Financial data processing method, device and medium based on machine learning model

Abstract

The application discloses a financial data processing method, device and medium based on a machine learning model, the method comprising: obtaining financial data of at least one object having a financial link with a financial product; performing feature extraction on the financial data through a machine learning model to obtain a plurality of object features; configuring at least one first label matched with the plurality of object features for the at least one object based on the plurality of object features; sorting the at least one first label based on the size of the reference value corresponding to each of the at least one first label and determining the first labels arranged in the top pre-set number of positions; configuring a second label used for comprehensively describing the first labels arranged in the top pre-set number of positions based on the first labels arranged in the top pre-set number of positions, and associating and outputting the object and the matched second label. In the foregoing manner, the application can mine the financial data to obtain the value of a user corresponding to the financial data in essence.

Description

Technical Field

[0001] This application relates to the field of financial technology, and in particular to financial data processing methods, devices and media based on machine learning models. Background Technology

[0002] With the application and development of technologies such as big data, blockchain, and artificial intelligence in the financial industry, the deep integration and close cooperation between finance and technology has become a trend in modern finance. This combination of finance and technology allows financial companies to improve service efficiency and reduce costs, while also providing better financial services, increasing service frequency, and expanding the scale of the financial services market.

[0003] However, financial companies often face the challenge of processing large amounts of financial data. Existing financial data processing methods merely scratch the surface of the data and fail to extract the actual value of the users corresponding to the financial data, thus failing to meet the needs of financial companies for precise marketing. Summary of the Invention

[0004] The main technical problem addressed in this application is to provide a financial data processing method, device, and medium based on machine learning models, which can mine financial data to obtain the value of users corresponding to the financial data.

[0005] To address the aforementioned technical problems, this application provides a technical solution: a financial data processing method based on a machine learning model. This method includes: acquiring financial data of at least one object that has a financial link with a financial product; extracting features from the financial data using a machine learning model to obtain several object features; configuring at least one first label for each object that matches the several object features; configuring corresponding reference values ​​for each first label; sorting the at least one first label based on the magnitude of its corresponding reference values ​​and determining the first label ranked in the top preset position; configuring a second label for a comprehensive description of the first label ranked in the top preset position for the matching object based on the first label ranked in the top preset position, and associating the output object with the matching second label.

[0006] To solve the above-mentioned technical problems, another technical solution adopted in this application is: to provide a computer device, which includes a processor, a memory and a communication circuit; the memory and the communication circuit are coupled to the processor, the memory stores a computer program, and the processor is able to execute the computer program to implement the financial data processing method based on the machine learning model provided in this application as described above.

[0007] To solve the above-mentioned technical problems, another technical solution adopted in this application is to provide a computer-readable storage medium that stores a computer program that can be executed by a processor to implement the financial data processing method based on a machine learning model provided in this application as described above.

[0008] The beneficial effects of this application are as follows: Unlike existing technologies, after obtaining financial data of at least one object with a financial link to a financial product, a machine learning model can be used to extract features from the financial data to obtain several object features. Based on these object features, at least one first label matching the object features can be assigned to at least one object. Since financial data includes various object features, extracting different object features through a machine learning model allows for the assignment of first labels to objects corresponding to the financial data, facilitating the segmentation and management of at least one object corresponding to the financial data. Then, corresponding reference values ​​can be assigned to each first label, and the at least one first label is sorted based on the magnitude of its respective reference value, determining the first label ranking within the first preset position. By assigning corresponding reference values ​​to the first labels and sorting them according to the magnitude of the reference values, the importance of different first labels can be more intuitively understood. Finally, based on the first labels ranking within the first preset position, a second label can be assigned to the matching object to comprehensively describe the first labels ranking within the first preset position, and the output object and the matching second label can be associated. The above scheme can integrate the more important first tags and configure second tags for objects to comprehensively describe the more important first tags. Ultimately, it can segment at least one object through the second tags, thereby mining financial data to obtain the characteristics of objects that actually correspond to the financial data. This is beneficial for conducting precise marketing to different objects based on the output results. Attached Figure Description

[0009] Figure 1 This is a flowchart illustrating an embodiment of the financial data processing method based on a machine learning model in this application;

