Risk assessment method and device based on resource transfer, equipment and storage medium

Abstract

The embodiment of the invention provides a risk assessment method and device based on resource transfer, equipment and a storage medium, and belongs to the technical field of financial science and technology. The method comprises the steps of determining candidate evaluation feature data based on a sample resource transfer category and initial evaluation feature data; inputting the candidate evaluation feature data into the initial risk evaluation model, and extracting numerical value type feature data and label type feature data from the candidate evaluation feature data; vectorizing the label type feature data to obtain a label type feature vector, and splicing the label type feature vector with the numerical value type feature data to obtain sample spliced data; performing risk assessment on the sample splicing data based on the risk assessment sub-model, and determining a model loss value based on the obtained prediction object risk category and the sample object risk category so as to construct a target risk assessment model; and performing risk assessment on the target assessment feature data of the target object based on the target risk assessment model. According to the embodiment of the invention, the efficiency of risk assessment based on resource transfer can be improved.

Description

Technical Field

[0001] The present application relates to the field of financial technology, and in particular to a risk assessment method and apparatus, device, and storage medium based on resource transfer. Background Art

[0002] Risk assessment of resource transfer refers to the process of analyzing and evaluating the object requesting resource transfer during the resource transfer process to determine the potential risks and likelihood of occurrence in the resource transfer process. For example, in the loan scenario of FinTech, the object requesting resource transfer can apply for a loan from a bank to ensure its current development. Before transferring resources to the object, the bank can first conduct a risk assessment on whether the object can repay the loan, thereby avoiding certain losses to the bank after the object cannot repay the loan. However, the risk assessment method based on resource transfer provided by the relevant technology usually conducts risk assessment based on all information of the object requesting resource transfer, and uses the Long Short-Term Memory (LSTM) network to generate a time series before conducting risk assessment, resulting in low efficiency of risk assessment. Therefore, how to improve the efficiency of risk assessment based on resource transfer has become a technical problem that needs to be solved urgently. Summary of the invention

[0003] The main purpose of the embodiments of the present application is to propose a risk assessment method and apparatus, device, and storage medium based on resource transfer, aiming to improve the efficiency of risk assessment based on resource transfer.

[0004] To achieve the above-mentioned purpose, a first aspect of an embodiment of the present application proposes a risk assessment method based on resource transfer, the method comprising:

[0005] Acquire a training sample of a sample object, wherein the training sample includes a sample resource transfer category of the sample object, initial assessment feature data, and a sample object risk category, wherein the sample object risk category is used to indicate a risk level of the sample object in performing resource transfer in a scenario corresponding to the sample resource transfer category;

[0006] Performing feature selection on the initial evaluation feature data based on the sample resource transfer category to determine candidate evaluation feature data;

[0007] Inputting the candidate assessment feature data into a preset initial risk assessment model, wherein the initial risk assessment model includes a data extraction layer, a vectorization layer, a splicing layer and a risk assessment sub-model;

[0008] Extracting the candidate evaluation feature data based on the data extraction layer to obtain numerical feature data and label feature data;

[0009] Performing vectorization processing on the label type feature data based on the vectorization layer to obtain a label type feature vector;

[0010] Based on the splicing layer, the numerical type feature data and the label type feature vector are spliced ​​to obtain sample splicing data;

[0011] Performing risk assessment on the sample splicing data based on the risk assessment sub-model to obtain a risk category of the predicted object;

[0012] Determine a model loss value based on the predicted object risk category and the sample object risk category, adjust parameters of the data selection model and the initial risk assessment model based on the model loss value, and construct a target risk assessment model based on the data selection model and the initial risk assessment model after parameter adjustment;

[0013] Target assessment feature data of a target object for which resource transfer is requested is acquired, and risk assessment is performed on the target assessment feature data based on the target risk assessment model.

[0014] In some embodiments, the initial evaluation feature data includes initial evaluation feature sub-data corresponding to the object evaluation feature, and the performing feature selection on the initial evaluation feature data based on the sample resource transfer category to determine the candidate evaluation feature data includes:

[0015] Performing feature evaluation on the initial evaluation feature sub-data and the sample resource transfer category based on a preset feature evaluation model to obtain a feature importance evaluation value of the object evaluation feature, wherein the feature importance evaluation value is used to characterize the importance of the object evaluation feature in the risk assessment scenario corresponding to the sample resource transfer category;

[0016] The initial evaluation feature sub-data is selected based on the feature importance evaluation value and the preset feature quantity to determine the candidate evaluation feature data.

[0017] In some embodiments, the selecting the initial evaluation feature sub-data based on the feature importance evaluation value and the preset feature quantity to determine the candidate evaluation feature data includes:

[0018] Obtaining feature preference weights of the object evaluation features and historical resource transfer scores of the sample object;

[0019] determining a target feature weight for the object assessment feature based on the feature preference weight and the historical resource transfer score;

[0020] Performing weighted calculation on the target feature weight and the feature importance evaluation value to obtain a target evaluation value;

[0021] The initial evaluation feature sub-data is selected based on the target evaluation value and the preset feature quantity to determine the candidate evaluation feature data.

