Abnormal Service Identification Method and Device

Abstract

This invention provides a method and apparatus for identifying abnormal business transactions, belonging to the field of artificial intelligence technology. The method includes: acquiring current feature data of the business; inputting the current feature data into a business classification model created based on historical feature data to obtain the business type; inputting the current feature data into a business transfer cost integration model corresponding to the business type, created based on historical feature data and historical business transfer costs, to obtain the business transfer prediction cost; and outputting the abnormal business based on a comparison between the business transfer prediction cost and a preset warning threshold range. This invention can provide timely warnings of abnormal business transactions, reducing bank transaction risks and minimizing potential losses caused by abnormal transactions.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and more specifically, to a method and apparatus for identifying abnormal business operations. Background Technology

[0002] When considering the costs of their business products, banking institutions must not only compare themselves horizontally with the general cost levels of the market and peers, but also vertically consider the potential costs brought about by internal and external risk factors. Currently, the prediction of business transfer costs is difficult to base on historical data over a long period and relies heavily on subjective human factors. Therefore, it is impossible to accurately identify abnormal transactions, which can increase the bank's transaction risk and lead to losses. Summary of the Invention

[0003] The main objective of this invention is to provide a method and apparatus for identifying abnormal business transactions, and to provide early warnings for abnormal transactions, thereby reducing bank transaction risks and minimizing potential losses caused by abnormal transactions.

[0004] To achieve the above objectives, embodiments of the present invention provide an abnormal service identification method, including:

[0005] Obtain the current feature data of the business, input the current feature data into the business classification model created based on historical feature data, and obtain the business type;

[0006] Input the current feature data into the integrated business transfer cost model created based on historical feature data and historical business transfer costs corresponding to the business type to obtain the business transfer prediction cost;

[0007] The abnormal business is output based on the comparison between the predicted cost of business transfer and the preset early warning threshold range.

[0008] This invention also provides an abnormal service identification device, comprising:

[0009] The business type module is used to obtain the current feature data of a business, input the current feature data into a business classification model created based on historical feature data, and obtain the business type.

[0010] The business transfer prediction cost module is used to input the current feature data into the business transfer cost integration model created based on historical feature data and historical business transfer costs corresponding to the business type, so as to obtain the business transfer prediction cost.

[0011] The abnormal business module is used to output abnormal business based on the comparison results between the business transfer prediction cost and the preset early warning threshold range.

[0012] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the abnormal service identification method.

[0013] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the abnormal service identification method.

[0014] The abnormal business identification method and apparatus of this invention first inputs the current feature data into a business classification model created based on historical feature data to obtain the business type. Then, it inputs the current feature data into a business transfer cost integration model created based on historical feature data and historical business transfer costs corresponding to the business type to obtain the business transfer prediction cost. Finally, it outputs the abnormal business based on the comparison result between the business transfer prediction cost and the preset warning threshold range, so as to provide timely warnings for abnormal businesses, reduce bank transaction risks, and reduce potential losses caused by business anomalies. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart of the abnormal service identification method in an embodiment of the present invention;

[0017] Figure 2 This is a flowchart of an abnormal service identification method in another embodiment of the present invention;

[0018] Figure 3 This is a flowchart illustrating the creation of a business classification model based on historical feature data in an embodiment of the present invention;

[0019] Figure 4 This is a flowchart illustrating the creation of a business transfer cost integration model based on historical feature data and historical business transfer costs in an embodiment of the present invention.

[0020] Figure 5 This is a flowchart illustrating the determination of target weights in an embodiment of the present invention;

[0021] Figure 6 This is a structural block diagram of the abnormal service identification device in an embodiment of the present invention;

[0022] Figure 7 This is a structural block diagram of an abnormal service identification device in another embodiment of the present invention;

[0023] Figure 8 yes Figure 7 Structure diagram of the business data server in China;

[0024] Figure 9 yes Figure 7 A block diagram of the structure of a machine learning server in China;

[0025] Figure 10 yes Figure 7 Structural diagram of the abnormal service identification server in China;

[0026] Figure 11 This is a structural block diagram of the computer device in an embodiment of the present invention. Detailed Implementation

[0027] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0028] Those skilled in the art will recognize that embodiments of the present invention can be implemented as a system, apparatus, device, method, or computer program product. Therefore, this disclosure can be specifically implemented in the following forms: entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0029] Given that existing technologies cannot identify abnormal transactions, increasing bank transaction risks and causing losses, this invention provides an abnormal transaction identification method. This method can provide timely warnings of abnormal transactions, reduce bank transaction risks, minimize potential losses caused by abnormal transactions, promote profitability in the banking industry, and mitigate adverse factors arising from the market itself and the structure of business products. The invention will be described in detail below with reference to the accompanying drawings.

