Data processing method and device, electronic equipment, storage medium and product

Abstract

The invention discloses a data processing method and device, electronic equipment, a storage medium and a product. The method comprises the steps of receiving multiple pieces of to-be-processed data of at least one target enterprise; for the multiple pieces of to-be-processed data, in response to an event that an early warning tag of the to-be-processed data is inconsistent with a preset tag, obtaining target data of the to-be-processed data under multiple target dimensions; inputting the target data into an early warning model, and outputting an early warning attribute value corresponding to the target data; determining a target risk attribute of the target data based on the early warning attribute value of the target data and the target function; and when the target risk attribute meets a preset condition, outputting target early warning information representing the risk attribute of a target enterprise to which the to-be-processed data belongs. According to the technical scheme of the invention, through a dual automatic confirmation mechanism of the preset tag and the early warning model, the manual confirmation amount of the early warning event for the target enterprise is greatly reduced, and the confirmation efficiency, real-time performance and accuracy of the early warning information are improved.

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of big data technology, and in particular, to a data processing method, device, electronic device, storage medium, and product. Background Art

[0002] With the development of the global economy, banks and other financial institutions are facing major development opportunities and high-risk challenges. Among them, loan risk is the main risk in the banking operation process. In order to achieve sustainable development of the banking economy, it is necessary to strengthen credit risk management.

[0003] At present, after some enterprises borrow certain resources from a financial institution, they can process and analyze the relevant data of these enterprises to evaluate whether these enterprises can return the resources on time. If there is a risk of not being able to return the resources on time, an early warning signal will be generated. However, there are a large number of false early warning signals among these signals, and professional technicians are usually required to identify these early warning signals to determine whether these early warning signals are true.

[0004] However, with the substantial increase in post-loan warning signals, the manual confirmation method has high time and labor costs, and it is impossible for humans to process warning signals in a timely manner. There are technical problems such as low efficiency in post-loan warning signal confirmation, poor real-time performance, and large accuracy deviation. Summary of the invention

[0005] The embodiments of the present invention provide a data processing method, device, electronic device, storage medium and product to reduce the manual confirmation amount of warning events for target enterprises and improve the confirmation efficiency, real-time and accuracy of warning information.

[0006] In a first aspect, an embodiment of the present invention provides a data processing method, the method comprising:

[0007] receiving a plurality of to-be-processed data of at least one target enterprise;

[0008] For the plurality of data to be processed, in response to an event that the warning label of the data to be processed is inconsistent with the preset label, obtaining target data of the data to be processed under multiple target dimensions; wherein the multiple target dimensions are determined according to the data processing results of the historical sample data under all the dimensions to be selected;

[0009] Input the target data into a pre-trained warning model, and output a warning attribute value corresponding to the target data;

[0010] Determining a target risk attribute of the target data based on the warning attribute value of the target data and the target function;

[0011] When the target risk attribute meets a preset condition, target warning information of the data to be processed is output; wherein the target warning information is used to characterize the risk attribute of the target enterprise to which the data to be processed belongs.

[0012] In a second aspect, an embodiment of the present invention further provides a data processing device, the device comprising:

[0013] An enterprise data receiving module, used for receiving a plurality of to-be-processed data of at least one target enterprise;

[0014] A target data acquisition module is used to acquire target data of the data to be processed under multiple target dimensions in response to an event that the warning label of the data to be processed is inconsistent with the preset label, for the multiple data to be processed; wherein the multiple target dimensions are determined according to the data processing results of the historical sample data under all the dimensions to be selected;

[0015] A warning attribute value determination module is used to input the target data into a pre-trained warning model and output the warning attribute value corresponding to the target data;

[0016] A risk attribute determination module, used to determine the target risk attribute of the target data based on the warning attribute value of the target data and the target function;

[0017] The warning information determination module is used to output the target warning information of the data to be processed when the target risk attribute meets the preset conditions; wherein the target warning information is used to characterize the risk attribute of the target enterprise to which the data to be processed belongs.

[0018] In a third aspect, an embodiment of the present invention further provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the data processing method as described in any one of the embodiments of the present invention is implemented.

[0019] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a data processing method as described in any one of the embodiments of the present invention.

[0020] In a fifth aspect, an embodiment of the present invention further provides a computer program product, including a computer program, which, when executed by a processor, implements a data processing method as described in any one of the embodiments of the present invention.

[0021] In an embodiment of the present invention, multiple data to be processed of at least one target enterprise are received, and then, for multiple data to be processed, in response to an event that the warning label of the data to be processed is inconsistent with the preset label, the target data of the data to be processed under multiple target dimensions are obtained, wherein the multiple target dimensions are determined according to the data processing results of the historical sample data under all dimensions to be selected, thereby inputting the target data into the pre-trained warning model, outputting the warning attribute value corresponding to the target data, and further, based on the warning attribute value of the target data and the objective function, determining the target risk attribute of the target data, and then, when the target risk attribute meets the preset conditions, outputting the risk attribute target warning information for characterizing the target enterprise to which the data to be processed belongs. The technical solution of the present invention, through the dual automatic confirmation mechanism of preset labels and warning models, greatly reduces the amount of manual confirmation of warning events for target enterprises, and improves the confirmation efficiency, real-time and accuracy of warning information. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0023] Figure 1 is a flow chart of a data processing method provided according to an embodiment of the present invention;

[0024] Figure 2 is a schematic diagram of another data processing method provided according to an embodiment of the present invention;

[0025] Figure 3 is a schematic diagram of another data processing method provided according to an embodiment of the present invention;

[0026] Figure 4 is a structural schematic diagram of a data processing device provided according to an embodiment of the present invention;

[0027] Figure 5 The present invention is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.

[0029] It should be noted that similar numbers and letters represent similar items in the following figures. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance. The acquisition, storage, use, and processing of data in the technical solution of this application comply with the relevant provisions of national laws and regulations.

[0030] It should be noted that in the embodiments of the present application, certain software, components, models and other existing solutions in the industry may be mentioned, and they should be regarded as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of the present application, but it does not mean that the applicant has or will necessarily use the solution.

[0031] The acquisition, storage, use, and processing of data in the technical solution of this application comply with the relevant provisions of national laws and regulations.

[0032] Figure 1 It is a flow chart of a data processing method provided according to an embodiment of the present invention. This embodiment can be applied to situations where the authenticity of post-loan warning information of a certain enterprise needs to be confirmed. The method can be executed by a data processing device, and the data processing device can be implemented in the form of hardware and / or software.

[0033] like Figure 1 As shown, the method includes:

[0034] S110, receiving a plurality of to-be-processed data of at least one target enterprise.

[0035] In this embodiment, after some enterprises borrow certain resources from a financial institution, it is possible to evaluate whether these enterprises can return the resources on schedule. If there is a risk of not being able to return the resources on schedule, an early warning signal will be generated for these enterprises, and these enterprises can be target enterprises. The data to be processed is the early warning information contained in the early warning signal.