[0010] Figure 2 This is a timing diagram illustrating an embodiment of the financial data processing method based on a machine learning model in this application;

[0011] Figure 3 This is a schematic diagram of the circuit structure of an embodiment of the computer device of this application;

[0012] Figure 4 This is a schematic diagram of the circuit structure of an embodiment of the computer-readable storage medium of this application. Detailed Implementation

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0014] With the application and development of technologies such as big data, blockchain, and artificial intelligence in the financial industry, the deep integration and close cooperation between finance and technology has become a trend in modern finance. This combination of finance and technology allows financial companies to improve service efficiency and reduce costs, while also providing better financial services, increasing service frequency, and expanding the scale of the financial services market.

[0015] Through long-term research, the inventors discovered that financial companies often face the challenge of processing massive amounts of financial data. Existing financial data processing methods merely scratch the surface, failing to extract the intrinsic value of the users corresponding to the data, thus failing to meet the needs of financial companies for precise marketing. For example, if the financial data is basic business registration information, this information may be incomplete. Consequently, the target information identified based on this basic data, and the marketing strategies and business scenarios developed for those targets, may not match the actual situation of the targets, leading to the loss of valuable leads and hindering business expansion.

[0016] The following embodiments of this application can be applied to banking systems, specifically to scenarios where banking systems process financial data.

[0017] like Figure 1 As shown, the embodiment of the financial data processing method based on a machine learning model in this application may include the following steps: S100: Obtain financial data of at least one object that has a financial link with a financial product. S200: Extract features from the financial data using a machine learning model to obtain several object features. S300: Configure at least one first label for at least one object that matches the several object features. S400: Configure corresponding reference values ​​for each first label, sort the at least one first label based on the magnitude of the reference values ​​corresponding to each first label, and determine the first label that ranks in the top preset position. S500: Configure a second label for the matching object based on the first label that ranks in the top preset position to comprehensively describe the first label that ranks in the top preset position, and associate the output object with the matching second label.

[0018] After acquiring financial data of at least one object linked to a financial product, a machine learning model can be used to extract features from the financial data to obtain several object features. Based on these object features, at least one first label matching the object features can be assigned to at least one object. Since financial data includes various object features, extracting different object features through a machine learning model allows for the assignment of first labels to objects corresponding to the financial data, facilitating the segmentation and management of at least one object corresponding to the financial data. Then, corresponding reference values ​​can be assigned to each first label, and the at least one first label is sorted based on the magnitude of its respective reference value, determining the first label ranking within a predetermined number of positions. By assigning corresponding reference values ​​to the first labels and sorting them according to their magnitude, the importance of different first labels can be more intuitively understood. Finally, based on the first labels ranking within the predetermined number of positions, a second label can be assigned to the matching objects to comprehensively describe the first labels ranking within the predetermined number of positions, and the output object and the matching second label can be associated. The above scheme can integrate the more important first tags and configure second tags for objects to comprehensively describe the more important first tags. Ultimately, it can segment at least one object through the second tags, thereby mining financial data to obtain the characteristics of objects that actually correspond to the financial data. This is beneficial for conducting precise marketing to different objects based on the output results.

[0019] The following provides a detailed description of this implementation, such as... Figure 2 As shown, this embodiment may include:

[0020] S100: Obtain financial data of at least one object that has a financial link with the financial product.

[0021] Financial products can include all products within the banking system that can be used for financial transactions. Specifically, financial products can include products that are held in custody by the banking system.

[0022] The object can include objects that have a financial link with the financial product. Specifically, a financial link can include a link between products or funds.

[0023] Financial data can include data generated from financial transactions corresponding to an object; specifically, it can include basic business data. In acquiring financial data, large amounts of data can be obtained through big data platforms, and then, by identifying financial links with financial products, at least one object corresponding to the financial data can be determined.

[0024] In one implementation, the following steps included in S100 can be used as a reference for how to obtain financial data of at least one object that has a financial link with the financial product:

[0025] S110: Obtain financial data based on a pre-set data platform.