[0030] Figure 1 This is a flowchart of the abnormal service identification method in an embodiment of the present invention. Figure 2 This is a flowchart of an abnormal service identification method in another embodiment of the present invention. For example... Figure 1 and Figure 2 As shown, the abnormal service identification methods include:

[0031] S101: Obtain the current feature data of the business, input the current feature data into the business classification model created based on historical feature data, and obtain the business type.

[0032] Before executing S101, the process also includes: acquiring business product data, performing basic validity checks on the business product data to ensure the accuracy of the acquired current feature data. Validity checks include data filtering, handling missing data, and removing outlier data. For example, if business amount data for a certain date is missing, which may indicate a data quality issue, then the feature value for that day will be the average of the business amounts for the two days before and after it. x i-1 x represents the business amount for day i-1. i+1 Let x be the business amount on day i+1. i Let represent the business amount on day i. If the data for a certain day increases or decreases sharply, with a significant difference from the days before and after, then the data is considered to be abnormal data caused by a sudden or accidental event, and this abnormal data will be removed.

[0033] The business product data includes bond business data and financial market trading data from the financial market trading system, loan business data from the loan business system, document business data from the document business system, and interbank clearing data from the interbank clearing system.

[0034] Table 1 is a feature data table in an embodiment of the present invention. As shown in Table 1, the feature data may include counterparty type, counterparty location region, final counterparty type, final counterparty location region, transaction amount, transaction currency, interest rate floating type, interest rate benchmark type, transaction accrual date, transaction maturity date, transaction cycle period, preset fee rate, and transaction purpose.

[0035] Table 1

[0036] Feature data Eigenvalues Counterparty type Individuals, companies or organizations Counterparty's location Country Code Final counterparty type Individuals, companies or organizations The region where the final counterparty is located Country Code Business amount Actual amount Business currency Currency ISO Code Interest rate floating type Fixed or floating Interest rate benchmark type benchmark interest rate Business commencement date Value Date Business Expiry Date Date of Expiry Business Cycle Cycle period Preset rate Initial business fee rate Business applications Business applications

[0037] S102: Input the current feature data into the business transfer cost integration model created based on historical feature data and historical business transfer costs corresponding to the business type to obtain the business transfer prediction cost.

[0038] Table 2 shows the data structure for business transfer forecasting costs. As shown in Table 2, the data structure for business transfer forecasting costs includes business type, business currency, forecast date, business transfer forecasting cost value, as well as the business number and data generation date.

[0039] Table 2

[0040] Busi_type varchar2(50) Business Type currtype varchar2(10) Business currency Predict_day varchar2(10) Predicted date Predict_value number(30,5) Forecast cost of business transfer Busi_no varchar2(100) Business Number Report_date Varchar2(10) Data generation date

[0041] S103: Output abnormal business based on the comparison result between the business transfer prediction cost and the preset early warning threshold range.

[0042] In practice, services that output service transfer prediction costs exceeding the preset warning threshold are considered abnormal services. Warning notifications are generated based on abnormal services and sent to relevant personnel.

[0043] Figure 1 The abnormal service identification method shown can be executed by a computer. Figure 1 As shown in the process, the abnormal business identification method of this embodiment first inputs the current feature data into a business classification model created based on historical feature data to obtain the business type. Then, it inputs the current feature data into a business transfer cost integration model created based on historical feature data and historical business transfer costs corresponding to the business type to obtain the business transfer prediction cost. Finally, it outputs the abnormal business based on the comparison result between the business transfer prediction cost and the preset warning threshold range, so as to provide timely warnings for abnormal businesses, reduce bank transaction risks, and reduce potential losses caused by business anomalies.

