Method, apparatus, device, and computer readable medium for processing data

Abstract

The application discloses a method, device and equipment for processing data and a computer readable medium, and relates to the technical field of artificial intelligence. A specific embodiment of the method comprises the following steps: dividing merchant historical data into category historical data, wherein the category historical data comprises multiple categories, i.e., merchant historical data, financial historical data, management historical data and guarantee historical data; generating random noise of the same category based on a noise weight of the category historical data of the same category, wherein the noise weight is obtained on the basis of the category historical data of the same category; constructing simulated historical data of the same category by combining the category historical data of the same category with the random noise of the same category; and training and establishing a credit evaluation model by using multiple categories of simulated historical data and the multiple categories of category historical data. The embodiment can increase the quantity of credit evaluation data and grant credit to merchants by training the model.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method, apparatus, device, and computer-readable medium for processing data. Background Technology

[0002] Agricultural wholesale markets are the main carriers of agricultural product circulation, handling over 70% of the total agricultural product circulation volume and playing a vital role in linking production and sales. As a key link and main channel for the circulation of fresh agricultural products, they are characterized by their massive scale, numerous related entities, and intensive capital transactions. Located at the center of the industrial chain, agricultural wholesale markets serve as a bridge connecting millions of producers and consumers, and are hubs for the flow of logistics, information, and capital.

[0003] In the process of realizing this invention, the inventors discovered that the prior art has at least the following problems: the credit rating data of small and micro enterprises or individual businesses such as agricultural product suppliers is insufficient, which makes it impossible to extend credit to the above-mentioned merchants. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a method, apparatus, device, and computer-readable medium for processing data, which can increase the amount of credit rating data and grant credit to merchants through training models.

[0005] To achieve the above objectives, according to one aspect of the present invention, a method for processing data is provided, comprising:

[0006] Merchant historical data is divided into category historical data, which includes the following categories: merchant historical data, financial historical data, management historical data, and guarantee historical data.

[0007] Random noise of the same type is generated based on the noise weights of the historical data of the same type; the noise weights are obtained based on the historical data of the same type.

[0008] Historical data of the same category, combined with random noise of the same category, are used to construct simulated historical data of the same category;

[0009] A credit rating model is trained and established using historical data from multiple types of simulations and historical data from the aforementioned multiple categories.

[0010] The process of categorizing merchant historical data into different types includes:

[0011] The images in the merchant's entered information are converted into image text, and the image text and the entered text in the merchant's entered information are used as the merchant's historical data;

[0012] Analyze the semantics of the merchant's historical data and classify the merchant's historical data into category historical data.

[0013] The process of converting images from merchant-entered information into image text includes:

[0014] The image is selected from the merchant's entered information;

[0015] The image in the merchant's entered information is converted into original image text, and invalid text in the original image text is deleted to obtain the image text.

[0016] The generation of random noise of the same category based on the noise weights added to historical data of the same category includes:

[0017] The noise-adding weights are obtained based on historical data of the same category.

[0018] Random noise of the same type is generated based on the noise weights added to the historical data of the same category.

[0019] The noise-adding weights obtained based on the historical data of the same category include:

[0020] Two categories of historical data are obtained from the same type of historical data.

[0021] A data feature model is trained using historical data from one of the two categories and the difference between the historical data from both categories.

[0022] The category historical data from the same type of historical data is input into the data feature model to obtain the noise weight.

[0023] The noise-adding weights obtained based on the historical data of the same category include:

[0024] The standard deviation of the historical data of the same category is used as the weighted noise.

[0025] The generation of random noise of the same category based on the noise weights added to historical data of the same category includes:

[0026] Within the random noise range of the historical data of the same category, the adjustment parameters are randomly selected.

[0027] Random noise of the same type is generated based on the noise weights and adjustment parameters of the historical data of the same category.

[0028] The historical data of the same category, combined with random noise of the same category, is used to construct simulated historical data of the same category, including:

[0029] Historical data of the same category, combined with random noise of the same category, are used to create new historical data of the same category;

[0030] If the historical data of the same category has the same characteristics as the newly created historical data of the same category, then the newly created historical data of the same category will be used as the simulated historical data of the same category.

[0031] The historical data of the same category, combined with random noise of the same category, is used to establish new historical data of the same category, including:

[0032] By combining a single historical data point of the same category with random noise of the same category, new historical data of the same category can be created.

[0033] If the historical data of the same category has the same characteristics as the newly created historical data of the same category, then the newly created historical data of the same category will be used as the simulated historical data of the same category, including:

[0034] If newly created historical data of the same type falls within the statistical range of historical data of the same type, then the characteristics of historical data of the same type and newly created historical data of the same type are determined to be consistent.

[0035] Newly created historical data of the same type are used as simulated historical data of the same type.

[0036] After constructing the simulated historical data of the same type, the method further includes:

[0037] Add the simulated historical data of the same type to the historical data of the same category;

[0038] Then, based on the noise weights added to the historical data of the same category, random noise of the same category is generated to reconstruct simulated historical data of the same category.

[0039] The step of generating random noise of the same category based on the noise weights of the historical data of the same category to reconstruct simulated historical data of the same category includes:

[0040] Then, based on the noise weights of the historical data of the same category, random noise of the same category is generated to reconstruct simulated historical data of the same category.