[0026] The pre-set data platform can include a platform connected to the banking system. The banking system can obtain financial data from the pre-set data platform. For example, the pre-set data platform can be a business data platform, from which the banking system can obtain basic business data.

[0027] S120: Identify at least one object that corresponds to financial data and has a financial link with a financial product.

[0028] After acquiring financial data from a pre-set data platform, at least one object corresponding to the financial data and linked to a financial product can be identified. Specifically, the banking system can determine at least one object that actually corresponds to the financial data through the financial links between financial products. Since the objects directly represented by the financial data acquired from the pre-set data platform may not actually be objects with financial links—for example, if the banking system acquires basic business registration data, and the object corresponding to this basic business registration data is not a physical enterprise but a product—then the financial links between the banking system's financial products and the acquired basic business registration data can be used to determine the physical enterprise that actually corresponds to the basic business registration data. This allows for the mining of financial data to extract the value of the object that actually corresponds to the financial data, thereby meeting the needs of financial companies for precision marketing.

[0029] In order to improve the efficiency of data processing, the banking system may acquire a large amount of financial data at the same time when obtaining financial data from a pre-set data platform. Therefore, the acquired financial data may correspond to multiple objects.

[0030] By identifying at least one object that corresponds to financial data and has a financial link with financial products, it is possible to deeply trace the financial chain of financial data. This is beneficial for comprehensively covering the panorama of objects associated with financial data, enabling the banking system to gain a more comprehensive understanding of the object, which is conducive to providing more precise marketing strategies for the object, and also facilitates the banking system to expand its customer acquisition channels.

[0031] In one implementation, after acquiring financial data of at least one object that has a financial link with the financial product, the following steps may be included:

[0032] S130: Preprocessing financial data.

[0033] Since this embodiment processes financial data based on a machine learning model, after acquiring the financial data, it can be input into the machine learning model for processing. To ensure that the financial data input into the machine learning model remains clean, it can be preprocessed before input to improve the accuracy and stability of the machine learning model's processing of financial data.

[0034] In one implementation, the following steps included in S130 can be used as a reference for how to preprocess financial data:

[0035] S131: Select at least a portion of financial data from the financial data that meets the preset conditions corresponding to the machine learning model.

[0036] Preset conditions can be conditions pre-defined in a machine learning model. Because different scripts can be used to train machine learning models during different data processing of financial data, the models can adapt to different processing modes. Specifically, preset conditions can be financial data with high relevance to the modeling of the machine learning model, used as input to the model.

[0037] In other words, during the preprocessing of financial data, at least a portion of the financial data with high modeling relevance to the current machine learning model can be selected and used as input to the machine learning model.

[0038] S132: Perform data cleaning on at least a portion of the financial data.

[0039] After filtering the financial data to obtain at least a portion of the filtered financial data, data cleaning can be performed on at least a portion of the financial data to ensure that at least a portion of the financial data is clean data before being included in the model.

[0040] In one implementation, the following steps included in S132 can be referenced regarding how to perform data cleaning on at least a portion of the financial data:

[0041] S1321: Fill in missing values ​​for at least a portion of the financial data.

[0042] S1322: Uniformly encode at least a portion of the filled financial data.

[0043] S1323: Binarize at least a portion of the uniformly coded financial data.

[0044] During the data cleaning process for at least a portion of financial data, missing values ​​can be imputed first. Since financial data obtained from a pre-set data platform is unprocessed, incomplete data may exist during the data retrieval process. Therefore, it is necessary to impute these missing parts to maintain the integrity of the financial data. Specifically, this can be achieved by adding supplementary code values.

[0045] After filling in the missing financial data, at least a portion of the filled financial data can be uniformly encoded. Since financial data obtained from a pre-defined data platform may have different encoding methods, and even after filling in the missing financial data using supplementary code values, there may still be differences in encoding methods, uniformly encoding at least a portion of the filled financial data can be used. This ensures that at least a portion of the financial data uses a unified encoding method when input into a machine learning model, thereby facilitating the processing of the financial data.

[0046] After encoding at least a portion of the financial data using a unified encoding method, the encoded financial data can be binarized, which facilitates better analysis and processing of the financial data.