[0044] Figure 3 This is a flowchart illustrating the creation of a business classification model based on historical feature data in an embodiment of the present invention. For example... Figure 3 As shown, creating a business classification model based on historical feature data includes:

[0045] Perform the following iterative processing:

[0046] S201: Determine the Gini coefficient of the historical feature data based on the Gini coefficient of each feature value in the historical feature data set.

[0047] For example, let the historical feature dataset be D, with D as the root node. The historical feature data X has K feature values, and the Gini coefficient of each feature value is: G k =p k (1-p k ); where G k p is the Gini coefficient of the k-th eigenvalue. k This represents the proportion of feature data corresponding to the k-th feature value out of the total number of feature data corresponding to all feature values. Historical features.

[0048] The sum of the Gini coefficients of each eigenvalue of the data, i.e. G is the Gini coefficient of historical feature data X.

[0049] The coefficient, K, represents the number of feature values ​​under the historical feature data X.

[0050] S202: Determine the split node based on the minimum Gini coefficient of each feature value in the historical feature data corresponding to the minimum Gini coefficient, and divide the historical feature dataset into two historical feature data subsets based on the split node.

[0051] In specific implementation, let Y be the historical feature data corresponding to the minimum Gini coefficient of each historical feature data. If the Gini coefficient of feature value y in historical feature data Y is the smallest, then y is used as the splitting node to split the historical feature dataset D into the first historical feature data subset D1 and the second historical feature data subset D2.

[0052] S203: Determine whether the number of historical feature data in the historical feature data subset is less than the preset feature quantity threshold or whether the Gini coefficient of one of the historical feature data is less than the preset Gini threshold.

[0053] S204: When the number of historical feature data in the subset of historical feature data is less than the preset feature number threshold or the Gini coefficient of one of the historical feature data is less than the preset Gini threshold, create a business classification model (business classification decision tree) based on the split node in the current iteration.

[0054] For example, when the number of historical feature data in D1 is less than the preset feature number threshold n, or when the Gini coefficient of one of the historical feature data in D1 is less than the preset Gini threshold m, the splitting of the first historical feature data subset D1 is stopped.

[0055] S205: When the number of historical feature data in the historical feature data subset is greater than or equal to the preset feature number threshold and the Gini coefficient of all historical feature data is greater than or equal to the preset Gini threshold, the historical feature dataset is replaced with the historical feature data subset, and the iterative processing continues.

[0056] For example, when the number of historical feature data in D2 is greater than or equal to the preset feature number threshold n, or the Gini coefficient of one of the historical feature data in D2 is greater than or equal to the preset Gini threshold m, the historical feature dataset D in S201 is replaced with the second historical feature data subset D2, and S201 is returned to continue splitting the second historical feature data subset D2.

[0057] Figure 4 This is a flowchart illustrating the creation of a business transfer cost integration model based on historical feature data and historical business transfer costs in an embodiment of the present invention. For example... Figure 4 As shown, the integrated business migration cost model, created based on historical feature data and historical business migration costs, includes:

[0058] Perform the following iterative processing:

[0059] S301: Create sampling sets for each business type based on the historical feature data in each subset of historical feature data.

[0060] In this invention, the Bagging ensemble learning method is used to divide historical feature data into various historical feature data subsets according to the business classification model. Taking one of the historical feature data subsets as an example, the historical feature data in this subset is randomly sampled T times to obtain T sample sets.

[0061] S302: Input each sample set into the preset gradient boosting tree model to obtain the historical business transfer cost prediction value for each sample set.

[0062] S303: Determine the loss function for each sampling set based on the predicted historical service transfer cost and the corresponding historical service transfer cost.

[0063] S304: Determine whether the loss function is less than the preset loss function threshold.

[0064] S305: When the loss function is less than the preset loss function threshold, determine that each gradient boosting tree model in the current iteration is the basic model of each sampling set.

[0065] For example, the T gradient boosting tree models in the current iteration are identified as the T basic models.

[0066] S306: When the loss function is greater than or equal to the preset loss function threshold, update each gradient boosting tree model according to the loss function of each sampling set and continue to perform iterative processing.

[0067] S307: Create an integrated business transfer cost model for each business type based on the basic model of each sampling set and the target weights of each basic model.