[0041] If the simulated historical data of the same type meets the preset construction conditions, then the construction of simulated historical data will stop.

[0042] The process of training and building a credit rating model using multiple types of simulated historical data and the aforementioned multiple types of category historical data includes:

[0043] Annotate the simulated historical data of the multiple categories and the category-specific historical data of the multiple categories;

[0044] Credit parameters of the simulated historical data and the category historical data of the multiple categories after labeling are determined by preset rating conditions;

[0045] To determine the credit parameters of the data, a credit rating model is trained and established.

[0046] According to a second aspect of the present invention, an apparatus for processing data is provided, comprising:

[0047] The segmentation module is used to divide merchant historical data into category historical data, which includes the following categories: merchant historical data, financial historical data, management historical data, and guarantee historical data.

[0048] The generation module is used to generate random noise of the same type based on the noise weights of the historical data of the same type; the noise weights are obtained based on the historical data of the same type.

[0049] A construction module is used to construct simulated historical data of the same category by combining historical data of the same category with random noise of the same category.

[0050] A module is established to train and build a credit rating model using multiple types of simulated historical data and the historical data of the multiple types of categories.

[0051] According to a third aspect of the present invention, an electronic device for processing data is provided, comprising:

[0052] One or more processors;

[0053] Storage device for storing one or more programs.

[0054] When the one or more programs are executed by the one or more processors, the one or more processors perform the methods described above.

[0055] According to a fourth aspect of the present invention, a computer-readable medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method as described above.

[0056] According to a fifth aspect of the present invention, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method as described above in the embodiments of the present invention.

[0057] One embodiment of the above invention has the following advantages or beneficial effects: Merchant historical data is divided into category historical data, which includes multiple categories such as merchant historical data, financial historical data, management historical data, and guarantee historical data; random noise of the same category is generated based on the noise weights of the category historical data of the same category, where the noise weights are obtained based on the category historical data of the same category; simulated historical data of the same category is constructed by combining the category historical data of the same category with the random noise of the same category; a credit rating model is trained and established using the simulated historical data of multiple categories and the category historical data of the multiple categories. Simulated historical data is constructed for merchant historical data, thereby increasing the amount of credit rating data, and credit is granted to merchants through model training.

[0058] The further effects of the aforementioned unconventional alternative methods will be explained below in conjunction with specific implementation methods. Attached Figure Description

[0059] The accompanying drawings are provided to better understand the invention and are not intended to unduly limit the scope of the invention. Wherein:

[0060] Figure 1 This is a schematic diagram of the main flow of a data processing method according to an embodiment of the present invention;

[0061] Figure 2 This is a flowchart illustrating the process of dividing merchant historical data into category historical data according to an embodiment of the present invention;

[0062] Figure 3 This is a schematic diagram illustrating the process of converting images in merchant-entered information into image text according to an embodiment of the present invention;

[0063] Figure 4 This is a schematic diagram of the process for generating random noise of the same type according to an embodiment of the present invention;

[0064] Figure 5 This is a schematic diagram of the process for obtaining noise weights according to an embodiment of the present invention;

[0065] Figure 6 This is a schematic diagram of the process for generating random noise of the same type according to an embodiment of the present invention;

[0066] Figure 7 This is a schematic diagram of the process for constructing simulated historical data of the same type according to an embodiment of the present invention;

[0067] Figure 8 This is a schematic diagram of the process for confirming that historical data of the same type and newly created historical data of the same type have the same characteristics, according to an embodiment of the present invention.

[0068] Figure 9 This is a schematic diagram of the process of reconstructing the same type of simulated historical data according to an embodiment of the present invention;

[0069] Figure 10 This is a schematic diagram of the process for stopping the construction of simulated historical data according to an embodiment of the present invention;

[0070] Figure 11 This is a schematic diagram of the process of training and building a credit rating model according to an embodiment of the present invention;

[0071] Figure 12 This is a schematic diagram of the main structure of a data processing apparatus according to an embodiment of the present invention;

[0072] Figure 13 This is an exemplary system architecture diagram in which embodiments of the present invention can be applied;

[0073] Figure 14 This is a schematic diagram of the structure of a computer system suitable for implementing terminal devices or servers of the present invention. Detailed Implementation

[0074] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of the present invention, including various details to aid understanding. These details should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description. The acquisition, storage, use, and processing of data in the technical solutions of this application all comply with relevant national laws and regulations.

[0075] Agricultural wholesale markets serve as a bridge connecting millions of producers and consumers, and are hubs for logistics, information, and capital flows. From a financial services perspective, agricultural wholesale markets hold a dominant position in the financial sector, leveraging the influence of various participants in the industrial chain, and have comprehensive demands for services such as payment, settlement, and credit.

[0076] Credit lines are typically extended based on credit rating data to support business development. However, credit rating data for the main players in agricultural wholesale markets, agricultural product suppliers, and other small and micro-sized enterprises or individual businesses is insufficient. This lack of data makes it impossible to ascertain their repayment ability, thus preventing credit lines from being extended to them.

[0077] In summary, insufficient credit rating data for small and micro enterprises or individual businesses such as agricultural product suppliers makes it impossible to extend credit to these businesses.