[0036] Specifically, after the target enterprise borrows a certain amount of resources from the target financial institution, the comprehensive risk monitoring and early warning system corresponding to the target financial institution can process and analyze the relevant data of the target enterprise in real time. When the comprehensive risk monitoring and early warning system feeds back an early warning signal, it indicates that the enterprise is at risk of not being able to return the resources as scheduled, and at this time, an early warning signal corresponding to the target enterprise is generated. However, after the early warning signal is generated, it is necessary to further determine whether the early warning is true. Therefore, when the early warning signal for the target enterprise is generated, the early warning information corresponding to the early warning signal is the data to be processed corresponding to the target enterprise, and the data to be processed can be obtained to determine whether the early warning signal corresponding to the target enterprise is true through further analysis and processing of the data to be processed.

[0037] S120. For a plurality of data to be processed, in response to an event that a warning label of the data to be processed is inconsistent with a preset label, obtaining target data of the data to be processed under a plurality of target dimensions.

[0038] In this embodiment, multiple target dimensions are determined based on the data processing results of historical sample data under all dimensions to be selected. The target data is the specific data content of the target enterprise under the target dimension. Among them, the historical sample data includes the data corresponding to each dimension to be selected for multiple enterprises in the historical period, the corresponding early warning signals, and the processing results corresponding to the early warning signals. Since the number of dimensions to be selected is large, not all data of the dimensions to be selected contribute to the confirmation processing results of the early warning signals. Therefore, all dimensions to be selected can be screened in advance, and a part of the dimensions to be selected can be determined as target dimensions. In this way, the accuracy of the early warning signal processing results can be guaranteed, and the amount of data calculation can be reduced.

[0039] Among them, the dimensions to be selected include but are not limited to: enterprise industrial and commercial data dimension, enterprise credit data dimension, enterprise person in charge credit data dimension, enterprise tax data dimension, enterprise violation fine data dimension, enterprise and enterprise actual controller account information data dimension in financial institutions, deposit account transaction information data dimension, loan information data dimension, business settlement data dimension, credit limit usage data dimension, loan overdue data dimension, financial report indicator data dimension, wage payment data dimension, tax payment data dimension, intra-bank business settlement change dimension, credit limit usage information change dimension, financial report indicator change dimension, industrial and commercial information change dimension, etc.

[0040] In this embodiment, the data to be processed includes a warning label. The warning label is used to characterize the reasons why a target enterprise has overdue risks. For example, the warning label may include: money laundering risk label, fraud risk label, judicial freezing risk label, blacklist label. In addition, the warning label may also include special attribute information corresponding to the target enterprise. For example, the warning label may be a whitelist label.

[0041] The preset label is a pre-set label type. For example, the preset label may include a first warning label under a first warning type and a second warning label under a second warning type. The first warning type indicates that the warning signal is true, and the first warning label may be at least one of a money laundering risk label, a fraud risk label, a judicial freezing risk label, and a blacklist label. The second warning type indicates that the warning signal is true and there is no risk, and the second warning label may be a whitelist label.

[0042] Specifically, if a target enterprise has multiple data to be processed, the processing process for each data to be processed is the same, and any one of the data to be processed is used as an example for illustrative description. If the warning label corresponding to the data to be processed is inconsistent with the first warning label, and the warning label is also inconsistent with the second warning label, it means that it is impossible to confirm that the data to be processed belongs to the first warning type or the second warning type. At this time, it is necessary to further determine which warning result the data to be processed belongs to. Therefore, the target data of the data to be processed under multiple target dimensions can be obtained to further determine the warning result corresponding to the data to be processed based on the target data.

[0043] S130: Input the target data into the pre-trained warning model, and output the warning attribute value corresponding to the target data.

[0044] Among them, the early warning model is a pre-trained neural network model. The early warning attribute value is used to characterize the credibility of the early warning signal. For example, the larger the early warning attribute value, the higher the credibility of the target enterprise's risk of overdue resource return. Conversely, the smaller the early warning attribute value, the lower the credibility of the target enterprise's risk of overdue resource return.

[0045] In this embodiment, the target data is input into a pre-trained warning model. After the warning model processes and analyzes the target data under the target dimension, the warning attribute value corresponding to the target data can be output.

[0046] S140: Determine the target risk attribute of the target data based on the warning attribute value of the target data and the target function.

[0047] The target risk attribute can be concretely represented by the risk attribute value. The objective function is a pre-set attribute value conversion function.

[0048] Specifically, in order to accurately convert the warning attribute value into a risk attribute value for quantifying the overdue risk of the target enterprise, an objective function can be pre-constructed. The objective function is determined in the following manner: based on the first preset warning attribute value, a first warning reference value and a second warning reference value are determined; based on the first warning reference value and the first preset risk attribute value corresponding to the first warning reference value, a first equation to be solved is constructed; based on the second warning reference value and the second preset risk attribute value corresponding to the second warning reference value, a second equation to be solved is constructed; based on the first equation to be solved and the second equation to be solved, a system of linear equations of two variables to be solved is constructed to solve the system of linear equations of two variables to be solved and obtain the first coefficient value and the second coefficient value; based on the first coefficient value and the second coefficient value, an objective function is constructed.

[0049] Among them, the first preset warning attribute value is a pre-set value. The first warning reference value is a reference value determined according to the first preset warning attribute value. The second warning reference value is twice the first warning reference value. The first preset risk attribute value represents the ideal value corresponding to the risk attribute value when the warning reference value is the first warning reference value. The second preset risk attribute value represents the ideal value corresponding to the risk attribute value when the warning reference value is the second warning reference value.

[0050] For ease of understanding, a specific example is used to illustrate how to determine the objective function. For example, if the first preset warning attribute value is 80%, the first warning reference value can be expressed as A1=ln(80% / 1-80%), the second warning reference value can be expressed as A2=2×ln(80% / 1-80%), the first preset risk attribute value is B1=600, and the second preset risk attribute value is B2=630. On this basis, A1 and B1 can be substituted into the standard form of the linear function y=ax+b, and the first equation to be solved is 600=a×ln(80% / 1-80%)+b; A2 and B2 can be substituted into the standard form of the linear function y=ax+b, and the second equation to be solved is 630=a×ln[2×ln(80% / 1-80%)]+b; thus, the set of two-variable linear equations to be solved can be constructed as:

[0051]

[0052] Further, the two-variable linear equation system is solved to obtain the value of a as scale, which is the first coefficient value; the value of b as location, which is the second coefficient value. Thus, the objective function can be expressed as:

[0053]

[0054] In the formula, Score represents the risk attribute value, and P represents the warning attribute value.

[0055] In this embodiment, when the warning attribute value of the target data is obtained, the warning attribute value can be substituted into the objective function, and the objective function after the value is substituted is solved to obtain the target risk attribute value corresponding to the target data.

[0056] S150. When the target risk attribute meets the preset conditions, output the target warning information of the data to be processed.