[0068] Figure 5 This is a flowchart illustrating the determination of target weights in an embodiment of the present invention. For example... Figure 5 As shown, determining the target weights includes:

[0069] Perform the following iterative processing:

[0070] S401: Input the historical feature data of each sampling set into the corresponding basic model to obtain the basic transfer cost prediction value of each basic model.

[0071] S402: Determine whether the current iteration count has reached the preset iteration count.

[0072] S403: When the current iteration count reaches the preset iteration count, determine the preset weights of each basic model as the target weights of each basic model.

[0073] In practice, 63.2% of the historical feature data in the sampling set can be used for training, and 36.8% of the historical feature data can be used for validation. The preset number of iterations and the preset loss function threshold can be adjusted based on the validation results to continuously optimize the business transfer cost integration model for each business type.

[0074] S404: When the current iteration count has not reached the preset iteration count, update the preset weights based on the deviation between the basic transfer cost prediction value of each basic model and the corresponding historical business transfer cost, and continue to perform iterative processing.

[0075] For example, let h be the deviation corresponding to the i-th basic model. i There are T basic models in total. The preset weights of the i-th basic model in the next iteration are updated as follows: Among them, W i The preset weights are the updated weights for the i-th basic model.

[0076] The specific process of this invention embodiment is as follows:

[0077] 1. Determine the Gini coefficient of the historical feature data based on the Gini coefficient of each feature value in the historical feature data set.

[0078] 2. Determine the splitting node based on the minimum Gini coefficient of each feature value in the historical feature data corresponding to the minimum Gini coefficient of the historical feature data, and divide the historical feature dataset into two historical feature data subsets based on the splitting node.

[0079] 3. When the number of historical feature data in the historical feature data subset is less than the preset feature number threshold or the Gini coefficient of one of the historical feature data is less than the preset Gini threshold, create a business classification model based on the split node in the current iteration; otherwise, replace the historical feature dataset with the historical feature data subset and return to step 1.

[0080] 4. Create sampling sets for each business type based on the historical feature data in each subset of historical feature data.

[0081] 5. Input each sample set into the preset gradient boosting tree model to obtain the predicted historical business transfer cost value for each sample set.

[0082] 6. Determine the loss function for each sampling set based on the predicted historical business transfer costs and the corresponding historical business transfer costs.

[0083] 7. When the loss function is less than the preset loss function threshold, determine that each gradient boosting tree model in the current iteration is the basic model of each sampling set; otherwise, update each gradient boosting tree model according to the loss function of each sampling set and return to step 4.

[0084] 8. Input the historical feature data of each sampling set into the corresponding basic model to obtain the basic transfer cost prediction value of each basic model.

[0085] 9. When the current iteration count reaches the preset iteration count, determine the preset weight of each basic model as the target weight of each basic model; otherwise, update the preset weight based on the deviation between the basic transfer cost prediction value of each basic model and the corresponding historical business transfer cost, and return to step 8.

[0086] 10. Create integrated business transfer cost models for each business type based on the basic models of each sampling set and the target weights of each basic model.

[0087] 11. Obtain the current feature data of the business, input the current feature data into the business classification model, and obtain the business type.

[0088] 12. Input the current feature data into the business transfer cost integration model corresponding to the business type to obtain the business transfer prediction cost.

[0089] 13. Output abnormal business based on the comparison results of the business transfer prediction cost and the preset early warning threshold range.

[0090] In summary, the abnormal service identification method of the present invention has the following beneficial effects:

[0091] 1. Supports multiple feature data, and uses decision trees to classify and identify feature data for business purposes. It can not only predict actual business but also predict virtual business and even predict business product data with missing feature information, and has strong compatibility.

[0092] 2. More accurate prediction of business transfer costs: After introducing machine learning algorithms, it can learn from training sets over a longer time period by combining internal and external information. In addition, by combining the Bagging ensemble learning method with the business classification method, the prediction results of multiple factors affecting business transfer costs can be accumulated. By iteratively updating the loss function, the prediction results can be gradually approximated to the true value of business transfer costs. Furthermore, by using the ensemble method, multiple independent prediction models are combined into a business transfer cost ensemble model, which is superior to the prediction results of any single classification and greatly improves the overall prediction accuracy.

[0093] 3. It has strong generalization ability and is generally applicable to quantitative indicators that have periodic patterns and rely on data from a long historical period for prediction.

