Business category identification method, device, equipment, storage medium and program product

Abstract

The present application relates to the field of big data technology, and in particular to a method, apparatus, device, storage medium, and program product for identifying business categories. The method comprises: obtaining multi-dimensional business operation data generated by a target user when processing financial business; inputting the multi-dimensional business operation data into a business category identification model to obtain a business risk category identification result of the financial business; the business category identification model comprises a learner module composed of a first type learner and a second type learner cascaded; wherein the first type learner is used to identify the acquired input data according to a first classification algorithm; and the second type learner is used to identify the acquired input data according to a second classification algorithm. The present application can accurately and efficiently identify the category of financial business.

Description

Technical Field

[0001] The present application relates to the field of big data technology, and in particular to a business category identification method, apparatus, device, storage medium, and program product. Background Art

[0002] With the development of internet technology and financial services, financial transactions are becoming increasingly convenient. Users can easily access online tools to conduct financial transactions. For example, they can apply for credit cards online and complete transactions such as credit card repayments online. However, while internet technology offers significant convenience to users, it also leads to a large number of illegal business operations, resulting in significant security risks when conducting financial transactions online.

[0003] Traditional technologies mainly rely on statistical methods, machine learning methods, etc. to identify risks in financial services on the network. However, traditional technologies are unable to accurately and efficiently identify risks for financial business processing data with huge data volumes and uneven data quality. Summary of the Invention

[0004] Based on this, it is necessary to provide a business category identification method, device, equipment, storage medium and program product that can accurately and efficiently identify business categories in response to the above technical problems.

[0005] In a first aspect, the present application provides a method for identifying a service category. The method comprises:

[0006] Obtain multi-dimensional business operation data generated by target users when handling financial business;

[0007] Inputting multi-dimensional business operation data into a preset business category identification model to obtain a business risk category identification result for the financial business; the business category identification model includes a learner module consisting of a first type learner and a second type learner in cascade;

[0008] The first type of learner is used to identify the acquired input data according to the first classification algorithm; the second type of learner is used to identify the acquired input data according to the second classification algorithm.

[0009] In one embodiment, the business category identification model further includes a feature extraction module, and the multi-dimensional business operation data is input into the business category identification model to obtain a business risk category identification result of the financial business, including:

[0010] Input the multi-dimensional business operation data into the business category recognition model, and perform multi-granularity scanning on the multi-dimensional business operation data through the feature extraction module to extract the operation data feature vector from the multi-dimensional business operation data;

[0011] The operational data feature vector is input into the learner module to obtain the business risk category identification result of the financial business.

[0012] In one embodiment, before inputting the multi-dimensional business operation data into a preset business category identification model, the method further includes:

[0013] Obtain the initial business operation data generated by the target user when processing financial business;

[0014] The initial business operation data is preprocessed to obtain multi-dimensional business operation data.

[0015] In one embodiment, preprocessing the initial business operation data to obtain multi-dimensional business operation data includes:

[0016] Verify the integrity of initial business operation data;

[0017] If the integrity verification of the initial business operation data passes, the initial business operation data is normalized to obtain multi-dimensional business operation data.

[0018] In one embodiment, the process of building a business category identification model includes:

[0019] Obtain a construction sample set and a test sample set, each of which contains sample multi-dimensional business operation data of multiple sample financial businesses and a business risk category label for each sample financial business;

[0020] Determine the training sample set according to the construction sample set and the test sample set;

[0021] The training sample set is input into the initial business category recognition model, and the initial business category recognition model is trained until the initial business category recognition model meets the preset convergence condition, thereby obtaining the business category recognition model.

[0022] In one embodiment, the initial business category identification model includes an initial feature extraction module and an initial learner module;

[0023] The training sample set is input into the initial service category identification model, and the initial service category identification model is trained until the initial service category identification model meets the preset convergence conditions, thereby obtaining a service category identification model, including:

[0024] Input the test sample set into the initial feature extraction module to obtain the test operation data feature vector, and input the test operation data feature vector into the initial learner module to obtain the test business risk identification result;

[0025] If the total number of learners in the initial learner module reaches a preset number, and the error between the test business risk identification result and the standard business risk identification result is less than a preset value, it is determined that the initial business category identification model meets the convergence condition and the business category identification model is obtained.

[0026] In one embodiment, the initial learner module includes a first type initial learner and at least one second type initial learner cascaded after the first type initial learner;

[0027] Input the test operation data feature vector into the initial learner module to obtain the test business risk identification results, including:

[0028] The test operation data feature vector is input into the first type initial learner to obtain the first test business risk identification result, and the test operation data feature vector and the first test business risk identification result are input into the first second type initial learner to obtain the second test business risk identification result, and so on, to obtain the test business risk identification result output by the last second type initial learner.

[0029] In one embodiment, the business risk category label includes a positive sample label and a negative sample label;

[0030] The training sample set is determined based on the construction sample set and the test sample set, including:

[0031] Expanding the construction sample set to obtain an updated construction sample set;

[0032] Determine a training sample set according to the updated construction sample set and test sample set;

[0033] The construction sample set is expanded to obtain an updated construction sample set, including:

[0034] Extracting first sample multi-dimensional business operation data with positive sample labels and second sample multi-dimensional business operation data with negative sample labels from each sample multi-dimensional business operation data according to the business risk category label of each sample financial business in the constructed sample set;

[0035] Repeating the sample screening step multiple times to obtain an updated constructed sample set; the number of the first sample multi-dimensional business operation data in the expanded constructed sample set is the same as the number of the second sample multi-dimensional business operation data;

[0036] The sample screening step is to extract the same amount of second sample multi-dimensional business operation data as the first sample multi-dimensional business operation data from the second sample multi-dimensional business operation data according to the amount of the first sample multi-dimensional business operation data.