[0022] In some embodiments, the risk assessment sub-model includes a preset number of hidden layers and a normalization layer, and the risk assessment of the sample splicing data based on the risk assessment sub-model to obtain the predicted object risk category includes:

[0023] Extracting data from the sample spliced ​​data based on the preset number of hidden layers to obtain sample hidden data;

[0024] Normalizing the sample hidden data based on the normalization layer to obtain a sample risk assessment probability;

[0025] The object category is divided based on the preset category threshold and the sample risk assessment probability to obtain the predicted object risk category.

[0026] In some embodiments, obtaining a training sample of a sample object includes:

[0027] Acquire an initial sample of initial objects, wherein the initial sample includes a risk category, a resource transfer status, and a resource transfer type of the initial object;

[0028] Selecting the initial object based on a preset time point, the risk category of the initial object, the resource transfer status, and the resource transfer type to determine the sample object;

[0029] Performing data selection on the resource transfer feature data of the sample object to obtain valid feature data;

[0030] Performing data preprocessing on the effective feature data to obtain the initial evaluation feature data of the sample object;

[0031] The training samples are constructed based on the initial evaluation feature data.

[0032] In some embodiments, data selection is performed on the resource transfer feature data of the sample object to obtain valid feature data, including:

[0033] Performing data missing rate detection on the resource transfer feature data of the sample object to obtain the feature data missing rate;

[0034] If the missing rate of the characteristic data is less than or equal to a preset missing rate threshold, determining a data filling value of the corresponding resource transfer characteristic data;

[0035] Performing data filling on the resource transfer characteristic data based on the data filling value to obtain filled characteristic data;

[0036] Performing correlation detection on the filled feature data to obtain feature correlation;

[0037] Data selection is performed on the resource transfer feature data based on the feature correlation to obtain the valid feature data.

[0038] In some embodiments, the performing data preprocessing on the effective feature data to obtain the initial evaluation feature data of the sample object includes:

[0039] Performing chi-square binning on the valid feature data to obtain initial binning;

[0040] Perform bin data change detection based on the initial binned data to obtain bin data change distribution;

[0041] Adjusting the initial binning based on the distribution of binning data changes to obtain target binning;

[0042] The effective feature data of the target bin is subjected to variable conversion based on a preset variable conversion function to obtain the initial evaluation feature data.

[0043] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application proposes a risk assessment device based on resource transfer, the device comprising:

[0044] An acquisition module, used to acquire a training sample of a sample object, wherein the training sample includes a sample resource transfer category of the sample object, initial assessment feature data, and a sample object risk category, wherein the sample object risk category is used to indicate a risk level of resource transfer by the sample object in a scenario corresponding to the sample resource transfer category;

[0045] A data selection module, configured to perform feature selection on the initial evaluation feature data based on the sample resource transfer category to determine candidate evaluation feature data;

[0046] An input module, used for inputting the candidate assessment feature data into a preset initial risk assessment model, wherein the initial risk assessment model includes a data extraction layer, a vectorization layer, a splicing layer and a risk assessment sub-model;

[0047] A data extraction module, used for extracting the candidate evaluation feature data based on the data extraction layer to obtain numerical type feature data and label type feature data;

[0048] A vectorization module, used for performing vectorization processing on the label type feature data based on the vectorization layer to obtain a label type feature vector;

[0049] A data splicing module, used for performing data splicing on the numerical type feature data and the label type feature vector based on the splicing layer to obtain sample splicing data;

[0050] A prediction module, used to perform risk assessment on the sample splicing data based on the risk assessment sub-model to obtain a risk category of a predicted object;

[0051] A parameter adjustment module, used to determine a model loss value based on the predicted object risk category and the sample object risk category, perform parameter adjustment on the data selection model and the initial risk assessment model based on the model loss value, and construct a target risk assessment model based on the data selection model and the initial risk assessment model after parameter adjustment;

[0052] The risk assessment module is used to obtain target assessment feature data of a target object for requesting resource transfer, and perform risk assessment on the target assessment feature data based on the target risk assessment model.

[0053] To achieve the above objectives, a third aspect of an embodiment of the present application proposes an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method described in the first aspect is implemented.

[0054] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in the first aspect is implemented.

[0055] The present application proposes a risk assessment method and apparatus, device, and storage medium based on resource transfer, which obtains training samples of sample objects, wherein the training samples include sample resource transfer categories, initial assessment feature data, and sample object risk categories of the sample objects, wherein the sample object risk category is used to indicate the risk level of resource transfer of the sample object in the scenario corresponding to the sample resource transfer category; performs feature selection on the initial assessment feature data based on the sample resource transfer category to determine candidate assessment feature data; inputs the candidate assessment feature data into a preset initial risk assessment model, wherein the initial risk assessment model includes a data extraction layer, a vectorization layer, a splicing layer, and a risk assessment sub-model; performs data extraction on the candidate assessment feature data based on the data extraction layer to obtain numerical type feature data and vectorized feature data; and obtains a vectorized feature data and a vectorized feature data. label type feature data; vectorize the label type feature data based on the vectorization layer to obtain the label type feature vector; splice the numerical type feature data and the label type feature vector based on the splicing layer to obtain the sample splicing data; perform risk assessment on the sample splicing data based on the risk assessment sub-model to obtain the risk category of the predicted object; determine the model loss value based on the risk category of the predicted object and the risk category of the sample object, adjust the parameters of the data selection model and the initial risk assessment model based on the model loss value, and build a target risk assessment model based on the data selection model and the initial risk assessment model after the parameter adjustment; obtain the target assessment feature data of the target object requesting resource transfer, and perform risk assessment on the target assessment feature data based on the target risk assessment model. Compared with the related art method of performing risk assessment based on all information of the object requesting resource transfer, the embodiment of the present application can first perform feature selection on the initial assessment feature data based on the sample resource transfer category to select candidate assessment feature data that is more relevant to the risk assessment process. Furthermore, the target risk assessment model is constructed and optimized based on the parameter-adjusted data selection model and the initial risk assessment model, which can better realize the effective identification and quantitative assessment of potential risks in the resource transfer process, thereby improving the efficiency of risk assessment based on resource transfer. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 is a flow chart of a risk assessment method based on resource transfer provided in an embodiment of the present application;

[0057] Figure 2 yes Figure 1 A flowchart of step S110 in FIG.