[0078] To address the issue of insufficient credit rating data, which prevents credit from being granted to the aforementioned merchants, the following technical solutions from the embodiments of the present invention can be adopted.

[0079] See Figure 1 , Figure 1 This is a schematic diagram of the main flow of a data processing method according to an embodiment of the present invention. Based on merchant historical data, simulated historical data is constructed to increase the amount of credit rating data. Figure 1 As shown, the specific steps include:

[0080] S101. Merchant historical data is divided into category historical data, which includes the following categories: merchant historical data, financial historical data, management historical data, and guarantee historical data.

[0081] In embodiments of the present invention, the object of processing is merchant historical data. Merchant historical data refers to data from the agricultural product business transactions of micro and small enterprises or individual businesses. These transactions are characterized by a high frequency of transactions and small transaction amounts. Therefore, existing credit granting methods are insufficient for extending credit to these micro and small enterprises and individual businesses.

[0082] In embodiments of the present invention, merchant historical data is started from multiple dimensions and then comprehensively evaluated.

[0083] See Figure 2 , Figure 2 This is a flowchart illustrating the process of dividing merchant historical data into category historical data according to an embodiment of the present invention. Specifically, it includes the following steps:

[0084] S201. Convert the images in the merchant's entered information into image text, and use the image text and the entered text in the merchant's entered information as the merchant's historical data.

[0085] In embodiments of the present invention, merchants typically input information using methods such as taking photos or scanning. The information input by merchants includes images and text, with the images containing merchant data. To improve operational efficiency, the images in the merchant input information can be converted into image text, and the image text and the text input in the merchant input information can be used as the merchant's historical data.

[0086] See Figure 3 , Figure 3 This is a schematic diagram illustrating the process of converting images in merchant-entered information into image text according to an embodiment of the present invention. Specifically, it includes the following steps:

[0087] S301. Filter out images from the merchant's entered information.

[0088] Merchant information includes both images and text. To improve the efficiency of processing merchant information, images are filtered out from the merchant information in file formats. As an example, file formats include one or more of the following: bmp, jpg, png, tif, gif, pcx, tga, exif, fpx, svg, psd, cdr, pcd, dxf, ufo, eps, ai, raw, WMF, webp, avif, and apng.

[0089] S302. Convert the images in the merchant's entered information into original image text, and delete invalid text in the original image text to obtain the image text.

[0090] Image processing can be used to convert images in merchant-entered information into raw image text. Raw image text is the text corresponding to the images in the merchant-entered information. To improve data processing speed and avoid the impact of invalid text, invalid text can be removed from the raw image text to obtain the final image text. As an example, preset keywords are used, and text not containing these preset keywords is considered invalid text. For example: Preset keyword: Merchant Identifier.

[0091] exist Figure 3 In one embodiment, invalid text is removed to improve the effectiveness of image text.

[0092] S202. Analyze the semantics in the merchant's historical data and divide the merchant's historical data into category historical data.

[0093] In embodiments of the present invention, to accurately identify merchants, merchant historical data can be divided into category historical data according to the semantics within the merchant historical data. As an example, Natural Language Processing (NLP) is used to analyze the semantics within the merchant historical data.

[0094] Category history data is a collection of historical data belonging to the same category. Category history data includes several categories, such as merchant history data, financial history data, management history data, and guarantee history data.

[0095] Merchant historical data is historical data that characterizes the features of a merchant. As an example, merchant historical data includes: merchant identifier, merchant address, and merchant business hours, etc.

[0096] Financial history data is historical data that characterizes a merchant's financial status. As an example, financial history data includes: invoice information, tax payment records, and financial statements.

[0097] Managing historical data refers to historical data that characterizes a merchant's daily operations. As an example, managing historical data includes order information, sales information, and social security payment records.

[0098] Guarantee history data is historical data that characterizes a merchant's performance. As an example, guarantee history data includes the guarantee amount, guarantee date, and guarantor.

[0099] In embodiments of the present invention, the categorized historical data includes several types, such as merchant historical data, financial historical data, management historical data, and guarantee historical data. It is understood that the categorized historical data includes at least two types. For historical data of the same type, simulated historical data for that type is constructed.

[0100] Moreover, categorical historical data is structured data, which facilitates the construction of simulated historical data from structured data, thereby increasing the speed of constructing simulated historical data.

[0101] S102. Based on the noise weights of historical data of the same category, generate random noise of the same category. The noise weights are obtained based on the historical data of the same category.

[0102] In embodiments of the present invention, simulated historical data is constructed based on historical data of the same category. It is understood that, for the same category, two types of data are involved: category historical data and simulated historical data.

[0103] As an example, categorized technical data includes merchant historical data and financial historical data. Specifically, simulated historical data of merchant historical data can be constructed using merchant historical data; simulated historical data of financial historical data can be constructed using financial historical data.

[0104] For each category, there is a corresponding noise weight. Random noise for that category is generated based on the noise weight.

[0105] See Figure 4 , Figure 4 This is a schematic diagram of a process for generating random noise of the same type according to an embodiment of the present invention. Specifically, it includes the following steps:

[0106] S401, Noise-adding weights obtained based on historical data of the same category.