[0037] In a second aspect, the present application also provides a service category identification device. The device includes:

[0038] The acquisition module is used to obtain the multi-dimensional business operation data generated by the target user when processing financial business;

[0039] An identification module is used to input multi-dimensional business operation data into a preset business category identification model to obtain a business risk category identification result for the financial business; the business category identification model includes a learner module consisting of a cascade of a first type learner and a second type learner;

[0040] The first type of learner is used to identify the acquired input data according to the first classification algorithm; the second type of learner is used to identify the acquired input data according to the second classification algorithm.

[0041] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, the following steps are performed:

[0042] Obtain multi-dimensional business operation data generated by target users when handling financial business;

[0043] Inputting multi-dimensional business operation data into a preset business category identification model to obtain a business risk category identification result for the financial business; the business category identification model includes a learner module consisting of a first type learner and a second type learner in cascade;

[0044] The first type of learner is used to identify the acquired input data according to the first classification algorithm; the second type of learner is used to identify the acquired input data according to the second classification algorithm.

[0045] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the following steps:

[0046] Obtain multi-dimensional business operation data generated by target users when handling financial business;

[0047] Inputting multi-dimensional business operation data into a preset business category identification model to obtain a business risk category identification result for the financial business; the business category identification model includes a learner module consisting of a first type learner and a second type learner in cascade;

[0048] The first type of learner is used to identify the acquired input data according to the first classification algorithm; the second type of learner is used to identify the acquired input data according to the second classification algorithm.

[0049] In a fifth aspect, the present application further provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the following steps:

[0050] Obtain multi-dimensional business operation data generated by target users when handling financial business;

[0051] Inputting multi-dimensional business operation data into a preset business category identification model to obtain a business risk category identification result for the financial business; the business category identification model includes a learner module consisting of a first type learner and a second type learner in cascade;

[0052] The first type of learner is used to identify the acquired input data according to the first classification algorithm; the second type of learner is used to identify the acquired input data according to the second classification algorithm.

[0053] The above-mentioned business category identification method, device, equipment, storage medium and program product retain the information richness and integrity of financial business data by obtaining the multi-dimensional business operation data generated by the target user when processing financial business; the first type learner and the second type learner analyze the multi-dimensional business operation data with different classification algorithms, and obtain classification results in different dimensions, so that the business risk category identification results are more accurate; the input of the latter layer of the cascade structure is the output of the previous layer of the cascade structure, so that the processing information of each cascade layer can be preserved as much as possible, and the classification and recognition advantages of each learner are combined to further improve the recognition accuracy of the business category identification model. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 A diagram illustrating an application environment of a service category identification method according to an embodiment;

[0055] Figure 2 1 is a flow chart of a method for preprocessing multi-dimensional business operation data in one embodiment;

[0056] Figure 3 Schematic diagram of a process for extracting features from multi-dimensional business operation data in one embodiment;

[0057] Figure 4 A flowchart illustrating steps of a multi-dimensional business operation data preprocessing method according to an embodiment;

[0058] Figure 5 A schematic flow chart of a method for constructing an initial business category identification model in another embodiment;

[0059] Figure 6 A flowchart of a method for training a model based on a test sample set in another embodiment;

[0060] Figure 7 A schematic flow chart of a method for expanding and constructing a sample set in another embodiment;

[0061] Figure 8 is a structural block diagram of a service category identification device in one embodiment;

[0062] Figure 9 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0064] The service category identification method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store data that server 104 needs to process. The data storage system can be integrated with server 104, or placed on a cloud or other network server. Terminal 102 and server 104 can each be used independently to perform the business category identification method, or terminal 102 and server 104 can jointly perform the business category identification method via a network. For example, terminal 102 obtains multi-dimensional business operation data generated by a target user when processing financial business, inputs the multi-dimensional business operation data into a preset business category identification model, and obtains a business risk category identification result for the financial business. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart car devices, etc. Portable wearable devices can include smart watches, smart bracelets, head-mounted devices, etc. Server 104 can be implemented as a standalone server or a server cluster consisting of multiple servers.

[0065] With the development of internet technology and financial services, financial transactions are becoming increasingly convenient. Users can easily access online application tools and conduct financial transactions. For example, they can apply for credit cards online and complete transactions such as credit card repayments online. However, while internet technology brings great convenience to users, it also leads to a large number of illegal transactions online, such as overdue credit card payments and malicious overdrafts. Assessing the security risks of these transactions based on transaction data to predict whether a user is likely to default is an important way to improve the security of financial transactions.

[0066] Traditionally, a large number of business operation risk assessments are completed through manual review. However, as the amount of financial business data continues to grow, in order to further improve efficiency, methods for identifying financial business risks based on statistical methods and machine learning have been developed. However, traditional technologies are unable to accurately and efficiently identify risks in financial business processing data with huge data volumes and uneven data quality.

[0067] In one embodiment, Figure 2 As shown, a business category identification method is provided, which is applied to Figure 1 Taking the server 104 in the example as an example, the following steps are included:

[0068] Step 202: Acquire multi-dimensional business operation data generated by the target user when processing financial business.

[0069] Among them, the multi-dimensional business operation data generated by the target user when processing financial business includes the current basic information of the target user and the operation records generated by the target user when processing financial business; the multi-dimensional business operation data in the embodiment of the present application refers to a one-dimensional sequence data composed of multiple dimensional features in the current basic information of the target user and multiple dimensional features in the operation records generated when the target user processes financial business, which are arranged in sequence.

[0070] Specifically, the user's current basic information refers to the user information of multiple dimensions collected in advance, which is used to assess the user's risk level; for example, age, marital status, years of employment, nature of work, employment status, housing status, income status, number of existing cards, proportion of disposable income, credit card guarantee amount, credit history, credit record, transaction level, loan status, recent consumption record, etc. For example, credit card use adopts the method of overdraft consumption first and repayment later. When the applicant has financial problems, has no repayment ability or intentionally defaults on repayment on time or the company goes bankrupt and cannot repay, it will cause economic losses to the issuing bank. Such risks constitute user credit risk. It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0071] Specifically, the operation records generated when a user handles financial business refer to the business operation record information of multiple dimensions collected from the business client (business system). This information is used to assess the risk level of the financial business operation currently being handled by the user; for example, transaction time, transaction success or failure, transaction account, transaction frequency, transaction quantity, transaction type, transaction channel, etc. For example, the use of counterfeit credit cards, the fraudulent use of other people's credit cards (such as cards lost and then fraudulently claimed and used), the theft of credit cards (such as cards stolen in the mail, card accounts stolen), card-absence fraud (i.e., using Trojan horse technology to steal payment information or trick cardholders into paying for fraud), such risks constitute operational risks. It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0072] Step 204 : Input the multi-dimensional business operation data into a preset business category identification model to obtain a business risk category identification result of the financial business.