[0057] Among them, the preset conditions are pre-set rules used to determine the target warning information corresponding to the target risk attributes. The target warning information is used to characterize the risk attributes of the target enterprise to which the data to be processed belongs. Specifically, the target warning information may include a true warning, a risk-free warning, or a transfer to the next step for confirmation. A true warning indicates that the target enterprise to which the data to be processed belongs has the risk of overdue resource return; a risk-free warning indicates that the target enterprise to which the data to be processed belongs does not have the risk of overdue resource return; a transfer to the next step for confirmation indicates that it is necessary to further determine whether the target enterprise has the risk of overdue resource return. Generally speaking, in this case, the data to be processed and related data corresponding to the target enterprise can be sent to the downstream business system so that professional technicians of the downstream business system can perform further confirmation tasks.

[0058] In this embodiment, the preset conditions may include a first preset threshold and a second preset threshold, and the second preset threshold is greater than the first preset threshold. When the target risk attribute value is obtained, the target risk attribute value, the second preset threshold and the first preset threshold may be further judged in numerical value to determine the target warning information of the data to be processed. For example, if the target risk attribute value is less than the first warning threshold, the target warning information of the data to be processed is that the warning is true; if the target risk attribute value is greater than the second warning threshold, the target warning information of the data to be processed is that the warning is risk-free; if the target risk attribute value is between the first warning threshold and the second warning threshold, the target warning information of the data to be processed is that the next step is transferred to confirm.

[0059] The technical solution provided in this embodiment receives multiple data to be processed of at least one target enterprise, and then, for multiple data to be processed, in response to the event that the warning label of the data to be processed is inconsistent with the preset label, obtains the target data of the data to be processed under multiple target dimensions, wherein the multiple target dimensions are determined according to the data processing results of the historical sample data under all the dimensions to be selected, thereby inputting the target data into the pre-trained warning model, outputting the warning attribute value corresponding to the target data, and further, based on the warning attribute value of the target data and the target function, determining the target risk attribute of the target data, and then, when the target risk attribute meets the preset conditions, outputting the risk attribute target warning information for characterizing the target enterprise to which the data to be processed belongs. The technical solution of the present invention, through the dual automatic confirmation mechanism of preset labels and warning models, greatly reduces the amount of manual confirmation of warning events of target enterprises, and improves the confirmation efficiency, real-time and accuracy of warning information.

[0060] Based on the above embodiments, this embodiment further describes how to determine the target dimension and the training process of the early warning model. Accordingly, the present invention proposes the following embodiments: Figure 21 is a flow chart of a data processing method provided according to an embodiment of the present invention. Figure 2 As shown, the data processing method comprises the following steps:

[0061] S210: Acquire historical sample data, wherein the historical sample data includes data content under at least one dimension to be selected.

[0062] In this embodiment, the dimensions to be selected include at least one of the target enterprise control user dimension, the target enterprise credit dimension, the historical transaction data dimension, the negative feedback dimension, the customer basic information dimension, the transaction account dimension, the loan dimension, and the change trend dimension under each dimension. Among them, the data content corresponding to the target enterprise control user dimension may include the resource data and personal credit information of the actual controller of the target enterprise in the financial institution; the data content corresponding to the target enterprise credit dimension may include the target enterprise credit data; the data content corresponding to the historical transaction data dimension may include the number of transactions of the target enterprise in the financial institution during the historical period, and the transaction records of each transaction. The data content corresponding to the negative feedback dimension may include the administrative penalty record information of the target enterprise. The data content corresponding to the customer basic information dimension may include the target enterprise's industrial and commercial information, tax payment information, business settlement information, financial report indicator information, wage payment, tax payment information, etc.; wherein the financial report indicator information may include profitability indicators, debt repayment indicators, and operating capacity indicators. Profitability indicators, for example, may include net sales rate, net asset rate, etc., which are used to evaluate the profitability and operating efficiency of the enterprise. Debt repayment indicators, for example, may include asset-liability ratio, current ratio, etc., which reflect the ability of the enterprise to repay long-term and short-term debts. Operating capacity indicators, for example, may include accounts receivable turnover rate, inventory turnover rate, etc., to measure the asset management and operational efficiency of the enterprise. The data content corresponding to the transaction account dimension may include the transaction information of each deposit account corresponding to the target enterprise. The data content corresponding to the loan dimension may include the loan record information of each account corresponding to the target enterprise, such as credit limit usage information, loan overdue information, etc. The change trend dimension refers to the month-on-month data of various dimensional data. In this embodiment, the historical data information of the target enterprise is obtained from multiple different dimensions, which improves the comprehensiveness of the sample data. Moreover, the dimensions to be selected include the data dimensions of the time point variables, and also include the dimensions used to characterize the change trends of various dimensions. Some hidden feature data that are difficult to be characterized by time point variables alone can be extracted, which further improves the comprehensiveness of the sample data.

[0063] In this embodiment, a plurality of different target enterprises have generated a large amount of data content in the historical period. In order to determine the target dimension that can efficiently and accurately characterize the resource repayment situation of the target enterprise from the numerous dimensions to be selected, a large amount of historical sample data corresponding to the enterprises can be obtained.

[0064] S220: For all data contents under the same dimension to be selected, determine the missing rate corresponding to the dimension to be selected according to all data contents.

[0065] In this embodiment, historical sample data corresponding to a large number of enterprises are obtained, but there are cases where some enterprises have missing data in some dimensions to be selected. In order to eliminate dimensions to be selected with a large amount of missing data and improve the availability of the target dimension, the missing rate corresponding to each dimension to be selected can be determined first. Specifically, the specific implementation method of determining the missing rate corresponding to the dimension to be selected based on all data content includes: determining the missing rate corresponding to the dimension to be selected based on the number and total number of data contents under each dimension to be selected.

[0066] Exemplarily, historical sample data corresponding to 1,000 enterprises are obtained, and 900 enterprises have data content under the transaction account dimension, so the missing rate corresponding to the transaction account dimension is 10%.

[0067] S230. For the dimensions to be selected whose missing rate meets the preset conditions, determine the information value and stability of the dimensions to be selected according to the data content under the dimensions to be selected, and determine the target dimension based on the information value and stability.

[0068] The higher the value of the information value, the stronger the predictive ability of the feature for the target variable, that is, the more helpful the feature is in distinguishing different categories or predicting targets. Information value is used to help evaluate the contribution of independent variables to model prediction. The stability of the dimension to be selected can be evaluated by calculating the group stability index. Stability is used to measure the difference between the model prediction distribution and the actual distribution. By screening the dimensions to be selected through information value and stability parameters, the target dimension obtained not only has a large contribution rate to the model prediction, but also has stable data features. In this way, feature screening is performed before model training to improve the performance and accuracy of the model.