[0094] Based on the same inventive concept, this embodiment of the invention also provides an abnormal service identification device. Since the principle of this device in solving the problem is similar to that of the abnormal service identification method, the implementation of this device can refer to the implementation of the method, and the repeated parts will not be described again.

[0095] Figure 6 This is a structural block diagram of the abnormal service identification device in an embodiment of the present invention. Figure 7 This is a structural block diagram of an abnormal service identification device in another embodiment of the present invention. Figure 8 yes Figure 7 A structural diagram of the business data server in China. Figure 9 yes Figure 7 A block diagram of the structure of a machine learning server. Figure 10 yes Figure 7 A structural diagram of the abnormal service identification server in China. (Example) Figures 6-10 As shown, the abnormal service identification device includes:

[0096] The business type module is used to obtain the current feature data of a business, input the current feature data into a business classification model created based on historical feature data, and obtain the business type.

[0097] The business transfer prediction cost module is used to input the current feature data into the business transfer cost integration model created based on historical feature data and historical business transfer costs corresponding to the business type, so as to obtain the business transfer prediction cost.

[0098] The abnormal business module is used to output abnormal business based on the comparison results between the business transfer prediction cost and the preset early warning threshold range.

[0099] In one embodiment, it further includes:

[0100] The business classification model creation module is used to perform the following iterative processes:

[0101] The Gini coefficient of the historical feature data is determined based on the Gini coefficient of each feature value under the historical feature data in the historical feature dataset.

[0102] The splitting node is determined based on the minimum Gini coefficient of each feature value under the historical feature data corresponding to the minimum Gini coefficient of the historical feature data. The historical feature dataset is then divided into two historical feature data subsets based on the splitting node.

[0103] If the number of historical feature data in the historical feature data subset is less than the preset feature number threshold or the Gini coefficient of one of the historical feature data is less than the preset Gini threshold, a business classification model is created based on the split node in the current iteration; otherwise, the historical feature dataset is replaced with the historical feature data subset, and the iterative processing continues.

[0104] In one embodiment, it further includes:

[0105] The business transfer cost integration model creation module is used to perform the following iterative processes:

[0106] Create sampling sets for each business type based on the historical feature data in each subset of historical feature data;

[0107] Each sample set is input into a preset gradient boosting tree model to obtain the historical business transfer cost prediction value for each sample set;

[0108] The loss function for each sampling set is determined based on the predicted historical business transfer costs and the corresponding historical business transfer costs.

[0109] When the loss function is less than the preset loss function threshold, each gradient boosting tree model in the current iteration is determined to be the basic model of each sampling set; otherwise, each gradient boosting tree model is updated according to the loss function of each sampling set, and the iterative process continues.

[0110] An integrated model for business transfer costs for each business type is created based on the basic model of each sampling set and the target weights of each basic model.

[0111] In one embodiment, it further includes:

[0112] The target weight determination module is used to perform the following iterative process:

[0113] By inputting the historical feature data of each sampling set into the corresponding basic model, the basic transfer cost prediction value of each basic model is obtained;

[0114] When the current iteration count reaches the preset iteration count, the preset weight of each basic model is determined as the target weight of each basic model; otherwise, the preset weight is updated according to the deviation between the basic transfer cost prediction value of each basic model and the corresponding historical business transfer cost, and the iteration process continues.

[0115] like Figures 7-10 As shown, in practical applications, the abnormal service identification device includes a service data server, a machine learning server, and an abnormal service identification server.

[0116] The business data server is used to receive and load business product data from various bank application servers (such as financial market trading systems, loan business systems, securities business systems, and interbank clearing systems), and to uniformly store and verify the business product data (to ensure the data quality of feature data).

[0117] like Figure 8 As shown, the business data server includes a source data loading module and a source data preprocessing module.

[0118] The source data loading module is used to receive and load business product data from various bank application servers.

[0119] The source data preprocessing module is used to perform basic legality checks on business product data by setting verification rules to ensure the correctness of feature data. Legality checks include data filtering, missing data handling, and abnormal data removal.

[0120] like Figure 9 As shown, the machine learning server trains machine learning models based on historical data and machine learning algorithms, and continuously improves the machine learning models by optimizing the model parameters. It includes a historical feature data import module, a historical business transfer cost import module, a decision tree analysis module, a machine learning process module, and a model validation module.