[0058] Figure 3 yes Figure 2 A flowchart of step S230 in FIG.

[0059] Figure 4 yes Figure 2 A flowchart of step S240 in FIG.

[0060] Figure 5 yes Figure 1 A flowchart of step S120 in FIG.

[0061] Figure 6 yes Figure 5 A flowchart of step S520 in FIG.

[0062] Figure 7 yes Figure 1 A flowchart of step S170 in FIG.

[0063] Figure 8 is a structural schematic diagram of a risk assessment device based on resource transfer provided in an embodiment of the present application;

[0064] Fig. 9 It is a hardware structure diagram of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0065] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0066] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0067] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0068] First, some nouns involved in this application are analyzed:

[0069] Artificial intelligence (AI) is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. AI is a branch of computer science that attempts to understand the essence of intelligence and produce a new type of intelligent machine that can respond in a similar way to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing and expert systems. AI can simulate the information process of human consciousness and thinking. AI is also a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0070] Natural language processing (NLP): NLP uses computers to process, understand and apply human languages ​​(such as Chinese, English, etc.). NLP is a branch of artificial intelligence and an interdisciplinary subject between computer science and linguistics. It is often referred to as computational linguistics. Natural language processing includes grammatical analysis, semantic analysis, and text understanding. Natural language processing is often used in technical fields such as machine translation, handwritten and printed character recognition, speech recognition and text-to-speech conversion, information intent recognition, information extraction and filtering, text classification and clustering, public opinion analysis and opinion mining. It involves data mining related to language processing, machine learning, knowledge acquisition, knowledge engineering, artificial intelligence research, and linguistic research related to language computing.

[0071] Light Gradient Boosting Machine (LightGBM): is an efficient and scalable gradient boosting framework that uses a decision tree-based learning algorithm and optimizes it in many ways to improve training speed and model performance.

[0072] Multi-Layer Perceptron (MLP): It is a basic feedforward neural network model and also a directional structured artificial neural network.

[0073] Long Short-Term Memory (LSTM): is a special type of recurrent neural network (RNN), which is mainly used to deal with long-term dependency problems in sequence data. LSTM effectively solves the gradient vanishing and gradient exploding problems encountered by traditional recurrent neural networks (RNN) when processing long sequence data by introducing a gating mechanism.

[0074] Risk assessment of resource transfer refers to the process of analyzing and evaluating the object requesting resource transfer during the resource transfer process to determine the potential risks and likelihood of occurrence in the resource transfer process. For example, in the loan scenario of FinTech, the object requesting resource transfer can apply for a loan from a bank to ensure its current development. Before transferring resources to the object, the bank can first conduct a risk assessment on whether the object can repay the loan, thereby avoiding certain losses to the bank after the object cannot repay the loan. However, the risk assessment method based on resource transfer provided by the relevant technology usually performs risk assessment based on all information of the object requesting resource transfer, and uses LSTM to generate time series before performing risk assessment, resulting in low efficiency of risk assessment. Therefore, how to improve the efficiency of risk assessment based on resource transfer has become a technical problem that needs to be solved urgently.

[0075] Based on this, the embodiments of the present application provide a risk assessment method and apparatus, device, and storage medium based on resource transfer, which can improve the detection accuracy of database operation statements, thereby improving the security of the database.

[0076] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0077] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0078] The risk assessment method based on resource transfer provided in the embodiment of the present application relates to the field of artificial intelligence technology. The risk assessment method based on resource transfer provided in the embodiment of the present application can be applied to the terminal, can also be applied to the server side, and can also be software running in the terminal or the server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or it can be configured as a server cluster or a distributed system composed of multiple physical servers, and can also be configured to provide cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN) and big data and artificial intelligence platforms and other basic cloud computing services. Cloud servers; software can be applications that implement risk assessment methods based on resource transfer, etc., but are not limited to the above forms.

[0079] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0080] It should be noted that in each specific implementation of the present application, when it comes to the need to perform relevant processing based on data related to the identity or characteristics of the object, such as object identity information, object resource transfer information, object feature information, etc., the permission or consent of the object will be obtained first, and the collection, use and processing of these data will comply with relevant laws, regulations and standards. In addition, when the embodiment of the present application needs to obtain sensitive personal information of the object, the separate permission or separate consent of the object will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the separate permission or separate consent of the object, the necessary object-related data for the normal operation of the embodiment of the present application will be obtained.