[0047] After imputing missing values, standardizing encoding, and binarizing at least some financial data, clean financial data can be obtained, which can then be fed into machine learning models for further processing.

[0048] S200: Several object features are obtained by extracting features from financial data through machine learning models.

[0049] After inputting financial data into a machine learning model, the model can extract features from the financial data to obtain several object features. Extracting these object features facilitates the classification of financial data and the assignment of different labels, thereby simplifying the processing of the financial data.

[0050] In one implementation, the steps included in S200 for extracting features from financial data using a machine learning model to obtain several object features can be referenced:

[0051] S210: Several object features are obtained by extracting features based on the data structure of financial data through a machine learning model.

[0052] Data structures can include the internal structure and patterns of data. Object features can include features extracted from the data structure of financial data. For example, object features can include numerical features, categorical features, and spatial features. In the process of feature extraction using machine learning models, the machine learning model can extract numerical features, categorical features, and spatial features based on the data structure of the input financial data.

[0053] In one implementation, the steps in S210 for extracting features of several objects based on the data structure of financial data using a machine learning model can be referenced:

[0054] S211: The data structure of financial data is determined by using a preset mathematical calculation method.

[0055] In the process of feature extraction based on the data structure of financial data using machine learning models, a pre-defined mathematical calculation method can be used to determine the data structure of the financial data. Specifically, the pre-defined mathematical calculation method can be a pre-set method in the machine learning model used to explore the structure and patterns of the financial data. For example, the pre-defined mathematical calculation method can be the sample mean, variance, quantiles, kurtosis, etc.

[0056] S212: Based on the data structure, feature extraction is performed to obtain several object features.

[0057] After determining the data structure of financial data using a pre-defined mathematical calculation method, feature extraction can be performed based on the determined data structure to obtain several object features.

[0058] S300: Configure at least one first label on at least one object based on several object characteristics, which matches the several object characteristics.

[0059] The first label can include labels configured for an object based on its characteristics. Specifically, since each object can correspond to multiple object characteristics, multiple first labels can be configured for each object based on all the object characteristics corresponding to that object. By configuring first labels that match the object characteristics, on the one hand, several objects and their corresponding financial data can be classified based on the first labels, which is beneficial for the processing of financial data; on the other hand, it is beneficial to configure more refined second labels for objects based on the first labels.

[0060] In one implementation, the following steps in S300 can be used to describe how to configure at least one first label that matches at least one object based on several object characteristics:

[0061] S310: Based on the preset requirements and the object characteristics of each object, configure at least one first label for each object that matches its respective object characteristics.

[0062] Pre-defined requirements may include business requirements corresponding to the machine learning model. In configuring at least one first label for several objects, each object can be configured with at least one first label matching its respective object characteristics based on its own object features and business requirements.

[0063] For example, if the object features extracted from the financial data corresponding to object A include numerical features, category features, and spatial features, then based on the numerical features of object A, two primary labels, namely, the amount of foreign investment and the frequency of foreign investment, can be assigned to object A. These two primary labels can be associated with the total amount and quantity of foreign investment in the financial data corresponding to object A in the past 3 months, the past 6 months, or the past year.

[0064] For example, a first label can be configured for whether object A is a government-guided fund based on the numerical characteristics of object A, and this first label can be associated with whether the financial product corresponding to object A is a government-guided fund.

[0065] For example, based on the category characteristics of object A, primary labels such as judicial risk level, operational risk level, financing background level, enterprise vitality level, outward investment level, investment region, and industry preference can be assigned to object A. The judicial risk level and operational risk level can be correlated with the number of risk data entries registered by object A in the past year. The financing background level can be correlated with object A's corresponding controlling background, shareholding background, investment background, and financing background. The enterprise vitality level can be correlated with object A's corresponding operating status, the number and amount of outward investments, and the number and amount of investments received. The outward investment level can include conservative, relatively conservative, stable, relatively aggressive, and aggressive investment. The investment region and industry preference can be correlated with the region or industry distribution of object A's invested enterprises that have the highest proportion.

[0066] For example, a first label can be configured for object A based on its spatial characteristics, and this first label can be associated with object A's registered amount and its region.