[0121] The Historical Feature Data Import module is used to import historical feature data from the historical feature dataset.

[0122] The Historical Business Transfer Cost Import module is used to match historical business transfer costs and historical feature data according to a point in time.

[0123] The decision tree analysis module includes a business classification model creation module.

[0124] The machine learning process module includes a business transfer cost integration model creation module and a target weight determination module.

[0125] The model validation module is used to validate the business transfer cost integration model on 36.8% of the historical feature data in the sampling set. Based on the validation results, the preset iteration number and preset loss function threshold are adjusted to continuously optimize the business transfer cost integration model for each business type.

[0126] like Figure 10 As shown, the abnormal service identification server includes a service transfer prediction cost management module, a service transfer prediction cost processing module, and an abnormal service early warning module.

[0127] The business transfer forecasting cost management module is used to configure the loss function threshold, preset number of iterations, warning threshold range, warning reminder method and frequency.

[0128] The business migration prediction cost processing module includes a business type module and a business migration prediction cost module. It is used to load data from the business data server every day, call the business classification model and business migration cost integration model trained by the machine learning server, and obtain the business migration prediction cost.

[0129] The abnormal business early warning module includes the abnormal business module.

[0130] In summary, the abnormal business identification device of this embodiment first inputs the current feature data into a business classification model created based on historical feature data to obtain the business type. Then, it inputs the current feature data into a business transfer cost integration model created based on historical feature data and historical business transfer costs corresponding to the business type to obtain the business transfer prediction cost. Finally, it outputs the abnormal business based on the comparison result between the business transfer prediction cost and the preset warning threshold range, so as to provide timely warnings for abnormal businesses, reduce bank transaction risks, and reduce potential losses caused by business anomalies.

[0131] This invention also provides a specific implementation of a computer device capable of implementing all the steps in the abnormal service identification method described above. Figure 11 This is a structural block diagram of the computer device in an embodiment of the present invention, see below. Figure 11 The computer equipment specifically includes the following:

[0132] Processor 1101 and memory 1102.

[0133] The processor 1101 is used to call the computer program in the memory 1102. When the processor executes the computer program, it implements all the steps in the abnormal service identification method in the above embodiments. For example, when the processor executes the computer program, it implements the following steps:

[0134] Obtain the current feature data of the business, input the current feature data into the business classification model created based on historical feature data, and obtain the business type;

[0135] Input the current feature data into the integrated business transfer cost model created based on historical feature data and historical business transfer costs corresponding to the business type to obtain the business transfer prediction cost;

[0136] The abnormal business is output based on the comparison between the predicted cost of business transfer and the preset early warning threshold range.

[0137] In summary, the computer device in this embodiment of the invention first inputs the current feature data into a business classification model created based on historical feature data to obtain the business type. Then, it inputs the current feature data into a business transfer cost integration model created based on historical feature data and historical business transfer costs corresponding to the business type to obtain the business transfer prediction cost. Finally, it outputs abnormal business based on the comparison result between the business transfer prediction cost and the preset warning threshold range, so as to provide timely warnings for abnormal business, reduce bank transaction risks, and reduce potential losses caused by business anomalies.

[0138] This invention also provides a computer-readable storage medium capable of implementing all steps of the abnormal service identification method in the above embodiments. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements all steps of the abnormal service identification method in the above embodiments. For example, when the processor executes the computer program, it implements the following steps:

[0139] Obtain the current feature data of the business, input the current feature data into the business classification model created based on historical feature data, and obtain the business type;

[0140] Input the current feature data into the integrated business transfer cost model created based on historical feature data and historical business transfer costs corresponding to the business type to obtain the business transfer prediction cost;

[0141] The abnormal business is output based on the comparison between the predicted cost of business transfer and the preset early warning threshold range.

[0142] In summary, the computer-readable storage medium of this embodiment first inputs the current feature data into a business classification model created based on historical feature data to obtain the business type. Then, it inputs the current feature data into a business transfer cost integration model created based on historical feature data and historical business transfer costs corresponding to the business type to obtain the business transfer prediction cost. Finally, it outputs abnormal business based on the comparison result between the business transfer prediction cost and the preset warning threshold range, so as to provide timely warnings for abnormal business, reduce bank transaction risks, and reduce potential losses caused by business anomalies.