[0107] For each category, a noise weight can be obtained. The noise weight is derived based on historical data within the same category. As an example, the standard deviation of historical data within the same category is used as the weighted noise. Using the standard deviation as weighted noise allows for the identification of differences between historical data within the same category.

[0108] See Figure 5 , Figure 5 This is a schematic diagram of the process for obtaining noise-adding weights according to an embodiment of the present invention. Specifically, it includes the following steps:

[0109] S501. Obtain historical data for two categories from the same type of historical data.

[0110] In one embodiment of the present invention, historical data of the same category can be used to train a model to obtain noise-adding weights.

[0111] Specifically, two categories of historical data are obtained from the same category of historical data. As an example, two categories of historical data are randomly selected from the same category of historical data. As another example, to improve the realism of the data, one category of historical data with a value below the average and one category of historical data with a value above the average can be selected.

[0112] S502. A data feature model is trained using historical data from one of the two categories and the difference between the two categories of historical data.

[0113] Using the difference between historical data from two categories, a data feature model can be trained using historical data from one of the two categories. As an example, historical data from one category is used as the input parameter to the data feature model, and the model is trained using the difference between the historical data from the two categories.

[0114] S503. Input the historical data of the same category into the data feature model to obtain the noise weights.

[0115] Once the data feature model has been trained, historical data from one category of the same type of historical data can be input into the data feature model to obtain noise-adding weights.

[0116] In one embodiment of the present invention, a preset number of historical data of the same category can be input into a data feature model to obtain a preset number of noise-adding weights. Then, the average of the preset data noise-adding weights is used as the final noise-adding weight.

[0117] exist Figure 5 In one embodiment, a data feature model is used to obtain the noise-adding weights in order to improve the authenticity of the noise-adding weights.

[0118] S402. Generate random noise of the same type based on the noise weights added to historical data of the same category.

[0119] Random noise of the same type can be generated by adding noise weights to historical data of the same category.

[0120] See Figure 6 , Figure 6 This is a schematic diagram of a process for generating random noise of the same type according to an embodiment of the present invention. Specifically, it includes the following steps:

[0121] S601. Within the random noise range of historical data of the same category, randomly select adjustment parameters.

[0122] Considering the randomness of noise, adjustment parameters can be randomly selected within the random noise range of historical data of the same category. The random noise range differs for different categories of historical data. A random noise range can be preset based on the category. Then, adjustment parameters can be randomly selected within this range.

[0123] S602. Generate random noise of the same type based on the noise weights and adjustment parameters of historical data of the same category.

[0124] Once the adjustment parameters are determined, random noise of the same category can be generated based on the noise weights and adjustment parameters of historical data of the same category. As an example, for the same category, the product of the noise weights and adjustment parameters can be used as random noise.

[0125] exist Figure 6 In one embodiment, randomness is considered on the basis of adding noise weights to generate random noise of the same kind.

[0126] S103. Historical data of the same category, combined with random noise of the same category, are used to construct simulated historical data of the same category.

[0127] Simulated historical data of the same type is constructed based on historical data of the same category and random noise of the same type. As an example, the sum or difference of historical data of the same category and random noise of the same type is used as simulated historical data of the same type.

[0128] See Figure 7 , Figure 7 This is a schematic diagram illustrating the process of constructing simulated historical data of the same type according to an embodiment of the present invention. Specifically, it includes the following steps:

[0129] S701. Combine historical data of the same category with random noise of the same category to create new historical data of the same category.

[0130] By combining historical data of the same category with random noise of the same category, new historical data of the same category can be created. As an example, the average of multiple technical data of the same category can be combined with random noise of the same category to create new historical data of the same category.

[0131] In one embodiment of the present invention, considering the randomness of data, a new historical data of the same category can be created by combining a historical data of the same category with random noise of the same category.

[0132] As an example, the sum or difference of a randomly acquired historical data point of the same category and random noise of the same category can be used as newly created historical data of the same category.

[0133] S702. If the characteristics of historical data of the same category are consistent with those of newly created historical data of the same category, then the newly created historical data of the same category shall be regarded as the simulated historical data of the same category.

[0134] To ensure that simulated historical data and category-specific historical data have the same characteristics, it is necessary to first determine that the characteristics of category-specific historical data and newly created historical data of the same category are consistent, and then use the newly created historical data of the same category as simulated historical data of the same category.

[0135] See Figure 8 , Figure 8 This is a schematic diagram illustrating the process of confirming the consistency of characteristics between historical data of the same category and newly created historical data of the same category, according to an embodiment of the present invention. Specifically, it includes the following steps:

[0136] S801. If newly created historical data of the same type falls within the statistical scope of historical data of the same category, then the characteristics of historical data of the same category and newly created historical data of the same type are determined to be consistent.

[0137] In embodiments of the present invention, whether the features of newly created historical data belong to the statistical range of category historical data is used as the criterion for determining whether the features of category historical data and newly created historical data are consistent. The statistical range of category historical data can be preset for each category. That is, the statistical range of historical data for different categories is different. As an example, the preset statistical range of category historical data includes the area between the upper quartile and the lower quartile of the category historical data.

[0138] If newly created historical data of the same type falls within the statistical scope of historical data of the same category, then the characteristics of historical data of the same category and newly created historical data of the same type are considered consistent. Conversely, if newly created historical data of the same type does not fall within the statistical scope of historical data of the same category, then the characteristics of historical data of the same category and newly created historical data of the same type are considered different.