[0069] In this embodiment, the dimension to be selected whose missing rate is less than the first preset missing rate threshold can be determined as the dimension to be selected that meets the preset conditions. Specifically, the specific implementation method of determining the information value and stability of the dimension to be selected according to the data content under the dimension to be selected can include the following steps: for the dimension to be selected whose missing rate is less than the first preset missing rate threshold in the preset condition, retrieve the data content under the dimension to be selected, and analyze and process the data content to obtain the information value under the dimension to be selected; obtain the dimension to be selected whose information value is greater than the preset value threshold, and determine the stability of the dimension to be selected according to the data content of the dimension to be selected.

[0070] In this embodiment, a first missing rate threshold can be pre-set, and the dimensions to be selected whose missing rate is less than the missing rate threshold can be screened. For these screened dimensions to be selected, the information value corresponding to each dimension to be selected is determined. First, binning is required, that is, the data of all target enterprises in the current dimension to be selected is divided into multiple intervals (bins), and then the information value of the dimension to be selected is determined by a formula. The calculation formula of the information value is:

[0071]

[0072] Where IV represents information value, N is the number of bins, Pi(good) is the proportion of good samples in the i-th bin, and Pi(bad) is the proportion of bad samples in the i-th bin.

[0073] It should be noted that for any historical sample data corresponding to an enterprise, if an enterprise meets any of the following conditions: first, the enterprise name is on the blacklist, and second, the enterprise has a historical warning risk that is true, then the sample label corresponding to the enterprise can be a bad sample. Otherwise, the sample label corresponding to the enterprise is a good sample.

[0074] Based on the above formula, the information value corresponding to each dimension to be selected can be obtained, and then the dimension to be selected whose information value is greater than the preset value threshold can be determined as the dimension to be processed. Furthermore, the stability of each dimension to be processed can be determined, and the dimension to be processed whose stability is less than or equal to the preset stability threshold can be determined as the target dimension. In this embodiment, for the case where the data missing rate meets the preset conditions, the dimension to be selected is double-screened by the information value parameter and the stability parameter, thereby improving the efficiency and applicability of feature screening.

[0075] Specifically, the method of determining stability may include: first, dividing the historical sample data of all enterprises into training sets and validation sets; second, binning, dividing the data into multiple intervals (bins); third, calculating the distribution ratio of each interval: for the two sample sets, respectively calculating the data ratio in each interval; fourth, determining the stability value according to the stability calculation formula, the stability calculation formula is:

[0076]

[0077] Where PSI represents the stability value and N is the number of bins.

[0078] In actual application, if the number of target dimensions obtained by the above method is too small, the dimensions to be selected can be reselected so that the number of target dimensions can meet the preset number threshold to ensure the accuracy and effectiveness of the early warning model. The specific implementation method of feature dimension reselection may include: if the number of target dimensions is less than the preset number threshold, reselect the dimensions to be aggregated whose missing rate is less than the second preset missing rate threshold; aggregate the data content in the dimensions to be aggregated to obtain multiple clusters, and obtain at least two aggregated dimensions from each cluster; determine the missing rates corresponding to at least two aggregated dimensions, and calculate the information value of the aggregated dimensions whose missing rate is less than the first preset missing rate threshold, so as to determine whether the aggregated dimension is the target dimension based on the information value.

[0079] The second preset missing rate threshold is greater than the first preset missing rate threshold.

[0080] In this embodiment, if the number of target dimensions is less than a preset number threshold, the dimensions to be selected whose missing rates are between the first preset missing rate threshold and the second preset missing rate threshold can be determined as dimensions to be aggregated. Further, the data contents of these dimensions to be aggregated are aggregated, and after obtaining multiple clusters, at least two aggregate dimensions with the highest reference indicators are obtained from each cluster based on the reference indicators of the data dimensions contained in each cluster (for example, information value, important attributes, etc.). Further, the missing rate corresponding to each clustering dimension is determined based on the number and total number of existing data contents in each clustering dimension. Then, the information value of the aggregated dimension whose missing rate is less than the first preset missing rate threshold is determined. For each aggregated dimension, if the information value is greater than the preset value threshold, it is determined as the target dimension, otherwise, it is not determined as the target dimension.

[0081] On the basis of the above embodiment, after determining the target dimension, the data of each enterprise under the target dimension is determined from all historical sample data as training sample data for subsequent training of the early warning model to be trained, so as to obtain a trained early warning model.

[0082] Specifically, the specific implementation methods of training the early warning model may include:

[0083] S1. For historical sample data, perform redundancy removal on the historical sample data according to the target dimension to obtain updated historical sample data.

[0084] In this embodiment, since the historical sample data corresponding to each enterprise includes a large amount of data content and some repeated data content, the historical sample data can be de-redundant to remove the repeated data content. Subsequently, the data content corresponding to the target dimension is extracted to obtain the updated historical sample data.

[0085] S2. For at least two warning models to be trained, the data content under the target dimension in the historical sample data is input into the warning model to be trained, and the predicted warning attributes are output.

[0086] The model structures or adopted algorithms of at least two early warning models to be trained are different. For example, at least two early warning models to be trained can be neural network models with different structures.

[0087] In this embodiment, the training process of each early warning model to be trained is consistent, and any early warning model to be trained is used as an example for explanation. And for the historical sample data corresponding to each enterprise, it can be trained in the S2-S4 method to obtain the required early warning model.

[0088] The early warning model to be trained is a model in which the model parameters are initial parameters or default parameters. The predicted early warning attribute is the early warning attribute output after the data content under the target dimension in the historical sample data corresponding to the current enterprise is input into the early warning model to be trained.

[0089] It should be noted that the model parameters in the early warning model to be trained do not meet the expected requirements. Therefore, there are certain differences between the predicted early warning attributes output based on the model parameters at this time and the actual label attributes. Therefore, the corresponding error loss value can be determined based on the predicted early warning attributes and actual label attributes corresponding to the historical sample data of each enterprise.

[0090] In this embodiment, the early warning model to be trained may be a logistic regression algorithm or a LightGBM algorithm model, etc. It should be noted that the early warning attribute corresponding to the historical sample data can be obtained, that is, the specific model type is not specifically limited.

[0091] S3. Determine a loss value according to the predicted warning attributes and the label attributes in the historical sample data, so as to modify the model parameters in the warning model to be trained based on the loss value.

[0092] It should be noted that the training parameters can be set to default values ​​before training the early warning model to be trained. When training the early warning model to be trained, the training parameters in the model can be corrected based on the output results of the early warning model to be trained, that is, the trained early warning model can be obtained by correcting the loss function in the early warning model to be trained. Each enterprise's corresponding historical sample data has a corresponding loss value, which is determined based on the predicted early warning attributes and the actual label attributes.

[0093] Specifically, after inputting the historical sample data corresponding to a certain enterprise into the early warning model to be trained, the early warning model to be trained can obtain the prediction early warning attribute corresponding to the historical sample data. According to the label attribute (i.e., the actual label attribute) in the historical sample data of the prediction early warning attribute, the loss value corresponding to the historical sample data can be determined, and the model parameters in the early warning model to be trained can be corrected by the back-propagation method.