[0143] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0144] Those skilled in the art will also understand that the various illustrative logical blocks, units, and steps listed in the embodiments of the present invention can be implemented by electronic hardware, computer software, or a combination of both. To clearly demonstrate the interchangeability of hardware and software, the functions of the various illustrative components, units, and steps described above have been generally described. Whether such functionality is implemented through hardware or software depends on the specific application and the overall system design requirements. Those skilled in the art can implement the described functions using various methods for each specific application, but such implementation should not be construed as exceeding the scope of protection of the embodiments of the present invention.

[0145] The various illustrative logic blocks, units, or devices described in the embodiments of this invention can be implemented or operate the described functions using a general-purpose processor, digital signal processor, application-specific integrated circuit (ASIC), field-programmable gate array or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof. The general-purpose processor can be a microprocessor; alternatively, it can be any conventional processor, controller, microcontroller, or state machine. The processor can also be implemented using a combination of computing devices, such as a digital signal processor and a microprocessor, multiple microprocessors, one or more microprocessors combined with a digital signal processor core, or any other similar configuration.

[0146] The steps of the methods or algorithms described in the embodiments of this invention can be directly embedded in hardware, a software module executed by a processor, or a combination of both. The software module can be stored in RAM, flash memory, ROM, EPROM, EEPROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium in the art. Exemplarily, the storage medium can be connected to the processor so that the processor can read information from and write information to the storage medium. Optionally, the storage medium can also be integrated into the processor. The processor and storage medium can be housed in an ASIC, which can be housed in a user terminal. Optionally, the processor and storage medium can also be housed in different components of the user terminal.

[0147] In one or more exemplary designs, the functions described in the embodiments of the present invention can be implemented in hardware, software, firmware, or any combination of these three. If implemented in software, these functions can be stored on a computer-readable medium or transmitted on a computer-readable medium in the form of one or more instructions or code. Computer-readable media include computer storage media and communication media that facilitate the transfer of computer programs from one place to another. Storage media can be any available media that can be accessed by a general-purpose or special-purpose computer. For example, such computer-readable media can include, but is not limited to, RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store program code in the form of instructions or data structures and other forms that can be read by a general-purpose or special-purpose computer, or a general-purpose or special-purpose processor. Furthermore, any connection can be suitably defined as a computer-readable medium, for example, if the software is transmitted from a website, server or other remote resource via a coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL) or wirelessly, such as infrared, wireless and microwave, it is also included in the defined computer-readable medium. The disks and discs mentioned include compressed disks, laser discs, optical discs, DVDs, floppy disks, and Blu-ray discs. Disks typically copy data magnetically, while disks typically copy data optically using lasers. Combinations of the above can also be contained in computer-readable media.

Claims

1. A method for identifying abnormal business operations, characterized in that, include: Obtain the current feature data of the business, and input the current feature data into the business classification model created based on historical feature data to obtain the business type; The current feature data is input into the integrated business transfer cost model created based on historical feature data and historical business transfer costs corresponding to the business type to obtain the business transfer prediction cost. The abnormal business is output based on the comparison between the predicted cost of business transfer and the preset early warning threshold range; Creating a business classification model based on historical feature data includes the following iterative process: determining the Gini coefficient of the historical feature data based on the Gini coefficient of each feature value in the historical feature data set; determining the split node based on the minimum Gini coefficient of each feature value in the historical feature data set corresponding to the minimum Gini coefficient; dividing the historical feature data set into two historical feature data subsets based on the split node; when the number of historical feature data in the historical feature data subset is less than a preset feature number threshold or the Gini coefficient of one of the historical feature data is less than the preset Gini threshold, creating a business classification model based on the split node in the current iteration; otherwise, replacing the historical feature data set with the historical feature data subset and continuing the iterative process. Creating an integrated business transfer cost model based on historical feature data and historical business transfer costs includes the following iterative processing: Creating sample sets for each business type based on historical feature data from each subset of historical feature data; inputting each sample set into a preset gradient boosting tree model to obtain the predicted historical business transfer cost for each sample set; determining the loss function for each sample set based on the predicted historical business transfer cost and the corresponding historical business transfer cost; when the loss function is less than a preset loss function threshold, determining each gradient boosting tree model in the current iteration as the basic model for each sample set; otherwise, updating each gradient boosting tree model based on the loss function of each sample set and continuing the iterative processing; creating an integrated business transfer cost model for each business type based on the basic model of each sample set and the target weights of each basic model. The historical and current feature data include counterparty type, counterparty location region, final counterparty type, final counterparty location region, transaction amount, transaction currency, interest rate floating type, interest rate benchmark type, transaction accrual date, transaction maturity date, transaction cycle period, preset fee rate, and transaction purpose. Perform the following iterative processing: By inputting the historical feature data of each sampling set into the corresponding basic model, the basic transfer cost prediction value of each basic model is obtained; When the current iteration count reaches the preset iteration count, the preset weight of each basic model is determined as the target weight of each basic model; otherwise, the preset weight is updated according to the deviation between the basic transfer cost prediction value of each basic model and the corresponding historical business transfer cost, and the iteration process continues. When the current iteration count reaches a preset iteration count, the preset weights of each basic model are determined as the target weights of each basic model; otherwise, the preset weights are updated based on the deviation between the predicted basic transfer cost of each basic model and the corresponding historical business transfer cost, and the iteration process continues, including: Let h be the deviation corresponding to the i-th basic model. i There are T basic models in total. The preset weights of the i-th basic model in the next iteration are updated as follows: Among them, W i The preset weights are the updated weights for the i-th basic model.

