threshold, and remove the data of the identified trading days from the transaction data; identify trading days on which the number of transactions is greater than a second quantity threshold, and truncate the transaction data that exceeds the second quantity threshold on the identified trading days.

[0075] Optionally, the device for extracting features from a loan credit model further includes a credit model training module 350 .

[0076] The credit model training module 350 is specifically configured to train the original credit model according to the user feature vector of the target user; and use the trained original credit model as a new credit model.

[0077] Optionally, the credit model training module 350 is further configured to combine the user feature vector of the target user with the original feature vector of the target user to obtain a combined feature vector of the target user; and input the combined feature vector of the target user into the original credit model for training.

[0078] The above-mentioned product after further explanation can also execute the data processing method for extracting loan credit model features provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0079] Example 4

[0080] Figure 4 This is a schematic diagram of a device for extracting features from a loan credit model provided by the fourth embodiment of the present invention. Figure 4 As shown, the device includes a processor 40, a memory 41, an input device 42 and an output device 43; the number of processors 40 in the device can be one or more. Figure 4 In the embodiment, a processor 40 is used as an example; the processor 40, the memory 41, the input device 42 and the output device 43 in the device can be connected by a bus or other means. Figure 4 The bus connection is taken as an example.

[0081] Memory 41, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the method for extracting features from a loan credit model in an embodiment of the present invention (e.g., the transaction data acquisition module 310, the feature vector conversion module 320, and the feature vector construction module 330 in the device for extracting features from a loan credit model). Processor 40 executes the software programs, instructions, and modules stored in memory 41 to execute various functional applications and data processing of the device, thereby implementing the aforementioned method for extracting features from a loan credit model.

[0082] The memory 41 may primarily include a program storage area and a data storage area. The program storage area may store an operating system and at least one application required for a function; the data storage area may store data created based on the use of the terminal, etc. Furthermore, the memory 41 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state memory device. In some instances, the memory 41 may further include memory remotely located relative to the processor 40, and these remote memories may be connected to the device via a network. Examples of the aforementioned networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0083] The input device 42 may be used to receive input digital or character information and generate key signal input related to user settings and function control of the terminal. The output device 43 may include a display device such as a display screen.

[0084] Example 5

[0085] A fifth embodiment of the present invention further provides a storage medium containing computer-executable instructions. When the computer-executable instructions are executed by a computer processor, the computer-executable instructions are used to perform a method for extracting features from a loan credit model. The method includes:

[0086] Obtain merchant transaction behavior sequence data;

[0087] Based on a set item vector conversion model, convert the transaction behavior sequence data into a merchant feature vector;

[0088] A user feature vector of the target user is constructed based on transaction behavior data between the target user and at least one merchant, and a merchant feature vector corresponding to the merchant.

[0089] Of course, the computer-executable instructions provided in the storage medium containing computer-executable instructions in the embodiments of the present invention are not limited to the operations of the method described above, but can also perform related operations in the loan credit model feature extraction method provided in any embodiment of the present invention.

[0090] Through the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented with the help of software and necessary general-purpose hardware, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods described in each embodiment of the present invention.

[0091] It is worth noting that in the embodiment of the above-mentioned search device, the various units and modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0092] Note that the above are only preferred embodiments of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments, and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments and may include many other equivalent embodiments without departing from the concept of the present invention. The scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for extracting features from a loan credit model, characterized in that: include: Obtain the target user's transaction data within a preset time period and read the merchants in the transaction data; Using the transaction data as a data source for the merchant's transaction behavior sequence data; Obtaining transaction behavior sequence data of a merchant; the merchant is a participant in the transaction that provides transaction goods or services; Based on a set item vector conversion model, convert the transaction behavior sequence data into a merchant feature vector; Constructing a user feature vector of the target user based on transaction behavior data between the target user and at least one merchant and a merchant feature vector corresponding to the merchant; the target user, as a participant in the transaction behavior, is the purchaser of the corresponding goods or services; The method of constructing a user feature vector of the target user based on transaction behavior data between the target user and at least one merchant and a merchant feature vector corresponding to the merchant includes: Determining the number of transactions between the target user and each merchant based on transaction behavior data of at least one merchant; Multiply the feature vector of each merchant by the corresponding number of transactions to obtain the intermediate feature vector of each merchant; A vector obtained by adding the intermediate feature vectors of all merchants is used as the user feature vector of the target user.

2. The method according to claim 1, characterized in that Before using the transaction data as a source of transaction behavior sequence data for the merchant, the following is also included: Identifying a trading day on which the number of transacting merchants is less than a first threshold number, and removing data on the identified trading day from the transaction data; A trading day in which the number of transactions is greater than a second quantity threshold is identified, and transaction data exceeding the second quantity threshold on the identified trading day is truncated.

3. The method according to claim 1, characterized in that Based on the set item vector conversion model, the transaction behavior sequence data is converted into a merchant feature vector, including: Processing the transaction behavior sequence data based on a serialization feature extraction model to obtain processed transaction behavior sequence data; The processed transaction behavior sequence data is input into the set item vector conversion model to obtain a merchant feature vector.

4. The method according to claim 1, wherein After constructing the user feature vector of the target user based on the transaction behavior data between the target user and at least one merchant and the merchant feature vector corresponding to the merchant, the method further includes: Training the original credit model according to the user feature vector of the target user; The original credit model after training is used as a new credit model.

5. The method according to claim 4, characterized in that The original credit model is trained according to the feature vector of the target user, including: Combining the user feature vector of the target user with the original feature vector of the target user to obtain a combined feature vector of the target user; The combined feature vector of the target user is input into the original credit model for training.

6. A device for extracting features from a loan credit model, characterized in that: The device comprises: The target user determination module is used to obtain the target user's transaction data within a preset time period and read the merchant in the transaction data; and use the transaction data as the data source of the merchant's transaction behavior sequence data; The transaction data acquisition module is used to obtain transaction behavior sequence data of merchants; the merchant is a participant in the transaction that provides transaction goods or services; A feature vector conversion module, configured to convert the transaction behavior sequence data into a merchant feature vector based on a set item vector conversion model; A feature vector construction module is configured to construct a user feature vector of the target user based on transaction behavior data between the target user and at least one merchant and a merchant feature vector corresponding to the merchant; the target user, as a participant in the transaction behavior, is the purchaser of the corresponding goods or services; Among them, the feature vector construction module is specifically used to: determine the number of transactions between the target user and each merchant based on the transaction behavior data of at least one merchant; multiply the feature vector of each merchant by the corresponding number of transactions to obtain the intermediate feature vector of each merchant; and use the vector obtained by adding the intermediate feature vectors of all merchants as the user feature vector of the target user.

7. A device for extracting features from a loan credit model, characterized in that: The device comprises: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the loan credit model feature extraction method as described in any one of claims 1-5.

8. A storage medium comprising computer-executable instructions, wherein the computer-executable instructions, when executed by a computer processor, are used to perform the method for extracting loan credit model features as described in any one of claims 1 to 5.

Images