[0081] See also Figure 1 , Figure 1is an optional flow chart of a risk assessment method based on resource transfer provided in an embodiment of the present application. In some embodiments, Figure 1 The method may include but is not limited to steps S110 to S190:

[0082] Step S110, obtaining a training sample of a sample object;

[0083] Step S120, performing feature selection on the initial evaluation feature data based on the sample resource transfer category to determine candidate evaluation feature data;

[0084] Step S130, inputting the candidate assessment feature data into a preset initial risk assessment model, the initial risk assessment model comprising a data extraction layer, a vectorization layer, a splicing layer and a risk assessment sub-model;

[0085] Step S140, extracting the candidate evaluation feature data based on the data extraction layer to obtain numerical feature data and label feature data;

[0086] Step S150, performing vectorization processing on the label type feature data based on the vectorization layer to obtain a label type feature vector;

[0087] Step S160, performing data splicing on the numerical type feature data and the label type feature vector based on the splicing layer to obtain sample splicing data;

[0088] Step S170, performing risk assessment on the sample splicing data based on the risk assessment sub-model to obtain a risk category of the predicted object;

[0089] Step S180, determining a model loss value based on the risk category of the predicted object and the risk category of the sample object, adjusting parameters of the data selection model and the initial risk assessment model based on the model loss value, and constructing a target risk assessment model based on the data selection model and the initial risk assessment model after the parameter adjustment;

[0090] Step S190: acquiring target assessment feature data of the target object for which resource transfer is requested, and performing risk assessment on the target assessment feature data based on a target risk assessment model.

[0091] In steps S110 to S190 of some embodiments, compared to the related art method of performing risk assessment based on all the information of the object requesting resource transfer, the embodiments of the present application can first perform feature selection on the initial assessment feature data based on the sample resource transfer category to select candidate assessment feature data that is more relevant to the risk assessment process. Further, the target risk assessment model is constructed and optimized based on the parameter-adjusted data selection model and the initial risk assessment model, which can better realize the effective identification and quantitative assessment of potential risks in the resource transfer process, thereby improving the efficiency of risk assessment based on resource transfer. In addition, because the present application will perform feature selection on the initial assessment feature data, it also avoids the influence of feature assessment data that is not related to the risk assessment process on the accuracy of risk assessment, thereby also improving the assessment accuracy and efficiency of risk assessment based on resource transfer. Based on this, the present application can create a strongly integrated risk control system, which can help financial institutions achieve digital transformation.

[0092] In step S110 of some embodiments, when training the model for implementing risk assessment, the present application may first obtain a training sample set, which may include multiple training samples, and each training sample may include a sample resource transfer category of the sample object, initial assessment feature data, and a sample object risk category. Among them, the sample resource transfer category may characterize the type of resource transfer specifically performed by the sample object, for example, rental loan, revolving loan, mortgage, etc. The initial assessment feature data is used to indicate feature indicator data related to the sample object, for example, the assessment feature may include asset information, liability information, key information, due diligence report information, object portrait information, repayment behavior information, second-generation credit information, etc., and each assessment feature may also include multiple indicator data, for example, asset information may include current asset information and non-current asset information; key information may include net assets, debt-to-asset ratio, current ratio, quick ratio, profitability index, etc.; due diligence report information may include company background, financial statements, management team, industry analysis, etc.

[0093] Since the training samples refer to the samples of the objects that have been risk assessed in the past, the risk category of the sample objects can be used to indicate the risk level of the sample objects in transferring resources in the scenario corresponding to the sample resource transfer category, and the object risk category can include normal risk objects (i.e., the borrower can perform the contract on time and in full), attention risk objects (i.e., the borrower currently has the ability to repay the principal and interest, but there are some adverse factors that may lead to the inability to repay the loan on time in the future, and there is a risk of overdue payment), secondary risk objects (i.e., the borrower has already defaulted and cannot guarantee timely repayment of the principal and interest in full, and may need to repay the loan through mortgage financing), suspicious risk objects (i.e., the borrower is unable to repay the principal and interest of the loan), loss risk objects (i.e., the borrower cannot repay the loan in any way), etc.; among them, normal risk objects and attention risk objects can be called general risk objects, dishonest objects, etc., while secondary risk objects, suspicious risk objects and loss risk objects can be called dishonest objects, that is, the borrower fails to perform the repayment obligations promised in the loan contract on time, resulting in overdue repayment or default. Based on this, in the actual training of this application, the sample object risk category corresponding to the selected sample object belongs to the general risk object category, that is, the impact of the dishonest object on the risk assessment is proposed.

[0094] See also Figure 2 , Figure 2 It is an optional flow chart of step S110 provided in an embodiment of the present application. In some embodiments of the present application, step S110 may specifically include steps S210 to S250:

[0095] Step S210, obtaining an initial sample of an initial object, the initial sample including the risk category, resource transfer status and resource transfer type of the initial object;

[0096] Step S220, selecting the initial object based on the preset time point, the risk category of the initial object, the resource transfer status and the resource transfer type, and determining the sample object;

[0097] Step S230, performing data selection on the resource transfer feature data of the sample object to obtain valid feature data;

[0098] Step S240, performing data preprocessing on the effective feature data to obtain initial evaluation feature data of the sample object;

[0099] Step S250: constructing training samples based on the initial evaluation feature data.