[0067] S400: Configure corresponding reference values ​​for each first label, sort at least one first label based on the magnitude of the reference values ​​corresponding to each first label, and determine the first label that is arranged in the first preset position.

[0068] The reference value can be a numerical value configured based on the importance of each first label. That is, the higher the reference value for a first label, the more important that first label can be considered. This importance can be related to the business requirements corresponding to the machine learning model.

[0069] After configuring at least one first label that matches the characteristics of each object, a corresponding reference value can be configured for each first label. Then, at least one first label can be sorted based on the size of the reference value corresponding to each first label and the first label ranked in the first preset position can be determined.

[0070] In one implementation, the following steps included in S400 can be used as a reference for configuring the corresponding reference values ​​for each first label:

[0071] S410: Use a filtering method to configure corresponding reference values ​​for each object's first tag.

[0072] In configuring reference values ​​for each primary tag, a filtering method can be used to assign reference values ​​to each primary tag corresponding to each object. That is, a filtering method can be used to score each primary tag corresponding to each object based on its importance, thus obtaining a reference value for each primary tag. For example, if the primary tag "Amount of Outward Investment" is of higher importance among multiple primary tags, a reference value of 80 points can be assigned to it; if the primary tag "Frequency of Outward Investment" is less important than "Amount of Outward Investment," a reference value of 75 points can be assigned to it.

[0073] In one implementation, after configuring the corresponding reference values ​​for each first tag, the following steps may be included:

[0074] S420: Sort all first labels corresponding to each object based on the size of the reference value corresponding to each first label.

[0075] After configuring corresponding reference values ​​for each first label, at least one first label can be sorted based on the magnitude of its respective reference value. Specifically, all first labels corresponding to each object can be sorted based on the magnitude of their respective reference values. Since each object can have multiple first labels, all first labels corresponding to each object need to be sorted according to the magnitude of their reference values. For example, if the first labels corresponding to object A include external investment amount, external investment frequency, and enterprise vitality level, and the corresponding reference values ​​are 80, 75, and 60 respectively, then the first labels are sorted based on the magnitude of their reference values ​​as external investment amount, external investment frequency, and enterprise vitality level. If the first labels corresponding to object B include external investment amount, financing background level, and operational risk level, and the corresponding reference values ​​are 80, 70, and 85 respectively, then the first labels are sorted based on the magnitude of their reference values ​​as operational risk level, external investment amount, and financing background level.

[0076] S430: Determine the first label that ranks first by a preset number of positions among all the first labels corresponding to each object.

[0077] After sorting all the first tags corresponding to each object, the first tag that ranks in the top preset position among all the first tags corresponding to each object can be determined. By sorting all the first tags corresponding to each object and determining the first tags that rank in the top preset position, the most important first tags among all the first tags can be identified more intuitively. This is beneficial for more quickly dividing several objects, and it also allows for the assignment of second tags to objects based on the first tags that rank in the top preset position.

[0078] For example, if the preset number of digits is 3, then after sorting all the first tags corresponding to object A according to the reference value, it can be determined that the top 3 of all the first tags corresponding to object A are the amount of foreign investment, the frequency of foreign investment, and the enterprise vitality level.

[0079] S500: Based on the first label arranged in the first preset number of positions, configure the matching object with a second label to comprehensively describe the first label arranged in the first preset number of positions, and associate the output object with the matching second label.

[0080] The second label may include a label used to comprehensively describe the first label ranked in the first preset position. For example, if the first label ranked in the first preset position is the amount of foreign investment, the frequency of foreign investment, and the enterprise vitality level, then the second label used to comprehensively describe these three first labels can be the investment activity level and the amount, and this second label is associated with the object's amount of foreign investment, the frequency of foreign investment, and the enterprise vitality level.

[0081] After determining the first label that ranks in the first preset position, a second label can be configured for the matching object based on the first label that ranks in the first preset position to comprehensively describe the first label that ranks in the first preset position. Finally, the machine learning model is used to associate the output object with the matching second label.