[0073] Among them, the business category identification model includes a learner module composed of a cascade of a first type learner and a second type learner; the first type learner is used to identify the acquired input data according to a first classification algorithm; the second type learner is used to identify the acquired input data according to a second classification algorithm.

[0074] The first type of learner can be a single base classifier or a strong learner ensembled from multiple base classifiers, which can be integrated in parallel. The second type of learner can also be a single base classifier or a strong learner ensembled from multiple base classifiers, which can be integrated in parallel. Examples of base classifiers include bagging learners and boosting learners. Bagging learners include random forests, while boosting learners include Xgboost, Catboost, Adaboost, and others. Both bagging and boosting learners are based on decision trees.

[0075] Among them, each base classifier includes multiple decision trees. The decision trees in the same type of base classifier are generated according to the same rules. Each decision tree continuously divides the feature space during the training process, which can grasp the internal details of the data more deeply. In the same base classifier, the decision trees generated based on different input samples or different selected sample feature attributes are also different; the base classifier obtains a comprehensive classification / prediction result by integrating the outputs of each decision tree, so that the classification / prediction effect of the base classifier exceeds the classification / prediction result of a single decision tree. In the process of training (generating) each decision tree, the randomness of sample selection and feature selection further increases the objectivity of the classification / prediction process and improves the recognition accuracy of the business category recognition model.

[0076] Specifically, the first classification algorithm corresponding to the first type learner is determined based on the base classifiers constituting the first type learner, and the second classification algorithm corresponding to the second type learner is determined based on the base classifiers constituting the second type learner. The first type learner and the second type learner are configured in different forms, including the following three configurations: 1) the base classifiers constituting the first type learner and the second type learner are of different types; 2) the number of base classifiers constituting the first type learner and the second type learner is different; 3) the type and number of base classifiers constituting the first type learner and the second type learner are the same, but the training processes are different, where the different training processes include different training samples, or different feature attributes of the selected training samples.

[0077] Therefore, after obtaining the multi-dimensional business operation data, the first type learner uses the first classification algorithm to learn the multi-dimensional business operation data to generate a first business risk category identification result; the second type learner uses the second classification algorithm to learn the multi-dimensional business operation data to generate a second business risk category identification result; the final generated business risk category identification result of the financial business is determined by combining the first business risk category identification result and the first business risk category identification result. Through different classification algorithms, the multi-dimensional business operation data is identified in different dimensions, which improves the accuracy of the business category identification model in identifying financial business.

[0078] The above-mentioned business category identification method retains the information richness and integrity of financial business data by obtaining the multi-dimensional business operation data generated by the target user when processing financial business; the first type learner and the second type learner analyze the multi-dimensional business operation data with different classification algorithms, and obtain classification results in different dimensions, so that the business risk category identification results are more accurate; the input of the latter layer of the cascade structure is the output of the previous layer of the cascade structure, so that the processing information of each cascade layer can be preserved as much as possible, and the classification and recognition advantages of each learner are combined to further improve the recognition accuracy of the business category identification model.

[0079] For multi-dimensional business operation data in the form of a one-dimensional sequence, the accuracy of the business category identification model is not only related to the input data features that can be effectively processed by the algorithms in each learner, but also to the order of each feature in the input data. Therefore, in one embodiment, the business category identification model also includes a feature extraction module. In step 204, the multi-dimensional business operation data is input into the business category identification model to obtain the business risk category identification results of the financial business, including:

[0080] The multi-dimensional business operation data is input into the business category identification model, and the multi-dimensional business operation data is scanned at multiple granularities through the feature extraction module to extract the operation data feature vector from the multi-dimensional business operation data; the operation data feature vector is input into the learner module to obtain the business risk category identification result of the financial business.

[0081] Multi-granularity scanning leverages the concept of convolutional neural networks and employs a sliding window to extract feature vectors. Assuming the multi-dimensional business operation data has a dimension of 300 and the sliding window dimension is 100, the sliding window is first used to sample one dimension of the multi-dimensional business operation data with a sampling step of 1. This generates 201 scan sub-input data. The scan sub-input data is then trained using Random Forest A and Random Forest B, respectively. For a binary classification task, each scan sub-input data generates a 2-dimensional probabilistic feature vector after passing through Random Forest A and Random Forest B. Thus, a total of 804-dimensional probabilistic feature vectors are obtained from the 201 scan sub-input data. Finally, these 804-dimensional probabilistic feature vectors are concatenated to form the data feature vector corresponding to the multi-dimensional business operation data. Furthermore, the length of the sliding window is adjustable, and multiple windows of different lengths can be used simultaneously for feature extraction. That is, multi-granularity scanning can generate probabilistic feature vectors of different granularities, which contain richer features and realize deep processing of the original input data. It not only expands the data dimension, but also increases the forward relationship between the input data, providing strong support for the classification and recognition of the subsequent cascaded first-type learners and second-type learners.

[0082] When acquiring multi-dimensional business operation data, the data types (dimensions) are varied and the data sources are also different. Therefore, when using the acquired data for data mining, there will be a lot of interference. Data preprocessing refers to cleaning the data before using it for modeling or data mining to ensure that the data can meet the input requirements of the model. Therefore, if Figure 3 As shown, before step 202 of acquiring the multi-dimensional business operation data generated by the target user when processing financial business, the method further includes step 201,

[0083] Step 201: Acquire initial business operation data generated by a target user when processing financial business; pre-process the initial business operation data to obtain multi-dimensional business operation data.