[0100] In step S210 of some embodiments, the training samples in the training sample set refer to samples selected and constructed from the initial samples. The resource transfer status may refer to the usage status of the current resource transfer account of the borrower, and may also refer to the status of the current resource transfer account of the borrower for borrowing and repaying within the set credit limit. The resource transfer type is the same as the above resource transfer type, and the initial object risk category is the same as the above sample object risk category, which will not be repeated. The initial object is all historical objects of the financial institution used for risk assessment.

[0101] In step S220 of some embodiments, the preset time point refers to a preset time node for selecting training samples. For example, if the time node is XX / XX / XXXX, then the characteristics, behaviors, transaction records and other data of customers before XX / XX / XXXX are analyzed, and the data after XX / XX / XXXX are temporarily not considered in this analysis, that is, the time node is used to evaluate or analyze a time snapshot of a certain state. Specifically, the present application can select the initial object based on the preset time point, the risk category of the initial object, the resource transfer status and the resource transfer type, and determine the sample object. For example, in the risk assessment scenario of the transfer of financial technology objects, the preset time point is XX / XX / XXXX. At this time, for the transfer of initial objects that have not been settled on XX / XX / XXXX, this application can exclude four types of initial objects from these initial objects: those with overdue days equal to 0 at the time of transfer (that is, if the overdue days of the initial object at the time of transfer are zero, indicating that the repayment can be made on time, then it will not be included in the subsequent analysis), revolving loans with normal credit limit status (which means that the corresponding initial object has not violated any terms of the loan agreement at present, and therefore will not be included in the analysis scope), mortgage (pledge) type, and defaulting object. The remaining initial objects are sample objects.

[0102] In step S230 of some embodiments, the present application may further perform data selection on the resource transfer feature data of the sample object based on feature engineering to obtain valid feature data.

[0103] It should be noted that the resource transfer characteristic data of the sample objects is the same as the description of the initial assessment characteristic data mentioned above, that is, multiple wide tables such as asset information, liability information, key information, due diligence report information, object portrait information, repayment behavior information, second-generation credit information, etc. can be extracted from the pre-set mart for storing data as the underlying data source, and the resource transfer characteristic data can be extracted from them.

[0104] It should be noted that before step S230, the present application can perform preliminary classification of the training samples corresponding to the sample objects, that is, the training samples corresponding to the sample objects can be divided into positive samples and negative samples, so as to ensure the balance of the number of positive and negative samples when forming the training sample set, and improve the accuracy of risk assessment. Among them, the present application can set the training samples that meet the following two conditions as positive samples: (1) objects that have repaid more than 5% within 5 months; (2) objects that have signed installment repayment agreements within 5 months; in addition, the present application can set the training samples that meet the following conditions as negative samples: objects that have repaid less than 5% within 5 months. The specific conditions for dividing positive and negative samples can be flexibly set according to actual needs and will not be elaborated on.

[0105] See also Figure 3 , Figure 3 It is an optional flow chart of step S230 provided in an embodiment of the present application. In some embodiments of the present application, step S230 may specifically include steps S310 to S350:

[0106] Step S310, performing data missing rate detection on the resource transfer feature data of the sample object to obtain the feature data missing rate;

[0107] Step S320, if the characteristic data missing rate is less than or equal to the preset missing rate threshold, determine the data filling value of the corresponding resource transfer characteristic data;

[0108] Step S330, performing data filling on the resource transfer characteristic data based on the data filling value to obtain filled characteristic data;

[0109] Step S340, performing correlation detection on the filled feature data to obtain feature correlation;

[0110] Step S350: selecting resource transfer feature data based on feature relevance to obtain valid feature data.

[0111] In steps S310 to S320 of some embodiments, when determining valid feature data, the present application may first clear the duplicate and redundant data in the resource transfer feature data of the sample object, and retain only the latest and unique data. The corresponding feature data missing rate is obtained by calculating the number of missing values ​​for the resource transfer feature data of each feature dimension (such as amount, time, type, etc.), and dividing it by the total data volume of the feature dimension. If the feature data missing rate is less than or equal to the preset missing rate threshold, the data filling value of the corresponding resource transfer feature data is determined. At this time, the data filling value can be determined according to the distribution characteristics of the data (such as mean, median, mode, etc.) or other statistical methods (such as interpolation, regression prediction, etc.) to determine the data filling value corresponding to each feature data. Specifically, it can be filled with the overall mean of all existing data, or it can be filled with the mode, or it can be determined according to different types of data. Corresponding different data filling values ​​to ensure the integrity and availability of the data. For example, the preset missing rate threshold is 50%. The present application can first remove feature data with a feature data missing rate greater than 50%, and perform data filling for feature data with a feature data missing rate less than or equal to 50%. Specifically, the numerical variables of the resource transfer feature data with a lower feature data missing rate can be filled with the overall mean, while the categorical variables can be filled with the mode.

[0112] In steps S330 to S350 of some embodiments, after the data filling value is determined, the missing values ​​in the original resource transfer feature data can be replaced by the determined data filling value to obtain the filled feature data. Further, the correlation between the filled feature data can be evaluated to identify redundant or irrelevant features. Specifically, the feature correlation between each two filled feature data can be used, such as using the Pearson correlation coefficient, the Spearman rank correlation coefficient, the information value (IV), etc., to determine the correlation between the features. Further, data selection is performed on the resource transfer feature data based on the feature correlation, that is, the feature variables in the resource transfer feature data whose IV values ​​corresponding to the feature correlation are greater than the preset correlation (such as 0.6, 0.7, etc.) can only retain one data with a larger feature correlation, and eliminate other correlation variables to prevent the occurrence of multicollinearity problems. In addition, variables with IV values ​​>0.1 or above can be retained (that is, research and analysis have found that variables with IV values ​​>0.1 or above are beneficial to classification) to determine the valid feature data corresponding to the training sample.