[0094] S4. The convergence of the loss function in the early warning model to be trained is taken as the training goal to obtain the early warning model to be screened.

[0095] Among them, the warning model to be screened is the model finally obtained through training and is used to determine the warning attribute value corresponding to the target data.

[0096] Specifically, the training error of the loss function, that is, the loss parameter, can be used as a condition for detecting whether the loss function has reached convergence, such as whether the training error is less than the preset error or whether the error change trend tends to be stable, or whether the current number of iterations is equal to the preset number. If the detection reaches the convergence condition, such as the training error of the loss function is less than the preset error or the error change tends to be stable, it indicates that the training of the early warning model to be trained is completed, and the iterative training can be stopped at this time. If it is detected that the convergence condition is not met at present, historical sample data can be further obtained to train the early warning model to be trained until the training error of the loss function is within the preset range. When the training error of the loss function reaches convergence, the early warning model to be trained can be used as the early warning model to be screened.

[0097] S5. Screen at least two early warning models to be screened according to pre-set screening conditions to obtain an early warning model.

[0098] In this embodiment, the screening conditions can be pre-determined according to the classification model evaluation index of each early warning model to be screened after training is completed. The classification model evaluation index includes but is not limited to the area under the ROC curve, the KS discrimination value (KS: Kolmogorov-Smirnov), the population stability value, etc.

[0099] In this embodiment, after completing the training process of each warning model to be trained, multiple warning models to be screened can be obtained, and the classification model evaluation index corresponding to each warning model to be screened can be determined. Further, the model performance score conversion can be performed according to the classification model evaluation index corresponding to each warning model to be screened, and the warning model to be screened with the highest model performance score is determined as the final warning model, so that a warning model with better classification performance can be obtained.

[0100] S240: Receive a plurality of to-be-processed data of at least one target enterprise.

[0101] S250. For a plurality of data to be processed, in response to an event that a warning label of the data to be processed is inconsistent with a preset label, obtaining target data of the data to be processed under a plurality of target dimensions.

[0102] Among them, multiple target dimensions are determined based on the data processing results of historical sample data under all dimensions to be selected.

[0103] S260: Input the target data into the pre-trained warning model, and output the warning attribute value corresponding to the target data.

[0104] S270: Determine the target risk attribute of the target data based on the warning attribute value of the target data and the target function.

[0105] S280. When the target risk attribute meets the preset conditions, output the target warning information of the data to be processed, wherein the target warning information is used to characterize the risk attribute of the target enterprise to which the data to be processed belongs.

[0106] The technical solution provided in this embodiment, when determining the target dimension, obtains historical sample data, wherein the historical sample data includes data content under at least one dimension to be selected, and then for all data content under the same dimension to be selected, determines the missing rate corresponding to the dimension to be selected based on all data content, thereby, for the dimension to be selected whose missing rate meets the preset conditions, determines the information value and stability of the dimension to be selected based on the data content under the dimension to be selected, and determines the target dimension based on the information value and stability. The technical solution of this embodiment, by screening the dimensions to be selected through information value and stability parameters, the target dimension obtained not only has a large contribution rate to model prediction, but also has stable data features, thereby performing feature screening before model training, improving the performance and accuracy of the model. When training the warning model based on the data content of the target dimension, the historical sample data is de-redundanted according to the target dimension to obtain the updated historical sample data. Subsequently, for at least two warning models to be trained, the data content under the target dimension in the historical sample data is input into the warning model to be trained, and the predicted warning attribute is output, wherein the model structure or the adopted algorithm of at least two warning models to be trained is different. Then, according to the predicted warning attribute and the label attribute in the historical sample data, the loss value is determined to correct the model parameters in the warning model to be trained based on the loss value. Further, the convergence of the loss function in the warning model to be trained is used as the training target to obtain the warning model to be screened. Finally, according to the pre-set screening conditions, at least two warning models to be screened are screened to obtain the warning model. In this embodiment, the model performance score is converted according to the classification model evaluation index corresponding to each warning model to be screened, and the warning model to be screened with the highest model performance score is determined as the final warning model, so that the warning model with better classification performance can be obtained.

[0107] Based on the above embodiments, S120 and S150 may be further described. Accordingly, the present invention provides the following embodiments: Figure 3 1 is a flow chart of a data processing method provided according to an embodiment of the present invention. Figure 3 As shown, the data processing method comprises the following steps:

[0108] S310, receiving a plurality of to-be-processed data of at least one target enterprise.

[0109] In this embodiment, the data to be processed may include a warning label, and different processing methods may be determined according to the consistency attribute of the warning label and the preset label. Specifically, the preset label may include a preset label under the first warning type and a preset label under the second warning type. The first warning type indicates that the warning signal is true. The second warning type indicates that the warning signal is risk-free. If the warning label meets the preset label under the first warning type, execute S320; if the warning label meets the preset label under the first warning type, execute S330; if the warning label neither meets the preset label under the first warning type nor the preset label under the first warning type, execute steps S340-S390. In this way, the risk attributes of the target enterprise are preliminarily determined through the warning label, which can filter out a part of the data processing volume, thereby improving the efficiency of confirming the overdue risk of the target enterprise.

[0110] S320: For a plurality of data to be processed, if the warning label carried by the data to be processed is consistent with a preset label under the first warning type, the data to be processed is determined to be target warning information.

[0111] Among them, the target warning information indicates the risk attribute of overdue resource return of the target enterprise to which the data to be processed belongs.

[0112] In this embodiment, the data to be processed carries at least one warning label. If there is a warning label that meets the preset label under the first warning type, it can be determined that the data to be processed is target warning information.

[0113] S330: For multiple data to be processed, if the warning label carried by the data to be processed is consistent with the preset label under the second warning type, it is determined that the warning of the data to be processed is true and risk-free.

[0114] Among them, the warning is true and there is no risk, which means that the warning information of the data to be processed is true, but the target enterprise to which the data to be processed belongs does not have the risk attribute of overdue resource return.

[0115] In this embodiment, if at least one warning label carried in the data to be processed meets the preset label under the second warning type, it can be determined that the warning of the data to be processed is true and risk-free.

[0116] S340. For multiple data to be processed, in response to an event that the warning label of the data to be processed is inconsistent with the preset label, obtain the target data of the data to be processed under multiple target dimensions, input the target data into a pre-trained warning model, and output the warning attribute value corresponding to the target data.

[0117] S350: Determine the target risk attribute of the target data based on the warning attribute value of the target data and the target function.

[0118] In this embodiment, a first warning threshold and a second warning threshold can be preset. For example, the first warning threshold is S1, the second warning threshold is S2, and S1 < S2. By presetting the warning threshold, the target warning information corresponding to the target enterprise can be efficiently determined.

[0119] S360. When the target risk attribute is less than the first warning threshold in the preset conditions, the target warning information of the data to be processed is output as warning without risk.