[0084] Among them, data preprocessing methods include screening, validity verification, normalization and other operations.

[0085] In one embodiment, Figure 4 As shown, step 201 pre-processes the initial business operation data to obtain multi-dimensional business operation data including:

[0086] Step 2011: Verify the integrity of the initial business operation data.

[0087] Among them, the original data from web crawling or databases may have varying degrees of missing data. In this case, it is necessary to delete or fill in the data items with missing values. When there is too much missing data or the original data volume is small, the data items with missing values cannot be directly deleted. When missing values need to be filled, there are many ways to fill the data, including median filling, mode filling, or filling using linear regression. In this embodiment, the initial business operation data with missing dimensions greater than a first threshold is deleted as incomplete data. Otherwise, it is retained as complete data. Specifically, the first threshold can be an empirical value.

[0088] Step 2012: If the integrity verification of the initial business operation data passes, the initial business operation data is normalized to obtain multi-dimensional business operation data.

[0089] Among them, data normalization (also known as data standardization) is the basic work of the pattern recognition system, because data from different sources often have different dimensions and units. If these data are directly used as input for subsequent modules, it may affect the efficiency and accuracy of data analysis. Therefore, data normalization operations are required to eliminate the problem of comparability between data. After the original data is normalized, all data points are in the same order of magnitude, which is more conducive to subsequent feature extraction and enables the subsequent business category recognition model to converge to the optimal solution faster. Specifically, the normalization process is to reduce the data to a specific threshold according to a certain ratio. The normalization process in the embodiment of the present application uses the maximum and minimum normalization method.

[0090] In one embodiment, Figure 5 As shown in FIG, the construction process of the business category identification model includes:

[0091] Step 300: Obtain a construction sample set and a test sample set.

[0092] The construction sample set and the test sample set both contain sample multi-dimensional business operation data for multiple sample financial businesses and business risk category labels for each sample financial business. The sample multi-dimensional business operation data for the multiple sample financial businesses can be open source data collected from multiple different users.

[0093] Step 302: Determine a training sample set based on the construction sample set and the test sample set.

[0094] Among them, the training sample set is constructed to train the learners in each cascade layer in the learner module, and the test training sample set is used to train the number of cascade layers of the learners in the learner module.

[0095] Step 304: input the training sample set into the initial service category recognition model, train the initial service category recognition model until the initial service category recognition model meets the preset convergence condition, and obtain the service category recognition model.

[0096] In one embodiment, Figure 6 As shown, the initial service category identification model includes an initial feature extraction module and an initial learner module; then step 304 inputs the training sample set into the initial service category identification model, trains the initial service category identification model until the initial service category identification model meets the preset convergence condition, and obtains the service category identification model, including:

[0097] Step 3041: Input the test sample set into the initial feature extraction module to obtain a test operation data feature vector, and input the test operation data feature vector into the initial learner module to obtain a test business risk identification result.

[0098] Among them, each sample in the test sample set is subjected to feature extraction by the initial feature extraction module to obtain the corresponding test sample operation data feature vector. The initial feature extraction module also uses the above-mentioned multi-granularity scanning feature extraction method to extract features from each sample.

[0099] Step 3042: If the total number of learners in the initial learner module reaches a preset number, and the error between the test business risk identification result and the standard business risk identification result is less than a preset value, it is determined that the initial business category identification model meets the convergence condition and a business category identification model is obtained.

[0100] Among them, for the two-class business category identification, the output result of the initial business category identification model is the classification result of whether the input data belongs to the preset first category or second category, and the corresponding business risk category labels include positive samples (fraud data) and negative samples (normal data); for the two-class prediction task, the output result of the initial business category identification model is the probability values of the input data belonging to each preset category, and the corresponding business risk category labels include positive sample probability interval (fraud data) and negative sample probability interval (normal data).

[0101] Specifically, after constructing each learner, each time a layer of cascade learners is added, the initial business category recognition model is tested with a test sample set. When the performance of the initial business category recognition model does not increase significantly, the cascade structure of the initial learner module automatically stops growing. For example, if the performance does not improve significantly within three layers of the cascade structure expansion, training is stopped. In this embodiment, the initial business category recognition model of the cascade structure adaptively determines its model complexity by terminating training at the appropriate time, which can prevent the model from overfitting and make the business category recognition model applicable to training data of different sizes, rather than being limited to large-scale training data.

[0102] In one embodiment, the initial learner module includes a first type initial learner and at least one second type initial learner cascaded after the first type initial learner. Step 3041 inputs the test operation data feature vector into the initial learner module to obtain a test business risk identification result, including:

[0103] The test operation data feature vector is input into the first type initial learner to obtain the first test business risk identification result, and the test operation data feature vector and the first test business risk identification result are input into the first second type initial learner to obtain the second test business risk identification result, and so on, to obtain the test business risk identification result output by the last second type initial learner.

[0104] Among them, the first test business risk identification result refers to the probability feature vector output by the first type of initial learner; for the binary classification task, if the first type of initial learner contains N1 base classifiers, and each base classifier outputs a probability vector of length 2, then the first type of initial learner outputs a concatenated probability feature vector of length 2*N1.

[0105] Set the feature number of the test operation data feature vector to M, then splice the test operation data feature vector with the feature vector of the first type initial learner, and input the probability feature vector with a length of M+2*N1 into the first second type initial learner; if the second type initial learner contains N2 base classifiers, for the binary classification task, the second type initial learner outputs the spliced probability feature vector with a length of 2*N2, that is, the second test business risk identification result.

[0106] The test operation data feature vector of length M is spliced with the probability vector of length 2*N2, and inputted together into the next second-type initial learner. Similarly, the test business risk identification result output by the last second-type initial learner is obtained as the test business risk identification result of the initial learner module.