[0113] In step S240 of some embodiments, data preprocessing is performed on the effective feature data to obtain initial evaluation feature data of the sample object. Specifically, the present application can use chi-square binning to perform data conversion on the effective feature data.

[0114] See also Figure 4, Figure 4 It is an optional flow chart of step S240 provided in an embodiment of the present application. In some embodiments of the present application, step S240 may specifically include steps S410 to S440:

[0115] Step S410, performing chi-square binning on the valid feature data to obtain initial binning;

[0116] Step S420, performing bin data change detection based on the initial binned data to obtain bin data change distribution;

[0117] Step S430, adjusting the initial bins based on the distribution of bin data changes to obtain target bins;

[0118] Step S440: Perform variable conversion on the effective feature data of the target bin based on a preset variable conversion function to obtain initial evaluation feature data.

[0119] In step S410 of some embodiments, the present application can discretize the effective feature data through the chi-square binning algorithm for subsequent analysis and modeling. Specifically, you can first select a valid feature data (such as a numerical variable), and then calculate the correlation between the effective feature data and the target variable (usually a label), and use the chi-square test to evaluate the correlation between the variable and the target. Furthermore, according to the chi-square statistic, select an appropriate number of bins to ensure that the sample size in each box is large enough to obtain the initial binning result, such as dividing the numerical feature into several categories (boxes).

[0120] In step S420 of some embodiments, in order to detect whether there are significant changes in the data after the initial binning and to evaluate the effectiveness of the binning, the present application can calculate the distribution of the data after the initial binning and compare it with the target variable, that is, count the number of samples in different bins and the performance of the target variable (such as the ratio of good to bad customers), identify whether there are significant changes in the binned data, for example, the target variable of a certain bin presents a different distribution, analyze the distribution corresponding to these changes, and obtain the distribution of changes in the binned data.

[0121] In step S430 of some embodiments, further, the interval trend of the variable after the initial binning is observed. If the distribution of binning data changes indicates that the binning result does not show a monotonic trend, the binning result can be adjusted to see whether it is necessary to merge the bins. The adjustment standard is that the distribution of binning data changes should show a monotonic increasing or decreasing trend. Therefore, the present application can merge or split bins with significant changes according to correlation and performance to achieve a more optimized binning goal.

[0122] In step S440 of some embodiments, further, after the variables are binned, the present application may perform variable conversion on the effective feature data of the target bin based on a preset variable conversion function (such as a WOETransformer function) to obtain initial evaluation feature data. For example, the WOETransformer function may be used to calculate the weight of evidence (WOE) for the binned variables, and the WOE is a conversion of the original variables in a coded form, and the converted data may be used to form a training sample set for the model.

[0123] In the above embodiments, the present application combines feature missing rate detection and correlation detection of feature engineering, and further combines chi-square binning to perform data preprocessing on effective feature data to obtain initial evaluation feature data of sample objects, which can better optimize the input feature data of the model and improve the evaluation accuracy and stability of the model.

[0124] In step S250 of some embodiments, further, a training sample for training a risk assessment model may be constructed based on the obtained initial assessment feature data of the sample object.

[0125] In step S120 of some embodiments, since each initial evaluation feature data also includes multiple indicator data, that is, the initial evaluation feature data includes initial evaluation feature sub-data corresponding to multiple object evaluation features. For the initial evaluation feature data of asset information, it can include two object evaluation features of current asset information and non-current asset information; for the initial evaluation feature data of key information, it can include object evaluation features such as net assets, debt-to-asset ratio, current ratio, quick ratio, profitability index, etc. Since the evaluation feature data required for risk assessment of different sample resource transfer categories are not exactly the same, there may be certain differences. Therefore, before inputting the evaluation feature data into the model, the present application can first perform feature selection on the initial evaluation feature data based on the sample resource transfer category to determine the candidate evaluation feature data corresponding to the sample resource transfer category. In this way, the amount of data input into the model can be reduced, the calculation speed can be improved, and the influence of invalid evaluation feature data on the evaluation accuracy can be avoided, thereby improving the efficiency of risk assessment based on resource transfer.

[0126] See also Figure 5 , Figure 5 It is an optional flow chart of step S120 provided in an embodiment of the present application. In some embodiments of the present application, step S120 may specifically include steps S510 to S520:

[0127] Step S510, performing feature evaluation on the initial evaluation feature sub-data and the sample resource transfer category based on a preset feature evaluation model to obtain a feature importance evaluation value of the object evaluation feature;

[0128] Step S520 , selecting the initial evaluation feature sub-data based on the feature importance evaluation value and the preset feature quantity, and determining candidate evaluation feature data.