[0120] S370. When the target risk attribute is greater than the second warning threshold in the preset conditions, the target warning information of the data to be processed is output as warning属实.

[0121] S380. When the target risk attribute is between the first warning threshold and the second warning threshold, the target warning information of the data to be processed is output as transfer to the next step for confirmation.

[0122] Based on the above example, in the case where the target risk attribute is K, if K < S1, it is determined that the target warning information corresponding to the target enterprise is warning without risk, that is, it indicates that there is no risk of overdue resource return for the target enterprise; if K < S2, it is determined that the target warning information corresponding to the target enterprise is warning属实, that is, it indicates that the target enterprise does have a risk of overdue resource return; if S1 ≤ K ≤ S2, the target warning information of its target enterprise is transfer to the next step for confirmation, that is, it indicates that it is necessary for the downstream business system to further confirm whether the target enterprise has a risk of overdue resource return.

[0123] S390. Send the data to be verified with the result of the target warning information being transfer to the next step for confirmation to the downstream system for the downstream system to analyze and process the data to be verified.

[0124] Among them, the data to be verified can include the original data to be processed, the data content of all dimensions of the target enterprise, and the target risk attribute, etc. The downstream system can be a business system that requires professional technical personnel to analyze the data to be verified to obtain the final risk warning result.

[0125] In this embodiment, in the case where the target warning information of the target enterprise is transfer to the next step for confirmation, the relevant data of the target enterprise can be integrated and packaged into the data to be verified. Thus, the data to be verified can be sent to the downstream system so that the professional technical personnel of the downstream system can perform further confirmation tasks to obtain the risk warning result corresponding to the target enterprise. In this way, through the dual screening of the preset label and the warning model, the amount of manual confirmation is greatly reduced, not only improving the confirmation efficiency of the warning information, but also improving the confirmation credibility of the warning information.

[0126] The technical solution provided by this embodiment can include a warning label in the data to be processed. If the warning label carried by the data to be processed is consistent with the preset label under the first warning type, the data to be processed is determined to be the target warning information; if the warning label carried by the data to be processed is consistent with the preset label under the second warning type, the data to be processed is determined to be a warning that is true and risk-free. In this embodiment, the risk attribute of the target enterprise is first preliminarily determined by the warning label, and a part of the data processing volume can be filtered out, thereby improving the efficiency of determining the overdue risk of the target enterprise. When the target risk attribute determines the target warning information of the data to be processed, when the target risk attribute is less than the first warning threshold in the preset condition, the target warning information of the data to be processed is output as a warning without risk; when the target risk attribute is greater than the second warning threshold in the preset condition, the target warning information of the data to be processed is output as a warning that is true; when the target risk attribute is between the first warning threshold and the second warning threshold, the target warning information of the data to be processed is output as a transfer to the next step confirmation, and the data to be verified whose target warning information is the result of the transfer to the next step confirmation is sent to the downstream system, so as to analyze and process the data to be verified based on the downstream system. In this embodiment, the target warning information corresponding to the target enterprise can be efficiently determined by presetting the warning threshold.

[0127] Figure 4 Schematic diagram of a data processing device according to an embodiment of the present invention. Figure 4 As shown, the device includes: an enterprise data receiving module 410, a target data acquiring module 420, a warning attribute value determining module 430, a risk attribute determining module 440 and a warning information determining module 450.

[0128] The enterprise data receiving module 410 is used to receive a plurality of to-be-processed data of at least one target enterprise;

[0129] The target data acquisition module 420 is used to acquire target data of the data to be processed under multiple target dimensions in response to an event that the warning label of the data to be processed is inconsistent with the preset label, for the multiple data to be processed; wherein the multiple target dimensions are determined according to the data processing results of the historical sample data under all the dimensions to be selected;

[0130] The warning attribute value determination module 430 is used to input the target data into the pre-trained warning model and output the warning attribute value corresponding to the target data;

[0131] A risk attribute determination module 440, configured to determine a target risk attribute of the target data based on the warning attribute value of the target data and an objective function;

[0132] The warning information determination module 450 is used to output the target warning information of the data to be processed when the target risk attribute meets the preset conditions; wherein the target warning information is used to characterize the risk attribute of the target enterprise to which the data to be processed belongs.

[0133] The technical solution of this embodiment receives multiple data to be processed of at least one target enterprise, and then, for multiple data to be processed, in response to the event that the warning label of the data to be processed is inconsistent with the preset label, obtains the target data of the data to be processed under multiple target dimensions, wherein the multiple target dimensions are determined according to the data processing results of the historical sample data under all the dimensions to be selected, thereby inputting the target data into the pre-trained warning model, outputting the warning attribute value corresponding to the target data, and further, determining the target risk attribute of the target data based on the warning attribute value of the target data and the target function, and then, when the target risk attribute meets the preset conditions, outputting the risk attribute target warning information for characterizing the target enterprise to which the data to be processed belongs. The technical solution of the present invention, when confirming the warning event of the target enterprise, greatly reduces the amount of manual confirmation through the dual automatic confirmation mechanism of the preset label and the warning model, and improves the confirmation efficiency, real-time and accuracy of the warning information.

[0134] Based on the above device, optionally, the device further includes: a target dimension determination module; the target dimension determination module includes:

[0135] A historical sample acquisition unit, configured to acquire the historical sample data, wherein the historical sample data includes data content under at least one dimension to be selected;

[0136] A missing rate determining unit, configured to determine, for all data contents under the same dimension to be selected, a missing rate corresponding to the dimension to be selected according to all data contents;

[0137] The target dimension determination unit is used to determine the information value and stability of the dimension to be selected according to the data content of the dimension to be selected, and determine the target dimension according to the information value and the stability.

[0138] Based on the above device, optionally, the dimensions to be selected include the target enterprise controlled user dimension, the target enterprise's credit dimension, the historical transaction data dimension, the negative feedback dimension, the customer basic information dimension, the transaction account dimension, the loan dimension and at least one of the change trend dimensions under each dimension.

[0139] On the basis of the above device, optionally, a missing rate determination unit is specifically used to determine the missing rate corresponding to the dimension to be selected according to the number and total number of data contents existing in each dimension to be selected.

[0140] Based on the above device, optionally, the target dimension determination unit includes:

[0141] an information value determination subunit, for retrieving data content under the dimension to be selected whose missing rate is less than the first preset missing rate threshold in the preset condition, and analyzing and processing the data content to obtain the information value under the dimension to be selected;

[0142] A stability determination subunit, used to obtain a dimension to be selected whose information value is greater than a preset value threshold, and determine the stability of the dimension to be selected according to the data content of the dimension to be selected;

[0143] Accordingly, determining the target dimension according to the information value and the stability includes:

[0144] The first target dimension determination subunit is used to take the dimension to be selected whose stability is less than or equal to a preset stability threshold as the target dimension.