[0107] Specifically, assuming that the last Type II initial learner has N3 base classifiers, the test business risk identification result output by the last Type II initial learner is a probability feature vector of length 2*N3. Taking N3 = 3 as an example, that is, the last Type II initial learner contains three base classifiers, if the probability vector generated by the first base classifier of the last Type II initial learner is [0.3, 0.7], the probability vector generated by the second base classifier of the last Type II initial learner is [0.6, 0.4], and the probability vector generated by the third base classifier of the last Type II initial learner is [0.8, 0.2], then the probability feature vector generated by the last Type II initial learner is [(0.3+0.6+0.8) / 3, (0.7+0.4+0.3) / 3], that is, [0.56, 0.44].

[0108] In this embodiment, during training, different first-type initial learners and second-type initial learners predict the category to which each sample multi-dimensional business operation data belongs based on their own model characteristics in the feature space. Each cascade layer in the cascade structure within the initial learner module receives the classification result processed by the learner in the upper cascade layer. In each cascade layer learner, in addition to being able to obtain the results of the classification algorithm of the learner at that layer for identifying the sample multi-dimensional business operation data, it can also integrate the recognition results output by the learner in the upper cascade layer, thereby making the integrated classification result obtained by the learner in the last cascade layer more accurate.

[0109] In fraudulent transaction identification, the vast majority of transactions are normal, and only a very small number of transactions are fraudulent transactions. Under this unbalanced classification, model prediction will extract less effective information from small positive samples (fraudulent transaction data), and over-reliance on data samples will lead to overfitting. Therefore, it is necessary to balance the positive samples (fraudulent transaction data) and negative samples (normal transaction data) in the training sample sets input to the first type initial learner and the second type initial learner. Therefore, in one embodiment, the business risk category label includes a positive sample label and a negative sample label. Step 302 determines the training sample set based on the constructed sample set and the test sample set, including: expanding the constructed sample set to obtain an updated constructed sample set; determining the training sample set based on the updated constructed sample set and the test sample set;

[0110] Among them, such as Figure 7 As shown, the construction sample set is expanded to obtain an updated construction sample set, including:

[0111] Step 3021 , based on the business risk category label of each sample financial business in the constructed sample set, extract first sample multi-dimensional business operation data with positive sample labels and second sample multi-dimensional business operation data with negative sample labels from each sample multi-dimensional business operation data.

[0112] Step 3022 , repeatedly performing the sample screening step multiple times to obtain an updated constructed sample set; the number of the first sample multi-dimensional business operation data in the expanded constructed sample set is the same as the number of the second sample multi-dimensional business operation data.

[0113] The sample screening step is to extract the same amount of second sample multi-dimensional business operation data as the first sample multi-dimensional business operation data from the second sample multi-dimensional business operation data according to the amount of the first sample multi-dimensional business operation data.

[0114] Specifically, the first sample multi-dimensional business operation data and the second sample multi-dimensional business operation data in each screening step are used as training sample subsets. The training sample subsets are used to train each base classifier in the first type of learner. Specifically, any base classifier is trained based on the training sample subset, including: Sa, for the training sample subset training, a greedy algorithm is used, starting from the root node, and the optimal features of the tree structure are recursively selected. The training sample subset training is divided according to this feature until the pre-set division stop condition (such as the maximum depth of the tree, etc.) is reached. At this point, all samples are assigned to leaf nodes to generate a decision tree. Sb, Repeat Sa, and establish multiple decision trees in sequence in the direction of the gradient descent of the loss function. Sc, Combine multiple decision trees to form a base classifier.

[0115] In this embodiment, positive samples refer to fraud samples, and negative samples refer to normal transaction samples. Repeated random sampling is used to achieve the purpose of balancing samples and improve the reuse rate of positive samples. The model trained in this way is more adaptable, and when another data set is used for classification prediction, the prediction effect of the model is more stable.

[0116] In one embodiment, inputting the training sample set into the initial service category recognition model includes: preprocessing the training sample set to obtain a preprocessed training sample set; and inputting the preprocessed training sample set into the initial service category recognition model.

[0117] Among them, preprocessing operations include missing value processing, normalization processing, outlier processing, collinearity processing, etc. Outliers refer to data in a certain variable that are significantly different from other data. Such data are also called outliers. There are many reasons for the occurrence of outliers. It may be an error in data collection, where data from other variables are collected as data for a certain variable. It may also be due to human negligence and changes to the original data. Collinearity refers to the existence of a high linear correlation between explanatory variables. The collinearity between explanatory variables can be divided into two cases: complete collinearity and incomplete collinearity. The reason for collinearity may be the sample data itself, or it may be that the variables do have a common trend of change. The existence of collinearity may lead to model failure or inaccurate model predictions. Therefore, collinearity detection is required before data mining.

[0118] When training the initial business category recognition model, first obtain the initial construction sample set and the test sample set, expand the initial construction sample set, and input the updated construction sample set and the test sample set as training sample sets into the initial business category recognition model. The construction sample set is used to train each learner in the initial learner module in the initial business category recognition model, so that each learner can learn and classify the training samples based on different base classifiers. The test sample set is used to construct a cascade structure of each learner in the initial learner module. If the performance is not significantly improved within the cascade structure expansion of 3 layers, the training is stopped to obtain a trained business category recognition model.

[0119] When applying the business category identification model, the multi-dimensional business operation data generated by the target user when handling financial business is obtained; the initial business operation data generated by the target user when handling financial business is obtained; the initial business operation data is pre-processed to obtain the multi-dimensional business operation data. The integrity of the initial business operation data is verified; if the integrity verification of the initial business operation data is passed, the initial business operation data is normalized to obtain the multi-dimensional business operation data. The multi-dimensional business operation data is input into the business category identification model to obtain the business risk category identification results of the financial business.

[0120] It should be understood that, although the steps in the flowcharts of the above embodiments are shown in sequence as indicated by the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be performed in other orders. Moreover, at least a portion of the steps in the flowcharts of the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily performed at the same time, but can be performed at different times. The execution order of these steps or stages is not necessarily to be performed in sequence, but can be performed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0121] Based on the same inventive concept, embodiments of the present application also provide a service category identification device for implementing the aforementioned service category identification method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations in the one or more service category identification device embodiments provided below can be found in the aforementioned limitations on the service category identification method and will not be further elaborated here.