[0082] In one implementation, regarding how to configure a second label for comprehensively describing the first label arranged in the first preset position to the matching object based on the first label arranged in the first preset position, and to associate the output object with the matching second label, the following steps included in S500 can be referred to:

[0083] S510: Using an unsupervised learning model, financial data that match the objects corresponding to the first label ranked in the first preset position are clustered according to intervals.

[0084] S520: Configure a second label for the object based on the clustering results.

[0085] S530: Associates the output object with a matching second label through a machine learning model.

[0086] In configuring a second label for an object, an unsupervised learning model can be used to cluster the financial data matching all objects with the first label ranked in the top preset position according to intervals. Then, the second label is configured for the object based on the clustering results. In other words, a machine learning model uses unsupervised learning to cluster financial data matching objects with the first label ranked in the top preset position together based on high similarity, and then classifies them hierarchically according to intervals. The second label can then be configured for the object based on the clustering results. After configuring the second label for the object, the machine learning model can be used to associate the output object with its matching second label.

[0087] For example, if the first label for object A, ranked in the top three, is the amount of outward investment, the frequency of outward investment, and the enterprise vitality level, the machine learning model, using unsupervised learning, clusters the financial data corresponding to these three metrics according to intervals. Based on the clustering results, it can assign a second label to object A that indicates high investment activity and a large amount of investment. In other words, after inputting the financial data corresponding to object A into the machine learning model, the model can output object A along with the second label indicating high investment activity and a large amount of investment.

[0088] In summary, after obtaining financial data of at least one object with a financial link to a financial product, this embodiment can extract features from the financial data using a machine learning model to obtain several object features. Based on these object features, at least one first label matching the object features can be assigned to at least one object. Since financial data includes various object features, extracting different object features using a machine learning model allows for the assignment of first labels to objects corresponding to the financial data, facilitating the segmentation and management of at least one object corresponding to the financial data. Then, corresponding reference values ​​can be assigned to each first label, and the at least one first label is sorted based on the magnitude of its respective reference value, determining the first label that ranks in the top preset position. By assigning corresponding reference values ​​to the first labels and sorting them according to the magnitude of the reference values, the importance of different first labels can be more intuitively understood. Finally, based on the first labels ranked in the top preset position, a second label can be assigned to the matching object to comprehensively describe the first labels ranked in the top preset position, and the output object and the matching second label can be associated. The above scheme can integrate the more important first tags and configure second tags for objects to comprehensively describe the more important first tags. Ultimately, it can segment at least one object through the second tags, thereby mining financial data to obtain the characteristics of objects that actually correspond to the financial data. This is beneficial for conducting precise marketing to different objects based on the output results.

[0089] like Figure 3 As shown in the embodiments of the computer equipment described in this application, the computer equipment 100 can be the aforementioned banking system. The computer equipment 100 may include a processor 110, a memory 120, and a communication circuit 130.

[0090] The memory 120 is used to store computer programs and may be RAM (Read-Only Memory), ROM (Random Access Memory), or other types of storage devices. Specifically, the memory may include one or more computer-readable storage media, which may be non-transitory. The memory may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In some embodiments, the non-transitory computer-readable storage media in the memory is used to store at least one line of program code.

[0091] Processor 110 is used to control the operation of computer device 100. Processor 110 may also be referred to as CPU (Central Processing Unit). Processor 110 may be an integrated circuit chip with signal processing capabilities. Processor 110 may also be a general-purpose processor, digital signal processor (DSP), application-specific integrated circuit (ASIC), off-the-shelf programmable gate array (FPGA), or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component. The general-purpose processor may be a microprocessor, or processor 110 may be any conventional processor.

[0092] The processor 110 is used to execute the computer program stored in the memory 120 to implement the financial data processing method based on the machine learning model described in the embodiments of the financial data processing method based on the machine learning model of this application.

[0093] The computer device 100 may also include a communication circuit 130, which is a device or circuit used by the computer device 100 to communicate with external devices, so that the processor 110 can interact with external devices via the communication circuit 130.

[0094] For a detailed description of the functions and execution processes of each functional module or component in the computer device embodiments of this application, please refer to the description in the embodiments of the financial data processing method based on machine learning models of this application, which will not be repeated here.