[0139] S802. Treat newly created historical data of the same type as simulated historical data of the same type.

[0140] Once it is determined that the characteristics of historical data of the same category are consistent with those of newly created historical data of the same category, the newly created historical data of the same category can be used as simulated historical data of the same category.

[0141] exist Figure 8 In the embodiments, the statistical range of category historical data is used to verify the newly created historical data in order to ensure the practicality of the newly created data.

[0142] Thus, a simulated historical dataset has been constructed. However, during data processing, the amount of data in this simulated historical dataset is far from sufficient to meet actual usage needs. Therefore, it is necessary to adopt… Figure 9 In this embodiment, simulated historical data is reconstructed.

[0143] See Figure 9 , Figure 9 This is a schematic diagram illustrating the process of reconstructing the same type of simulated historical data according to an embodiment of the present invention. Specifically, it includes the following steps:

[0144] S901. Add simulated historical data of the same type to the historical data of the same category.

[0145] Simulated historical data of the same type can be added to the historical data of the same category. It is understood that the current category historical data includes not only real category historical data but also simulated historical data.

[0146] S902. Based on the noise weights of historical data of the same category, generate random noise of the same category to reconstruct simulated historical data of the same category.

[0147] Constructing simulated historical data requires adding noise weights, and then, according to S102 and S103, that is, generating random noise of the same type based on the noise weights of historical data of the same category, in order to reconstruct simulated historical data of the same type.

[0148] See Figure 10 , Figure 10 This is a schematic diagram illustrating the process of stopping the construction of simulated historical data according to an embodiment of the present invention. Specifically, it includes the following steps:

[0149] S1001. Based on the noise weights of historical data of the same category, generate random noise of the same category to reconstruct simulated historical data of the same category.

[0150] Then, based on the noise weights added to historical data of the same category, random noise of the same category is generated to reconstruct simulated historical data of the same category. This process is repeated multiple times to obtain multiple simulated historical data of the same category.

[0151] S1002. If the simulated historical data of the same type meets the preset construction conditions, then the construction of simulated historical data will stop.

[0152] If the simulated historical data of the same type meets the preset construction conditions, the construction of simulated historical data will stop; similarly, if the simulated historical data of the same type does not meet the preset construction conditions, the construction of simulated historical data will stop.

[0153] As an example, the preset construction conditions include a threshold for the number of simulated historical data of the same type. That is, if the number of simulated historical data of the same type is greater than or equal to the number of simulated historical data of the same type, the construction of simulated historical data will stop; if the number of simulated historical data of the same type is less than the threshold for the number of simulated historical data of the same type, the construction of simulated historical data will continue.

[0154] exist Figure 9 In one embodiment, simulated historical data is constructed multiple times to meet the construction conditions.

[0155] S104. Use multiple types of simulated historical data and multiple types of category historical data to train and establish a credit rating model.

[0156] Multiple types of simulated historical data are constructed using S102 and S103. Simulated historical data can compensate for the limited quantity of categorical historical data. A credit rating model is trained and built using both simulated and categorical historical data, and then used to assess a merchant's credit limit.

[0157] See Figure 11 , Figure 11 This is a schematic diagram illustrating the process of training and establishing a credit rating model according to an embodiment of the present invention. Specifically, it includes the following steps:

[0158] S1101, label multiple types of simulated historical data and multiple types of category historical data.

[0159] Data annotation is the process of processing unprocessed data such as voice, images, text, and video into machine-readable information.

[0160] In embodiments of the present invention, preset labeling conditions can be used to label multiple types of simulated historical data and multiple types of category historical data. As an example, if there are no default records in the data, the merchant gains 1 point according to the preset labeling conditions; if the data shows good tax payment records, the merchant gains 2 points.

[0161] S1102. By setting up rating conditions, determine the credit parameters of multiple types of simulated historical data and multiple types of category historical data after labeling.

[0162] For historical data simulated across multiple categories and historical data across multiple categories after labeling, credit parameters can be preset for rating conditions. As an example, preset Level 1 rating conditions include: a merchant's total score greater than 90 points, with the information parameter set to "high"; preset Level 2 rating conditions include: a merchant's total score less than or equal to 90 points but greater than or equal to 60 points, with the information parameter set to "medium"; and preset Level 3 rating conditions include: a merchant's total score less than 60 points, with the information parameter set to "low".

[0163] S1103. Determine the credit parameters of the data, train and establish a credit rating model.

[0164] A credit rating model is trained and established using multiple types of simulated historical data, multiple types of historical data, and credit parameters of the data.

[0165] As an example, multiple types of simulated historical data and multiple types of category historical data are used as input parameters for the credit rating model. The credit rating model is trained using the credit parameters of the data. Finally, the trained credit rating model is obtained.

[0166] Among them, the credit rating model can use a recurrent neural network (RNN). Credit rating data often has temporal correlation, meaning that the output of the network at a certain time is related not only to the input at the current time, but also to the output at one or several previous times. RNN is a better choice for time series analysis.

[0167] exist Figure 11 In one embodiment, simulated historical data and category-specific historical data are used to train and build a credit rating model.