[0122] In one embodiment, Figure 8 As shown, a service category identification device 100 is provided, comprising: an acquisition module 110 and an identification module 120, wherein:

[0123] The acquisition module 110 is used to acquire multi-dimensional business operation data generated by the target user when processing financial business;

[0124] Identification module 120, used to input multi-dimensional business operation data into a preset business category identification model to obtain business risk category identification results for financial businesses; the business category identification model includes a learner module consisting of a cascade of a first type learner and a second type learner;

[0125] The first type of learner is used to identify the acquired input data according to the first classification algorithm; the second type of learner is used to identify the acquired input data according to the second classification algorithm.

[0126] In one embodiment, the business category identification model further includes a feature extraction module. The identification module 120 includes:

[0127] An input unit, configured to input the multi-dimensional business operation data into the business category recognition model, and perform multi-granularity scanning on the multi-dimensional business operation data through a feature extraction module to extract an operation data feature vector from the multi-dimensional business operation data;

[0128] The analysis unit is used to input the operation data feature vector into the learner module to obtain the business risk category identification result of the financial business.

[0129] In one embodiment, the recognition device further includes a pre-processing module, including:

[0130] A collection unit, used to obtain the initial business operation data generated by the target user when processing financial business;

[0131] The processing unit is used to pre-process the initial business operation data to obtain multi-dimensional business operation data.

[0132] In one embodiment, the processing unit is further configured to verify the integrity of the initial business operation data;

[0133] If the integrity verification of the initial business operation data passes, the initial business operation data is normalized to obtain multi-dimensional business operation data.

[0134] In one embodiment, the recognition device further comprises a building module, the building module comprising:

[0135] A sample storage unit is used to obtain a construction sample set and a test sample set, where the construction sample set and the test sample set both contain sample multi-dimensional business operation data of multiple sample financial businesses and a business risk category label for each sample financial business;

[0136] An expansion unit, configured to determine a training sample set based on a construction sample set and a test sample set;

[0137] The training unit is used to input the training sample set into the initial business category recognition model, train the initial business category recognition model until the initial business category recognition model meets the preset convergence condition, and obtain the business category recognition model.

[0138] In one embodiment, the training unit includes:

[0139] The first training subunit is used to input the test sample set into the initial feature extraction module to obtain the test operation data feature vector, and input the test operation data feature vector into the initial learner module to obtain the test business risk identification result;

[0140] The second training subunit is used to determine that the initial business category identification model meets the convergence condition and obtain the business category identification model if the total number of learners in the initial learner module reaches a preset number and the error between the test business risk identification result and the standard business risk identification result is less than a preset value.

[0141] In one embodiment, the initial learner module includes a first type initial learner and at least one second type initial learner cascaded after the first type initial learner;

[0142] The first training sub-unit is also used to input the test operation data feature vector into the first type initial learner to obtain a first test business risk identification result, and input the test operation data feature vector and the first test business risk identification result into the first second type initial learner to obtain a second test business risk identification result, and so on, to obtain the test business risk identification result output by the last second type initial learner.

[0143] In one embodiment, the business risk category label includes a positive sample label and a negative sample label; the expansion unit is further used to expand the constructed sample set to obtain an updated constructed sample set; determine the training sample set based on the updated constructed sample set and the test sample set; the expansion unit is further used to:

[0144] Extracting first sample multi-dimensional business operation data with positive sample labels and second sample multi-dimensional business operation data with negative sample labels from each sample multi-dimensional business operation data according to the business risk category label of each sample financial business in the constructed sample set;

[0145] Repeating the sample screening step multiple times to obtain an updated constructed sample set; the number of the first sample multi-dimensional business operation data in the expanded constructed sample set is the same as the number of the second sample multi-dimensional business operation data;

[0146] The sample screening step is to extract the same amount of second sample multi-dimensional business operation data as the first sample multi-dimensional business operation data from the second sample multi-dimensional business operation data according to the amount of the first sample multi-dimensional business operation data.

[0147] Each module in the above-mentioned service category identification device can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in the computer device in the form of software, so that the processor can call and execute the corresponding operations of each module.

[0148] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 9As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a service category identification method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad provided on the computer device housing, or an external keyboard, touchpad or mouse.

[0149] Those skilled in the art will understand that Figure 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0150] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0151] Obtain multi-dimensional business operation data generated by target users when handling financial business;

[0152] Inputting multi-dimensional business operation data into a preset business category identification model to obtain a business risk category identification result for the financial business; the business category identification model includes a learner module consisting of a first type learner and a second type learner in cascade;

[0153] The first type of learner is used to identify the acquired input data according to the first classification algorithm; the second type of learner is used to identify the acquired input data according to the second classification algorithm.

[0154] In one embodiment, when the processor executes the computer program, it also implements the following steps: inputting the multi-dimensional business operation data into the business category identification model, performing multi-granularity scanning on the multi-dimensional business operation data through the feature extraction module to extract the operation data feature vector from the multi-dimensional business operation data; inputting the operation data feature vector into the learner module to obtain the business risk category identification result of the financial business.

[0155] In one embodiment, when the processor executes the computer program, it further implements the following steps: obtaining initial business operation data generated by the target user when processing financial business; and preprocessing the initial business operation data to obtain multi-dimensional business operation data.

[0156] In one embodiment, when the processor executes the computer program, it further implements the following steps: verifying the integrity of the initial business operation data; if the integrity verification of the initial business operation data passes, normalizing the initial business operation data to obtain multi-dimensional business operation data.

[0157] In one embodiment, when the processor executes the computer program, it also implements the following steps: obtaining a construction sample set and a test sample set, where the construction sample set and the test sample set both contain sample multi-dimensional business operation data of multiple sample financial businesses, and a business risk category label for each sample financial business; determining a training sample set based on the construction sample set and the test sample set; inputting the training sample set into the initial business category identification model, training the initial business category identification model until the initial business category identification model meets the preset convergence conditions, and obtaining a business category identification model.