[0095] In the several embodiments provided in this application, it should be understood that the disclosed computer device 100 and the financial data processing method based on machine learning models can be implemented in other ways. For example, the embodiments of the computer device 100 described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0096] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0097] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0098] See Figure 4 If the integrated units described above are implemented as software functional units and sold or used as independent products, they can be stored in computer-readable storage medium 200. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions / computer programs to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods of various embodiments of this invention. The aforementioned storage medium includes various media such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks, as well as electronic terminals such as computers, mobile phones, laptops, tablets, and cameras that have the aforementioned storage media.

[0099] The execution process of program data in computer-readable storage media can be described with reference to the above embodiments of the financial data processing method based on machine learning models in this application, and will not be repeated here.

[0100] The above description is merely an embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A method for processing financial data based on a machine learning model, characterized in that, include: Obtain financial data of at least one object that has a financial link with a financial product; The machine learning model is used to extract features from the financial data to obtain several object features; Based on the aforementioned object features, configure at least one first label for the at least one object that matches the aforementioned object features; Configure a corresponding reference value for each of the first tags, sort the at least one first tags based on the magnitude of the reference value corresponding to each of the at least one first tags, and determine the first tags that are arranged in the first preset number of positions; Based on the first label arranged in the first preset number of positions, configure the matching object with a second label that comprehensively describes the first label arranged in the first preset number of positions, and output the object and the matching second label in association; The step of configuring at least one first label matching the plurality of object features for at least one object includes: Based on preset requirements and the object characteristics corresponding to each object, configure at least one first tag for each object that matches its respective object characteristics; The step of configuring corresponding reference values ​​for each of the first tags includes: A filtering method is used to configure the corresponding reference value for each of the first tags corresponding to each of the objects; The step of sorting the at least one first label based on the magnitude of the reference value corresponding to each of the at least one first label and determining the first label arranged in the first preset position includes: Sort all the first tags corresponding to each object based on the size of the reference value corresponding to each first tag; Determine the first label that is ranked at the first preset position among all the first labels corresponding to each object; The configuration of the object based on the first tag arranged in the first preset position to comprehensively describe the second tag arranged in the first preset position, and the associated output of the object and the matching second tag, includes: An unsupervised learning model is used to cluster the financial data that match the objects corresponding to the first labels ranked in the first preset number of positions according to intervals. The second label is configured for the object based on the clustering results; The machine learning model associates and outputs the object with the matching second label.

2. The method according to claim 1, characterized in that, The process of extracting features from the financial data using the machine learning model yields several object features, including: The machine learning model extracts features from the financial data based on the data structure to obtain several object features.

3. The method according to claim 2, characterized in that, The feature extraction based on the data structure of the financial data yields several object features, including: The data structure of the financial data is determined using a preset mathematical calculation method; Based on the data structure, feature extraction is performed to obtain several object features.

4. The method according to claim 1, characterized in that, The acquisition of financial data of at least one object that has a financial link with a financial product includes: The financial data is obtained based on a pre-set data platform; Identify at least one of the objects that corresponds to the financial data and has a financial link with the financial product.

5. The method according to claim 1, characterized in that, After obtaining financial data of at least one object that has a financial link with the financial product, the process includes: The financial data is preprocessed.

6. The method according to claim 5, characterized in that, The preprocessing of the financial data includes: Filter out at least a portion of the financial data that meets the preset conditions corresponding to the machine learning model from the financial data; Data cleaning is performed on at least a portion of the financial data.

7. The method according to claim 6, characterized in that, The data cleaning of at least a portion of the financial data includes: Impute missing values ​​for at least a portion of the financial data; At least a portion of the financial data after filling is uniformly encoded; Binarization is performed on at least a portion of the financial data after unified encoding.

8. A computer device, comprising: It includes a processor, a memory, and a communication circuit; the memory and the communication circuit are coupled to the processor, the memory stores a computer program, and the processor is capable of executing the computer program to implement the financial data processing method based on a machine learning model as described in any one of claims 1-7.

9. A computer readable storage medium, characterized in that, The system contains a computer program that can be executed by a processor to implement the financial data processing method based on a machine learning model as described in any one of claims 1-7.

Images