[0168] In the above embodiments, merchant historical data is divided into category historical data, which includes multiple categories such as merchant historical data, financial historical data, management historical data, and guarantee historical data. Random noise of the same category is generated based on the noise weights added to the category historical data of the same type. Simulated historical data of the same type is constructed by combining the random noise of the same type with the simulated historical data of the same type. A credit rating model is trained and established using the simulated historical data of multiple types and the multiple category historical data. Simulated historical data is constructed for merchant historical data to increase the amount of credit rating data, and credit is granted to merchants through model training.

[0169] See Figure 12 , Figure 12 This is a schematic diagram of the main structure of a data processing apparatus according to an embodiment of the present invention. The data processing apparatus can implement a data processing method, such as... Figure 12 As shown, the data processing device specifically includes:

[0170] The segmentation module 1201 is used to segment merchant historical data into category historical data, which includes the following categories: merchant historical data, financial historical data, management historical data, and guarantee historical data.

[0171] The generation module 1202 is used to generate random noise of the same type based on the noise weights of the historical data of the same type; the noise weights are obtained based on the historical data of the same type.

[0172] Module 1203 is used to construct simulated historical data of the same category by combining the same category of historical data with random noise of the same category.

[0173] Module 1204 is established to train and build a credit rating model using multiple types of simulated historical data and the multiple types of category historical data.

[0174] In one embodiment of the present invention, the segmentation module 1201 is specifically used to convert the image in the merchant's input information into image text, and use the image text and the input text in the merchant's input information as the merchant's historical data;

[0175] Analyze the semantics of the merchant's historical data and classify the merchant's historical data into category historical data.

[0176] In one embodiment of the present invention, the segmentation module 1201 is specifically used to filter out the image from the merchant's entered information;

[0177] The image in the merchant's entered information is converted into original image text, and invalid text in the original image text is deleted to obtain the image text.

[0178] In one embodiment of the present invention, the generation module 1202 is specifically used to obtain the noise-adding weight based on the historical data of the same category;

[0179] Random noise of the same type is generated based on the noise weights added to the historical data of the same category.

[0180] In one embodiment of the present invention, the generation module 1202 is specifically used to obtain two categories of historical data from the same type of historical data;

[0181] A data feature model is trained using historical data from one of the two categories and the difference between the historical data from both categories.

[0182] The category historical data from the same type of historical data is input into the data feature model to obtain the noise weight.

[0183] In one embodiment of the present invention, the generation module 1202 is specifically used to use the standard deviation of the historical data of the same category as the weighted noise.

[0184] In one embodiment of the present invention, the generation module 1202 is specifically used to randomly select adjustment parameters within the random noise range of the historical data of the same category;

[0185] Random noise of the same type is generated based on the noise weights and adjustment parameters of the historical data of the same category.

[0186] In one embodiment of the present invention, the construction module 1203 is specifically used to build new historical data of the same category by combining the same category of historical data with random noise of the same category.

[0187] If the historical data of the same category has the same characteristics as the newly created historical data of the same category, then the newly created historical data of the same category will be used as the simulated historical data of the same category.

[0188] In one embodiment of the present invention, the construction module 1203 is specifically used to combine a piece of historical data of the same category with random noise of the same category to establish new historical data of the same category.

[0189] In one embodiment of the present invention, the construction module 1203 is specifically used to determine that the newly created historical data of the same type belongs to the statistical range of the historical data of the same type.

[0190] Newly created historical data of the same type are used as simulated historical data of the same type.

[0191] In one embodiment of the present invention, the construction module 1203 is specifically used to add the simulated historical data of the same type to the historical data of the same category;

[0192] Then, based on the noise weights added to the historical data of the same category, random noise of the same category is generated to reconstruct simulated historical data of the same category.

[0193] In one embodiment of the present invention, the construction module 1203 is specifically used to generate random noise of the same type based on the noise weight of the historical data of the same type, so as to reconstruct the simulated historical data of the same type.

[0194] If the simulated historical data of the same type meets the preset construction conditions, then the construction of simulated historical data will stop.

[0195] In one embodiment of the present invention, the establishment module 1204 is specifically used to annotate the multiple types of simulated historical data and the multiple types of category historical data;

[0196] Credit parameters of the simulated historical data and the category historical data of the multiple categories after labeling are determined by preset rating conditions;

[0197] To determine the credit parameters of the data, a credit rating model is trained and established.

[0198] Figure 13 An exemplary system architecture 1300 is shown, in which a method or apparatus for processing data can be applied according to embodiments of the present invention.

[0199] like Figure 13 As shown, system architecture 1300 may include terminal devices 1301, 1302, and 1303, network 1304, and server 1305. Network 1304 is used as a medium to provide a communication link between terminal devices 1301, 1302, and 1303 and server 1305. Network 1304 may include various connection types, such as wired or wireless communication links or fiber optic cables, etc.

[0200] Users can use terminal devices 1301, 1302, and 1303 to interact with server 1305 via network 1304 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 1301, 1302, and 1303, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social media platform software, etc. (for example only).

[0201] Terminal devices 1301, 1302, and 1303 can be various electronic devices with displays and web browsing capabilities, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0202] Server 1305 can be a server that provides various services, such as a backend management server that supports shopping websites browsed by users using terminal devices 1301, 1302, and 1303 (for example only). The backend management server can analyze and process data such as received product information query requests, and feed back the processing results (such as target push information and product information - for example only) to the terminal devices.