[0158] In one embodiment, when the processor executes the computer program, it also implements the following steps: inputting the test sample set into the initial feature extraction module to obtain a test operation data feature vector, and inputting the test operation data feature vector into the initial learner module to obtain a test business risk identification result; if the total number of learners in the initial learner module reaches a preset number, and the error between the test business risk identification result and the standard business risk identification result is less than a preset value, it is determined that the initial business category identification model meets the convergence condition, and a business category identification model is obtained.

[0159] In one embodiment, when the processor executes the computer program, it also implements the following steps: testing the business risk identification result, and inputting the test operation data feature vector and the first test business risk identification result into the first second-type initial learner to obtain the second test business risk identification result, and so on, to obtain the test business risk identification result output by the last second-type initial learner.

[0160] In one embodiment, when the processor executes the computer program, it also implements the following steps: expanding the constructed sample set to obtain an updated constructed sample set; determining the training sample set based on the updated constructed sample set and the test sample set; wherein, expanding the constructed sample set to obtain the updated constructed sample set includes: extracting first sample multidimensional business operation data with positive sample labels and second sample multidimensional business operation data with negative sample labels from each sample multidimensional business operation data based on the business risk category label of each sample financial business in the constructed sample set; repeatedly performing the sample screening step to obtain the updated constructed sample set; the number of the first sample multidimensional business operation data and the number of the second sample multidimensional business operation data in the expanded constructed sample set are the same; wherein, the sample screening step is to extract the second sample multidimensional business operation data, which is the same in number as the first sample multidimensional business operation data, from the second sample multidimensional business operation data based on the number of the first sample multidimensional business operation data.

[0161] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0162] Obtain multi-dimensional business operation data generated by target users when handling financial business;

[0163] Inputting multi-dimensional business operation data into a preset business category identification model to obtain a business risk category identification result for the financial business; the business category identification model includes a learner module consisting of a first type learner and a second type learner in cascade;

[0164] The first type of learner is used to identify the acquired input data according to the first classification algorithm; the second type of learner is used to identify the acquired input data according to the second classification algorithm.

[0165] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented: multi-dimensional business operation data is input into a business category recognition model, and the multi-dimensional business operation data is scanned at multiple granularities through a feature extraction module to extract an operation data feature vector from the multi-dimensional business operation data; and the operation data feature vector is input into a learner module to obtain a business risk category recognition result of the financial business.

[0166] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: obtaining initial business operation data generated by the target user when processing financial business; and preprocessing the initial business operation data to obtain multi-dimensional business operation data.

[0167] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: verifying the integrity of the initial business operation data; if the integrity verification of the initial business operation data passes, normalizing the initial business operation data to obtain multi-dimensional business operation data.

[0168] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: obtaining a construction sample set and a test sample set, each of which contains sample multi-dimensional business operation data of multiple sample financial businesses, and a business risk category label for each sample financial business; determining a training sample set based on the construction sample set and the test sample set; inputting the training sample set into the initial business category identification model, training the initial business category identification model until the initial business category identification model meets the preset convergence conditions, and obtaining a business category identification model.

[0169] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: inputting the test sample set into the initial feature extraction module to obtain the test operation data feature vector, and inputting the test operation data feature vector into the initial learner module to obtain the test business risk identification result; if the total number of learners in the initial learner module reaches a preset number, and the error between the test business risk identification result and the standard business risk identification result is less than a preset value, then it is determined that the initial business category identification model meets the convergence condition, and the business category identification model is obtained.

[0170] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: testing the business risk identification result, and inputting the test operation data feature vector and the first test business risk identification result into the first second-type initial learner to obtain the second test business risk identification result, and so on, to obtain the test business risk identification result output by the last second-type initial learner.

[0171] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: expanding the constructed sample set to obtain an updated constructed sample set; determining the training sample set based on the updated constructed sample set and the test sample set; wherein, expanding the constructed sample set to obtain the updated constructed sample set includes: extracting first sample multidimensional business operation data with positive sample labels and second sample multidimensional business operation data with negative sample labels from each sample multidimensional business operation data according to the business risk category label of each sample financial business in the constructed sample set; repeatedly performing the sample screening step to obtain the updated constructed sample set; the number of the first sample multidimensional business operation data and the number of the second sample multidimensional business operation data in the expanded constructed sample set are the same; wherein, the sample screening step is to extract the second sample multidimensional business operation data with the same number as the first sample multidimensional business operation data from the second sample multidimensional business operation data according to the number of the first sample multidimensional business operation data.

[0172] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0173] Obtain multi-dimensional business operation data generated by target users when handling financial business;

[0174] Inputting multi-dimensional business operation data into a preset business category identification model to obtain a business risk category identification result for the financial business; the business category identification model includes a learner module consisting of a first type learner and a second type learner in cascade;

[0175] The first type of learner is used to identify the acquired input data according to the first classification algorithm; the second type of learner is used to identify the acquired input data according to the second classification algorithm.

[0176] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented: multi-dimensional business operation data is input into a business category recognition model, and the multi-dimensional business operation data is scanned at multiple granularities through a feature extraction module to extract an operation data feature vector from the multi-dimensional business operation data; and the operation data feature vector is input into a learner module to obtain a business risk category recognition result of the financial business.

[0177] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: obtaining initial business operation data generated by the target user when processing financial business; and preprocessing the initial business operation data to obtain multi-dimensional business operation data.

[0178] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: verifying the integrity of the initial business operation data; if the integrity verification of the initial business operation data passes, normalizing the initial business operation data to obtain multi-dimensional business operation data.

[0179] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: obtaining a construction sample set and a test sample set, each of which contains sample multi-dimensional business operation data of multiple sample financial businesses, and a business risk category label for each sample financial business; determining a training sample set based on the construction sample set and the test sample set; inputting the training sample set into the initial business category identification model, training the initial business category identification model until the initial business category identification model meets the preset convergence conditions, and obtaining a business category identification model.