[0145] Based on the above device, optionally, the target dimension determination unit further includes:

[0146] a to-be-aggregated dimension determination subunit, configured to, if the number of the target dimensions is less than a preset number threshold, reselect the to-be-aggregated dimensions whose missing rate is less than a second preset missing rate threshold, wherein the second preset missing rate threshold is greater than the first preset missing rate threshold;

[0147] An aggregation dimension determination subunit is used to aggregate the data content in the dimension to be aggregated to obtain multiple clusters, and obtain at least two aggregation dimensions from each cluster;

[0148] The second target dimension determination subunit is used to determine the missing rates corresponding to at least two aggregation dimensions, and calculate the information value of the aggregation dimension whose missing rate is less than the first preset missing rate threshold, so as to determine whether the aggregation dimension is the target dimension based on the information value.

[0149] Based on the above device, optionally, the device further includes: an early warning model determination module. The early warning model determination module includes:

[0150] A historical sample updating unit, configured to perform redundancy removal processing on the historical sample data according to the target dimension to obtain updated historical sample data;

[0151] A prediction and warning attribute determination unit, for at least two warning models to be trained, inputting the data content under the target dimension in the historical sample data into the warning models to be trained, and outputting the prediction and warning attributes; wherein the model structures or the adopted algorithms of the at least two warning models to be trained are different;

[0152] A model parameter correction unit, used to determine a loss value according to the predicted warning attribute and the label attribute in the historical sample data, so as to correct the model parameters in the to-be-trained warning model based on the loss value;

[0153] A unit for determining a warning model to be screened, used for taking the convergence of the loss function in the warning model to be trained as a training target to obtain the warning model to be screened;

[0154] The early warning model determination unit is used to screen at least two early warning models to be screened according to preset screening conditions to obtain the early warning model.

[0155] On the basis of the above-mentioned device, optionally, the target data acquisition module 420 is also specifically used to determine that the data to be processed is the target warning information if the warning label carried by the data to be processed is consistent with the preset label under the first warning type; if the warning label carried by the data to be processed is consistent with the preset label under the second warning type, it is determined that the warning of the data to be processed is true and risk-free.

[0156] On the basis of the above-mentioned device, optionally, the warning information determination module 450 is further specifically used to output the target warning information of the data to be processed as a warning of no risk when the target risk attribute is less than the first warning threshold in the preset condition; output the target warning information of the data to be processed as a warning of true when the target risk attribute is greater than the second warning threshold in the preset condition; and output the target warning information of the data to be processed as confirmation of transfer to the next step when the target risk attribute is between the first warning threshold and the second warning threshold.

[0157] On the basis of the above device, the device may further include: a module for sending data to be verified, which is used to send the data to be verified whose target warning information is the result of the transfer to the next step of confirmation to a downstream system, so as to analyze and process the data to be verified based on the downstream system.

[0158] Based on the above device, optionally, the device further includes: an objective function determination module. The objective function determination module includes:

[0159] A reference value determination unit, configured to determine a first warning reference value and a second warning reference value based on a first preset warning attribute value; wherein the second warning reference value is twice the first warning reference value;

[0160] A first formula determination unit, configured to construct a first equation to be solved based on the first warning reference value and a first preset risk attribute value corresponding to the first warning reference value;

[0161] A second formula determination unit is used to construct a second equation to be solved by using a second warning reference value and a second preset risk attribute value corresponding to the second warning reference value; wherein the second warning reference value is twice the first warning reference value;

[0162] A coefficient determination unit, configured to construct a system of linear equations of two variables to be solved based on the first equation to be solved and the second equation to be solved, so as to solve the system of linear equations of two variables to be solved and obtain a first coefficient value and a second coefficient value;

[0163] The objective function determination unit is used to construct an objective function based on the first coefficient value and the second coefficient value.

[0164] The data processing device provided in the embodiment of the present invention can execute the data processing method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0165] Figure 5 The present invention is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. Figure 5 A block diagram of an exemplary electronic device 50 suitable for implementing exemplary embodiments of the present invention is shown. Figure 5 The electronic device 50 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0166] like Figure 5 As shown, the electronic device 50 is in the form of a general computing device. The components of the electronic device 50 may include but are not limited to: one or more processors or processing units 501, a system memory 502, and a bus 503 connecting different system components (including the system memory 502 and the processing unit 501).

[0167] Bus 503 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor or a local bus using any of a variety of bus structures. For example, these architectures include but are not limited to Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus and Peripheral Component Interconnect (PCI) bus.

[0168] The electronic device 50 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the electronic device 50, including volatile and non-volatile media, removable and non-removable media.

[0169] System memory 502 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 504 and / or cache memory 505. Electronic device 50 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 506 may be used to read and write non-removable, non-volatile magnetic media ( Figure 5 not shown, usually called a "hard drive"). Although Figure 5 Not shown in the figure, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, a DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to the bus 503 via one or more data medium interfaces. The memory 502 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of the embodiments of the present invention.

[0170] A program / utility 508 having a set (at least one) of program modules 507 may be stored, for example, in the memory 502, such program modules 507 including but not limited to an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. The program modules 507 generally perform the functions and / or methods of the embodiments described herein.

[0171] The electronic device 50 may also communicate with one or more external devices 509 (e.g., keyboards, pointing devices, displays 510, etc.), may communicate with one or more devices that enable a user to interact with the electronic device 50, and / or communicate with any device that enables the electronic device 50 to communicate with one or more other computing devices (e.g., a network card, a modem, etc.). Such communication may be performed through an input / output (I / O) interface 511. Furthermore, the electronic device 50 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through a network adapter 512. As shown, the network adapter 512 communicates with other modules of the electronic device 50 through the bus 503. It should be understood that although Figure 5Not shown, other hardware and / or software modules may be used in conjunction with the electronic device 50, including but not limited to: microcode, device drivers, de-redundancy processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0172] The processing unit 501 executes various functional applications and data processing based on streaming processing by running the programs stored in the system memory 502, for example, implementing the data processing method provided in the embodiment of the present invention.

[0173] An embodiment of the present invention further provides a storage medium containing computer executable instructions, and when the computer executable instructions are executed by a computer processor, they are used to perform a data processing method, the method comprising:

[0174] receiving a plurality of to-be-processed data of at least one target enterprise;

[0175] For the plurality of data to be processed, in response to an event that the warning label of the data to be processed is inconsistent with the preset label, obtaining target data of the data to be processed under multiple target dimensions; wherein the multiple target dimensions are determined according to the data processing results of the historical sample data under all the dimensions to be selected;

[0176] Input the target data into a pre-trained warning model, and output a warning attribute value corresponding to the target data;

[0177] Determining a target risk attribute of the target data based on the warning attribute value of the target data and the target function;

[0178] When the target risk attribute meets a preset condition, target warning information of the data to be processed is output; wherein the target warning information is used to characterize the risk attribute of the target enterprise to which the data to be processed belongs.