[0203] It should be noted that the data processing method provided in the embodiments of the present invention is generally executed by the server 1305, and correspondingly, the data processing device is generally located in the server 1305.

[0204] It should be understood that Figure 13 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0205] A computing program product according to an embodiment of the present invention includes a computer program, which, when executed by a processor, implements the data processing method provided in the embodiment of the present invention.

[0206] The following is for reference. Figure 14 It shows a schematic diagram of the structure of a computer system 1400 suitable for implementing a terminal device of the present invention. Figure 14 The terminal device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0207] like Figure 14 As shown, the computer system 1400 includes a central processing unit (CPU) 1401, which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 1402 or programs loaded from storage section 1408 into random access memory (RAM) 1403. The RAM 1403 also stores various programs and data required for the operation of the system 1400. The CPU 1401, ROM 1402, and RAM 1403 are interconnected via a bus 1404. An input / output (I / O) interface 1405 is also connected to the bus 1404.

[0208] The following components are connected to I / O interface 1405: an input section 1406 including a keyboard, mouse, etc.; an output section 1407 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 1408 including a hard disk, etc.; and a communication section 1409 including a network interface card such as a LAN card, modem, etc. The communication section 1409 performs communication processing via a network such as the Internet. A drive 1410 is also connected to I / O interface 1405 as needed. Removable media 1411, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 1410 as needed so that computer programs read from them can be installed into storage section 1408 as needed.

[0209] In particular, according to the embodiments disclosed in this invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 1409, and / or installed from removable medium 1411. When the computer program is executed by central processing unit (CPU) 1401, it performs the functions defined above in the system of this invention.

[0210] It should be noted that the computer-readable medium shown in this invention can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this invention, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0211] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0212] The modules described in the embodiments of the present invention can be implemented in software or hardware. The described modules can also be housed in a processor; for example, a processor can be described as including a partitioning module, a generation module, a construction module, and a creation module. The names of these modules do not necessarily limit the module itself; for example, a partitioning module can also be described as "used to partition merchant historical data into category historical data, which includes multiple categories such as merchant historical data, financial historical data, management historical data, and guarantee historical data."

[0213] In another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiments; or it may exist independently and not assembled into the device. The computer-readable medium carries one or more programs, which, when executed by the device, cause the device to include:

[0214] Merchant historical data is divided into category historical data, which includes the following categories: merchant historical data, financial historical data, management historical data, and guarantee historical data.

[0215] Random noise of the same type is generated based on the noise weights of the historical data of the same type; the noise weights are obtained based on the historical data of the same type.

[0216] Historical data of the same category, combined with random noise of the same category, are used to construct simulated historical data of the same category;

[0217] A credit rating model is trained and established using historical data from multiple types of simulations and historical data from the aforementioned multiple categories.

[0218] According to the technical solution of this invention, merchant historical data is divided into category historical data, which includes multiple categories such as merchant historical data, financial historical data, management historical data, and guarantee historical data. Random noise of the same category is generated based on the noise weights added to the category historical data of the same category, where the noise weights are obtained from the category historical data of the same category. Simulated historical data of the same category is constructed by combining the random noise of the same category with the simulated historical data of the same category. A credit rating model is trained and established using the simulated historical data of multiple categories and the category historical data of the multiple categories. Simulated historical data is constructed for merchant historical data to increase the amount of credit rating data, and credit is granted to merchants through model training.

[0219] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

[0220] It should be noted that the collection, analysis, use, transmission, and storage of user personal information involved in the technical solution of this invention all comply with relevant laws and regulations, are used for legitimate and reasonable purposes, and are not shared, disclosed, or sold outside of these legitimate uses, and are subject to supervision and management by regulatory authorities. Necessary measures should be taken to prevent unauthorized access to such personal information data, ensure that personnel authorized to access personal information data comply with relevant laws and regulations, and ensure the security of user personal information. Once this user personal information data is no longer needed, the risk should be minimized by restricting or even prohibiting data collection and / or deleting the data.

[0221] When applicable, including in certain relevant applications, data deidentification is used to protect user privacy, such as by removing specific identifiers (e.g., date of birth), controlling the amount or specificity of stored data (e.g., collecting location data at the city level rather than at the specific address level), controlling how data is stored, and / or other methods of deidentification.