[0129] In step S510 and step S520 of some embodiments, the feature importance evaluation value is used to characterize the importance of each object evaluation feature in the risk assessment scenario corresponding to the sample resource transfer category. The object evaluation feature here refers to the specific feature index corresponding to the initial evaluation feature sub-data. The present application can first input the initial evaluation feature sub-data and the sample resource transfer category into a preset feature evaluation model for feature evaluation, and obtain the feature importance of the object evaluation feature corresponding to each initial evaluation feature sub-data by learning the importance of the initial evaluation feature sub-data in the risk assessment scenario corresponding to the sample resource transfer category, that is, the feature importance evaluation value. Among them, the larger the feature importance evaluation value, the higher the importance of the corresponding object evaluation feature in the corresponding risk assessment scenario. Further, the present application can arrange multiple initial evaluation feature sub-data in descending order based on the feature importance evaluation value to obtain a feature sub-data sequence, and select a preset number of initial evaluation feature sub-data from high to low according to the feature importance evaluation value, and use the selected initial evaluation feature sub-data as candidate evaluation feature data. The preset feature quantity may be a specific quantity or a percentage of the quantity contained in the feature sub-data sequence, such as selecting the first 30% of the initial evaluation feature sub-data in the feature sub-data sequence as candidate evaluation feature data.

[0130] It should be noted that the feature evaluation model constructed in this application can be constructed based on the LightGBM structure. The feature evaluation model constructed based on LightGBM in this application combined with other technical features has the following advantages over the traditional classification model: (1) The feature evaluation model uses a histogram-based algorithm to make the model training faster and more efficient; (2) Lower memory usage: Using discrete bins to save and replace continuous values ​​results in less memory usage; (3) Higher accuracy (compared to any other boosting algorithm), the leave-wise splitting method produces a more complex tree than the leave-wise splitting method. In addition, overfitting can be prevented by setting the |max-depth| parameter; (4) Better big data processing capabilities. Compared with XGBoost, this application can reduce training time and has the ability to process big data.

[0131] See also Figure 6 , Figure 6It is an optional flow chart of step S520 provided in an embodiment of the present application. In some embodiments of the present application, step S520 may specifically include steps S610 to S640:

[0132] Step S610, obtaining feature preference weights of object evaluation features and historical resource transfer scores of sample objects;

[0133] Step S620, determining a target feature weight of the object evaluation feature based on the feature preference weight and the historical resource transfer score;

[0134] Step S630, performing weighted calculation on the target feature weight and the feature importance evaluation value to obtain a target evaluation value;

[0135] Step S640 , selecting the initial evaluation feature sub-data based on the target evaluation value and the preset feature quantity to determine candidate evaluation feature data.

[0136] In step S610 of some embodiments, the feature preference weight refers to the importance of the corresponding object evaluation feature set in advance to the sample object evaluation. For example, the preference weight of each object evaluation feature can be calculated using statistical analysis or machine learning methods (such as feature selection algorithms). The historical resource transfer score refers to a historical score calculated based on the historical resource transfer data of the sample object, reflecting the performance of the corresponding sample object in historical resource transfer or resource utilization. The higher the feature preference weight, the more important the object evaluation feature is to the risk assessment of resource transfer for the sample object. The higher the historical resource transfer score, the lower the risk of the sample object in the resource transfer scenario and the better the performance.

[0137] In step S620 of some embodiments, in order to determine the comprehensive impact of each object assessment feature on the risk assessment of the sample object, the target feature weight of the object assessment feature is determined based on the feature preference weight and the historical resource transfer score, which can be achieved by a simple weighted average, linear combination, or a more complex model (such as a regression model), which will not be described in detail. The target feature weight is the target weight of each object assessment feature in the risk assessment, reflecting its relative importance in the assessment process.

[0138] In step S630 of some embodiments, a weighted calculation is performed on the target feature weight and the feature importance evaluation value, that is, a final evaluation value can be generated by combining the target feature weight through a certain scoring mechanism or model, where the target evaluation value = ∑ (target feature weight × feature importance evaluation value) to quantify the risk assessment result of the sample object.

[0139] In step S640 of some embodiments, further, the most representative and valuable features are selected from all object evaluation features according to the obtained target evaluation value combined with the preset number of features (i.e., the number of features desired to be selected). In this way, the feature dimension can be reduced while retaining the important features that best reflect the object characteristics, thereby improving the effectiveness of subsequent analysis or models.

[0140] In the above embodiments, the present application can comprehensively analyze the feature importance and the historical performance of the object to screen out the most valuable features for evaluation, so as to improve the decision accuracy and model effect.

[0141] In step S130 of some embodiments, further, the present application may input the candidate assessment feature data into a preset initial risk assessment model, which includes a data extraction layer, a vectorization layer, a splicing layer and a risk assessment sub-model, that is, the present application may use the constructed initial risk assessment model to perform risk assessment training on the candidate assessment features.

[0142] In step S140 of some embodiments, data extraction is performed on the candidate evaluation feature data based on the data extraction layer, and valid information in the candidate evaluation feature data can be extracted. The data extraction layer can divide the candidate features into numerical type features (i.e., numerical variables) and label type features (i.e., categorical variables), and output the corresponding feature data.

[0143] In step S150 of some embodiments, the extracted tag type feature data may be vectorized in the vectorization layer, and specifically, a word vectorization method such as One-Hot Encoding and Word Embedding may be used to generate a vector representation of the tag type feature so that the model can understand and calculate it. In this way, the tag type feature data may be converted into a format that can be used for model calculation.

[0144] In step S160 of some embodiments, the vectorized label type feature vector and the numerical type feature data can be further spliced ​​to merge different types of feature data into a unified data format for subsequent processing. Therefore, the spliced ​​data can be used as the input of the risk assessment sub-model.