[0179] The computer storage medium of the embodiment of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0180] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0181] Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0182] Computer program code for performing the operation of embodiments of the present invention may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0183] An embodiment of the present invention further provides a computer program product, including a computer program, which, when executed by a processor, implements the data processing method provided in any embodiment of the present application.

[0184] In the process of implementation, the computer program product can be written in one or more programming languages ​​or a combination thereof to perform the computer program code of the present invention, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, using an Internet service provider to connect through the Internet).

[0185] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A data processing method, characterized in that: include: receiving a plurality of to-be-processed data of at least one target enterprise; For the plurality of data to be processed, in response to an event that the warning label of the data to be processed is inconsistent with the preset label, obtaining target data of the data to be processed under multiple target dimensions; wherein the multiple target dimensions are determined according to the data processing results of the historical sample data under all the dimensions to be selected; Input the target data into a pre-trained warning model, and output a warning attribute value corresponding to the target data; Determining a target risk attribute of the target data based on the warning attribute value of the target data and the target function; When the target risk attribute meets a preset condition, target warning information of the data to be processed is output; wherein the target warning information is used to characterize the risk attribute of the target enterprise to which the data to be processed belongs.

2. The method according to claim 1, characterized in that Also includes: Acquire the historical sample data, wherein the historical sample data includes data content under at least one dimension to be selected; For all data contents under the same dimension to be selected, determining the missing rate corresponding to the dimension to be selected according to all data contents; For the dimensions to be selected whose missing rate meets the preset conditions, the information value and stability of the dimensions to be selected are determined according to the data content under the dimensions to be selected, and the target dimension is determined based on the information value and the stability.

3. The method according to any one of claims 1-2, characterized in that: The dimensions to be selected include at least one of the target enterprise controlled user dimension, the target enterprise's credit dimension, the historical transaction data dimension, the negative feedback dimension, the customer basic information dimension, the transaction account dimension, the loan dimension, and the change trend dimension under each dimension.

4. The method according to claim 2, characterized in that: The determining the missing rate corresponding to the dimension to be selected according to all the data contents includes: According to the quantity and total quantity of data contents existing in each dimension to be selected, the missing rate corresponding to the dimension to be selected is determined.

5. The method according to claim 3, characterized in that: For the dimension to be selected whose missing rate meets the preset condition, determining the information value and stability of the dimension to be selected according to the data content of the dimension to be selected, includes: For a dimension to be selected whose missing rate is less than the first preset missing rate threshold in the preset condition, retrieve the data content under the dimension to be selected, and analyze and process the data content to obtain the information value under the dimension to be selected; Acquire a dimension to be selected whose information value is greater than a preset value threshold, and determine the stability of the dimension to be selected according to the data content of the dimension to be selected; Accordingly, determining the target dimension according to the information value and the stability includes: The dimension to be selected whose stability is less than or equal to a preset stability threshold is used as the target dimension.

6. The method according to claim 5, characterized in that Also includes: If the number of the target dimensions is less than a preset number threshold, reselecting the dimensions to be aggregated whose missing rate is less than a second preset missing rate threshold, wherein the second preset missing rate threshold is greater than the first preset missing rate threshold; Aggregate the data content in the dimension to be aggregated to obtain multiple clusters, and obtain at least two aggregation dimensions from each cluster; Determine the missing rates corresponding to at least two aggregation dimensions, and calculate the information value of the aggregation dimension whose missing rate is less than the first preset missing rate threshold, so as to determine whether the aggregation dimension is the target dimension based on the information value.

7. The method according to claim 3, characterized in that After determining the target dimension, the method further includes: For the historical sample data, performing redundancy removal processing on the historical sample data according to the target dimension to obtain updated historical sample data; For at least two warning models to be trained, the data content under the target dimension in the historical sample data is input into the warning model to be trained, and the predicted warning attribute is output; wherein the model structures or the adopted algorithms of the at least two warning models to be trained are different; Determine a loss value according to the predicted warning attribute and the label attribute in the historical sample data, so as to modify the model parameters in the to-be-trained warning model based on the loss value; Taking the convergence of the loss function in the early warning model to be trained as a training target to obtain an early warning model to be screened; At least two early warning models to be screened are screened according to preset screening conditions to obtain the early warning model.

8. The method according to claim 1, characterized in that The event in response to the early warning label of the data to be processed being inconsistent with the preset label includes: If the warning label carried by the data to be processed is consistent with the preset label under the first warning type, then determining that the data to be processed is the target warning information; If the warning label carried by the data to be processed is consistent with the preset label under the second warning type, it is determined that the warning of the data to be processed is true and risk-free.

9. The method according to claim 1, characterized in that: When the target risk attribute meets the preset condition, outputting the target warning information of the data to be processed includes: When the target risk attribute is less than the first warning threshold in the preset condition, outputting the target warning information of the data to be processed as warning of no risk; When the target risk attribute is greater than the second warning threshold in the preset condition, outputting the target warning information of the data to be processed as a warning is true; When the target risk attribute is between the first warning threshold and the second warning threshold, outputting the target warning information of the data to be processed is transferring to the next step for confirmation.

10. The method according to claim 9, characterized in that The method further comprises: The target warning information is the data to be verified as the result of transferring to the next step of confirmation, which is sent to the downstream system, so that the data to be verified is analyzed and processed based on the downstream system.

11. The method according to claim 1, characterized in that: The method further comprises: Based on the first preset warning attribute value, determining a first warning reference value and a second warning reference value; wherein the second warning reference value is twice the first warning reference value; constructing a first equation to be solved based on the first warning reference value and a first preset risk attribute value corresponding to the first warning reference value; Constructing a second equation to be solved by using a second warning reference value and a second preset risk attribute value corresponding to the second warning reference value; Based on the first equation to be solved and the second equation to be solved, construct a system of linear equations of two variables to be solved, so as to solve the system of linear equations of two variables to be solved and obtain a first coefficient value and a second coefficient value; Based on the first coefficient value and the second coefficient value, an objective function is constructed.

12. A data processing method and device, characterized in that: include: An enterprise data receiving module, used for receiving a plurality of to-be-processed data of at least one target enterprise; A target data acquisition module is used to acquire target data of the data to be processed under multiple target dimensions in response to an event that the warning label of the data to be processed is inconsistent with the preset label, for the multiple data to be processed; wherein the multiple target dimensions are determined according to the data processing results of the historical sample data under all the dimensions to be selected; A warning attribute value determination module is used to input the target data into a pre-trained warning model and output the warning attribute value corresponding to the target data; A risk attribute determination module, used to determine the target risk attribute of the target data based on the warning attribute value of the target data and the target function; The warning information determination module is used to output the target warning information of the data to be processed when the target risk attribute meets the preset conditions; wherein the target warning information is used to characterize the risk attribute of the target enterprise to which the data to be processed belongs.

13. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the data processing method according to any one of claims 1 to 11 when executing the computer program.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the data processing method as described in any one of claims 1 to 11 is implemented.

15. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the computer program implements the data processing method according to any one of claims 1 to 11.

Images