Claims

1. A method for processing data, characterized in that, include: Merchant historical data is divided into category historical data, which includes the following categories: merchant historical data, financial historical data, management historical data, and guarantee historical data. The merchant historical data includes: merchant identifier, merchant address, and merchant business hours; the financial historical data includes: invoice information, tax payment records, and financial statements; the management historical data includes order information, sales information, and social security payment records; the guarantee historical data includes guarantee amount, guarantee date, and guarantor. Random noise of the same type is generated based on the noise weights of the historical data of the same category. The noise weights are obtained based on the historical data of the same category. Historical data of the same category, combined with random noise of the same category, are used to construct simulated historical data of the same category; A credit rating model was trained and built using multiple types of simulated historical data and multiple types of historical data. The generation of random noise of the same category based on the noise weights added to historical data of the same category includes: The noise-adding weights are obtained based on historical data of the same category. Based on the noise weights added to the historical data of the same category, random noise of the same category is generated; The noise-adding weights obtained based on the historical data of the same category include: Two categories of historical data are obtained from the same type of historical data. A data feature model is trained using historical data from one of the two categories and the difference between the historical data from both categories. The category historical data from the same type of historical data is input into the data feature model to obtain the noise weights; The generation of random noise of the same category based on the noise weights added to historical data of the same category includes: Within the random noise range of the historical data of the same category, the adjustment parameters are randomly selected. The product of the noise weights and the adjustment parameters of the historical data of the same category is used as the random noise of the same category; The historical data of the same category, combined with random noise of the same category, is used to construct simulated historical data of the same category, including: Historical data of the same category, combined with random noise of the same category, are used to create new historical data of the same category; If the historical data of the same category has the same characteristics as the newly created historical data of the same category, then the newly created historical data of the same category will be used as the simulated historical data of the same category. If the historical data of the same category has the same characteristics as the newly created historical data of the same category, then the newly created historical data of the same category will be used as the simulated historical data of the same category, including: If newly created historical data of the same type falls within the statistical range of historical data of the same type, then the characteristics of historical data of the same type and newly created historical data of the same type are determined to be consistent. Newly created historical data of the same type are used as simulated historical data of the same type.

2. The data processing method according to claim 1, characterized in that, The process of categorizing merchant historical data into different types includes: The images in the merchant's entered information are converted into image text, and the image text and the entered text in the merchant's entered information are used as the merchant's historical data; Analyze the semantics of the merchant's historical data and classify the merchant's historical data into category historical data.

3. The data processing method according to claim 2, characterized in that, The process of converting images from merchant-entered information into image text includes: The image is selected from the merchant's entered information; The image in the merchant's entered information is converted into original image text, and invalid text in the original image text is deleted to obtain the image text.

4. The data processing method according to claim 1, characterized in that, The noise-adding weights obtained based on the historical data of the same category include: The standard deviation of the historical data of the same category is used as weighted noise.

5. The method for processing data according to claim 1, characterized in that, The historical data of the same category, combined with random noise of the same category, is used to establish new historical data of the same category, including: By combining a single historical data point of the same category with random noise of the same category, new historical data of the same category can be created.

6. The method for processing data according to claim 1, characterized in that, After constructing the same type of simulated historical data, the following is also included: Add the simulated historical data of the same type to the historical data of the same category; Then, based on the noise weights added to the historical data of the same category, random noise of the same category is generated to reconstruct simulated historical data of the same category.

7. The method for processing data according to claim 6, characterized in that, The step of generating random noise of the same category based on the noise weights of the historical data of the same category to reconstruct simulated historical data of the same category includes: Then, based on the noise weights of the historical data of the same category, random noise of the same category is generated to reconstruct simulated historical data of the same category. If the simulated historical data of the same type meets the preset construction conditions, then the construction of simulated historical data will stop.

8. The method for processing data according to claim 1, characterized in that, The process of training and building a credit rating model using multiple types of simulated historical data and the aforementioned multiple types of category historical data includes: Annotate the simulated historical data of the multiple categories and the category-specific historical data of the multiple categories; Credit parameters of the simulated historical data and the category historical data of the multiple categories after labeling are determined by preset rating conditions; To determine the credit parameters of the data, a credit rating model is trained and established.

9. A data processing apparatus, characterized in that, include: The segmentation module is used to divide merchant historical data into category historical data, which includes the following categories: merchant historical data, financial historical data, management historical data, and guarantee historical data. The merchant historical data includes: merchant identifier, merchant address, and merchant business hours; the financial historical data includes: invoice information, tax payment records, and financial statements; the management historical data includes order information, sales information, and social security payment records; the guarantee historical data includes guarantee amount, guarantee date, and guarantor. A generation module is configured to generate random noise of the same type based on the noise weights of historical data of the same category, wherein the noise weights are obtained based on the historical data of the same category; and generate random noise of the same type based on the noise weights of the historical data of the same category; and obtain two categories of historical data from the historical data of the same category; train a data feature model using the difference between one category of historical data and the two categories of historical data; input the category of historical data from the historical data of the same category into the data feature model to obtain the noise weights; and randomly select an adjustment parameter within the random noise range of the historical data of the same category; and use the product of the noise weights of the historical data of the same category and the adjustment parameter as the random noise of the same type. A construction module is used to construct simulated historical data of the same category by combining historical data of the same category with random noise of the same category; and to establish new historical data of the same category by combining historical data of the same category with random noise of the same category; if the characteristics of the historical data of the same category and the new historical data of the same category are consistent, then the new historical data of the same category is used as the simulated historical data of the same category; and if the new historical data of the same category falls within the statistical range of the historical data of the same category, then it is determined that the characteristics of the historical data of the same category and the new historical data of the same category are consistent; and the new historical data of the same category is used as the simulated historical data of the same category. A module is established to train and build a credit rating model using historical data from multiple types of simulations and historical data from multiple categories.

10. An electronic device for processing data, characterized in that, include: One or more processors; 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 method as described in any one of claims 1-8.

11. A computer-readable medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-8.

12. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-8.

Images