2. An abnormal service identification device, characterized in that, include: The business type module is used to obtain the current feature data of the business, and input the current feature data into the business classification model created based on historical feature data to obtain the business type; The business transfer prediction cost module is used to input the current feature data into the business transfer cost integration model created based on historical feature data and historical business transfer costs corresponding to the business type, so as to obtain the business transfer prediction cost. The abnormal business module is used to output abnormal business based on the comparison results between the business transfer prediction cost and the preset early warning threshold range; The abnormal service identification device further includes: The business classification model creation module performs the following iterative process: It determines the Gini coefficient of historical feature data based on the Gini coefficient of each feature value in the historical feature dataset; it determines a split node based on the minimum Gini coefficient of each feature value in the historical feature data corresponding to the minimum Gini coefficient; it divides the historical feature dataset into two historical feature data subsets based on the split node; when the number of historical feature data in the historical feature data subset is less than a preset feature quantity threshold or the Gini coefficient of one of the historical feature data is less than the preset Gini threshold, it creates a business classification model based on the split node in the current iteration; otherwise, it replaces the historical feature dataset with the historical feature data subset and continues the iterative process. The abnormal service identification device further includes: The business transfer cost integration model creation module performs the following iterative processing: creating sample sets for each business type based on historical feature data from each subset of historical feature data; inputting each sample set into a preset gradient boosting tree model to obtain the predicted historical business transfer cost for each sample set; determining the loss function for each sample set based on the predicted historical business transfer cost and the corresponding historical business transfer cost; when the loss function is less than a preset loss function threshold, determining that each gradient boosting tree model in the current iteration is the basic model for each sample set; otherwise, updating each gradient boosting tree model based on the loss function of each sample set and continuing the iterative processing; and creating a business transfer cost integration model for each business type based on the basic model of each sample set and the target weights of each basic model. The historical and current feature data include counterparty type, counterparty location region, final counterparty type, final counterparty location region, transaction amount, transaction currency, interest rate floating type, interest rate benchmark type, transaction accrual date, transaction maturity date, transaction cycle period, preset fee rate, and transaction purpose. The target weight determination module is used to perform the following iterative process: By inputting the historical feature data of each sampling set into the corresponding basic model, the basic transfer cost prediction value of each basic model is obtained; When the current iteration count reaches the preset iteration count, the preset weight of each basic model is determined as the target weight of each basic model; otherwise, the preset weight is updated according to the deviation between the basic transfer cost prediction value of each basic model and the corresponding historical business transfer cost, and the iteration process continues. Let h be the deviation corresponding to the i-th basic model. i There are T basic models in total. The preset weights of the i-th basic model in the next iteration are updated as follows: Among them, W i The preset weights are the updated weights for the i-th basic model.

3. A computer device, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the abnormal service identification method according to claim 1.

4. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the abnormal service identification method according to claim 1.

Images