[0180] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: inputting the test sample set into the initial feature extraction module to obtain the test operation data feature vector, and inputting the test operation data feature vector into the initial learner module to obtain the test business risk identification result; if the total number of learners in the initial learner module reaches a preset number, and the error between the test business risk identification result and the standard business risk identification result is less than a preset value, then it is determined that the initial business category identification model meets the convergence condition, and the business category identification model is obtained.

[0181] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: testing the business risk identification result, and inputting the test operation data feature vector and the first test business risk identification result into the first second-type initial learner to obtain the second test business risk identification result, and so on, to obtain the test business risk identification result output by the last second-type initial learner.

[0182] In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: expanding the constructed sample set to obtain an updated constructed sample set; determining the training sample set based on the updated constructed sample set and the test sample set; wherein, expanding the constructed sample set to obtain the updated constructed sample set includes: extracting first sample multidimensional business operation data with positive sample labels and second sample multidimensional business operation data with negative sample labels from each sample multidimensional business operation data according to the business risk category label of each sample financial business in the constructed sample set; repeatedly performing the sample screening step to obtain the updated constructed sample set; the number of the first sample multidimensional business operation data and the number of the second sample multidimensional business operation data in the expanded constructed sample set are the same; wherein, the sample screening step is to extract the second sample multidimensional business operation data with the same number as the first sample multidimensional business operation data from the second sample multidimensional business operation data according to the number of the first sample multidimensional business operation data.

[0183] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0184] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processor involved in the various embodiments provided herein may be, but are not limited to, a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, and the like.

[0185] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0186] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for identifying a service category, characterized in that: The method comprises: Obtain the initial business operation data generated by the target user when processing financial business; Verifying the integrity of the initial business operation data; If the integrity verification of the initial business operation data passes, normalizing the initial business operation data to obtain multi-dimensional business operation data generated by the target user when processing financial business; Inputting the multi-dimensional business operation data into a preset business category identification model to obtain a business risk category identification result of the financial business; the business category identification model includes a learner module consisting of a first type learner and a second type learner in cascade; The first type learner is used to identify the acquired input data according to a first classification algorithm; the second type learner is used to identify the acquired input data according to a second classification algorithm.

2. The method according to claim 1, characterized in that The business category identification model further includes a feature extraction module, which inputs the multi-dimensional business operation data into the business category identification model to obtain a business risk category identification result of the financial business, including: Inputting the multi-dimensional business operation data into the business category identification model, and performing multi-granularity scanning on the multi-dimensional business operation data by the feature extraction module to extract an operation data feature vector from the multi-dimensional business operation data; The operation data feature vector is input into the learner module to obtain a business risk category identification result of the financial business.

3. The method according to claim 1 or 2, characterized in that The process of constructing the business category identification model includes: Obtaining a construction sample set and a test sample set, wherein the construction sample set and the test sample set both include sample multi-dimensional business operation data of a plurality of sample financial businesses and a business risk category label of each of the sample financial businesses; Determine a training sample set according to the constructed sample set and the test sample set; The training sample set is input into an initial service category identification model, and the initial service category identification model is trained until the initial service category identification model meets a preset convergence condition, thereby obtaining the service category identification model.

4. The method according to claim 3, characterized in that The initial business category identification model includes an initial feature extraction module and an initial learner module; The training sample set is input into the initial service category identification model, and the initial service category identification model is trained until the initial service category identification model meets the preset convergence condition, thereby obtaining the service category identification model, including: Inputting the test sample set into the initial feature extraction module to obtain a test operation data feature vector, and inputting the test operation data feature vector into the initial learner module to obtain a test business risk identification result; If the total number of learners in the initial learner module reaches a preset number, and the error between the test business risk identification result and the standard business risk identification result is less than a preset value, it is determined that the initial business category identification model meets the convergence condition and the business category identification model is obtained.

5. The method according to claim 4, characterized in that The initial learner module includes a first type initial learner and at least one second type initial learner cascaded after the first type initial learner; Inputting the test operation data feature vector into the initial learner module to obtain the test business risk identification result includes: The test operation data feature vector is input into the first type initial learner to obtain a first test business risk identification result, and the test operation data feature vector and the first test business risk identification result are input into the first second type initial learner to obtain a second test business risk identification result, and so on to obtain the test business risk identification result output by the last second type initial learner.

6. The method according to claim 3, characterized in that The business risk category labels include positive sample labels and negative sample labels; The determining of the training sample set according to the constructed sample set and the test sample set includes: Expanding the constructed sample set to obtain an updated constructed sample set; Determine the training sample set according to the updated construction sample set and the test sample set; The construction sample set is expanded to obtain an updated construction sample set, including: Extracting first sample multi-dimensional business operation data with positive sample labels and second sample multi-dimensional business operation data with negative sample labels from each of the sample multi-dimensional business operation data according to the business risk category label of each of the sample financial businesses in the constructed sample set; Repeating the sample screening step multiple times to obtain an updated constructed sample set; the number of the first sample multi-dimensional business operation data in the expanded constructed sample set is the same as the number of the second sample multi-dimensional business operation data; The sample screening step is to extract, from the second sample multidimensional business operation data, the same number of second sample multidimensional business operation data as the first sample multidimensional business operation data according to the number of the first sample multidimensional business operation data.

7. A service category identification device, characterized in that: The device comprises: The acquisition module is used to obtain the multi-dimensional business operation data generated by the target user when processing financial business; an identification module, configured to input the multi-dimensional business operation data into a preset business category identification model to obtain a business risk category identification result for the financial business; the business category identification model includes a learner module consisting of a first type learner and a second type learner in cascade; The first type learner is used to identify the acquired input data according to the first classification algorithm; the second type learner is used to identify the acquired input data according to the second classification algorithm; The identification device further includes a pre-processing module, including: a collection unit for acquiring initial business operation data generated by the target user when processing financial business; A processing unit, configured to pre-process the initial business operation data to obtain multi-dimensional business operation data; The processing unit is further used to verify the integrity of the initial business operation data; If the integrity verification of the initial business operation data passes, the initial business operation data is normalized to obtain multi-dimensional business operation data.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images