[0145] In step S170 of some embodiments, the sample splicing data is passed to the risk assessment sub-model, and after model reasoning and calculation, the predicted object risk category of the sample object is output. This category reflects the risk level of the object and provides a basis for subsequent steps.

[0146] It should be noted that the risk assessment submodel of the present application can be constructed based on the MLP network structure, and the risk assessment submodel includes a preset number of hidden layers and normalization layers. The preset number can be 3 layers, 4 layers, etc., and can be flexibly set according to the required evaluation efficiency and accuracy.

[0147] It should be noted that the application can construct a normalization layer based on different activation functions. The activation function plays a very important role in the neural network and can be used to map the output of the neuron to a specific range, such as Sigmoid function, Rectified Linear Unit (ReLU), Tanh function, etc. For example, the risk assessment submodel of the application can set the hidden layer to 3 layers, and the activation function selects the ReLU function to prevent the data from having gradient vanishing or gradient exploding.

[0148] See also Figure 7 , Figure 7 It is an optional flow chart of step S170 provided in an embodiment of the present application. In some embodiments of the present application, step S170 may specifically include steps S710 to S730:

[0149] Step S710, extracting data from the sample spliced ​​data based on a preset number of hidden layers to obtain sample hidden data;

[0150] Step S720, normalizing the sample hidden data based on the normalization layer to obtain the sample risk assessment probability;

[0151] Step S730, classify the object into categories based on the preset category threshold and the sample risk assessment probability to obtain the predicted object risk category.

[0152] In steps S710 to S730 of some embodiments, after the sample splicing data is input into the risk assessment sub-model, data extraction can be performed on the sample splicing data based on a preset number of hidden layers (these layers are layers between the input layer and the output layer in the neural network), and forward propagation is performed through these hidden layers. Each layer performs a nonlinear transformation on the input data through a certain activation function, thereby extracting higher-level features. In this way, more effective feature information can be extracted from the sample splicing data through a preset number of hidden layers. Further, the sample hidden data is normalized based on the normalization layer to convert the sample hidden data to a unified scale for subsequent processing and analysis. Among them, normalization is a data preprocessing technology used to scale data to a specific range or distribution. After the sample hidden data passes through the normalization layer, its value will be adjusted to a specific range, such as [0,1] or [-1,1], or it will conform to a certain standard distribution (such as a normal distribution). Finally, the normalized data is called the sample risk assessment probability, which is used to represent the sample risk assessment probability corresponding to each training sample, that is, the degree of risk. Finally, the final category of each training sample can be determined based on the sample risk assessment probability and the preset category threshold. The preset category threshold is set according to business needs or prior knowledge, and is used to distinguish the boundaries of different categories. It can also be adjusted according to the model accuracy. For example, the sample risk assessment probability represents the probability of each training sample defaulting. The preset category threshold is 0.5. At this time, for training samples with a sample risk assessment probability greater than 0.5, they can be judged as suspected default samples, indicating that the risk category of the predicted object of the training sample is a risk object.

[0153] It should be noted that this application can first use the feature evaluation model based on LightGBM for feature selection, and then use Convolutional Neural Networks (CNN) to generate feature vectors (i.e., vectorization layer), and combine with the risk assessment sub-model based on MLP to predict risk categories. Compared with the model structure and processing methods used in related technologies, it can better improve the assessment accuracy and efficiency of risk assessment based on resource transfer. In addition, this application is a classification prediction. Compared with the regression prediction used in related technologies, the model objectives and parameters are completely different. Both in data processing and classification prediction capabilities, they are much stronger than linear models, which can better improve the accuracy of assessment.

[0154] In step S180 of some embodiments, further, a model loss value can be determined based on the predicted object risk category and the sample object risk category using calculation methods such as cross entropy loss and mean square error to compare the difference between the risk category predicted by the model and the true label category. According to the calculated model loss value, the parameters of the data selection model and the initial risk assessment model are adjusted. This process may involve back propagation and optimization algorithms (such as gradient descent) to improve the performance of the model.

[0155] In step S190 of some embodiments, further, the constructed target risk assessment model can be used to perform risk assessment on the new object. Specifically, the relevant feature data of the target object to be transferred can be obtained (through data extraction and preprocessing), and then input into the constructed target risk assessment model for risk assessment. At this time, the target risk assessment model can process the target assessment feature data and output the risk assessment result of the object to help decide whether the resource transfer is safe or not.

[0156] It should be noted that this application supports parallel learning, that is, the features with high importance after screening can be simultaneously input into the initial risk assessment model for training, so as to obtain the probability value of default for each sample.

[0157] The embodiment of the present application provides a risk assessment method based on resource transfer, which can continuously optimize the risk prediction ability for specific objects through feature engineering, model training and evaluation. The advantages of deep learning model feature generation and integrated model classification prediction are utilized to combine the advantages of the two models; in the first stage, discrete data is regarded as text, and embedding processing is performed on it to convert it into word vectors, and then it is spliced ​​with numerical data, and it is input into the initial risk assessment model built based on CNN. By superimposing hidden layers, many features that cannot be mined by the human eye are generated, and then classification prediction is performed. Compared with the related technology of risk assessment based on all information of the object requesting resource transfer, the embodiment of the present application can better realize the effective identification and quantitative evaluation of potential risks in the resource transfer process, thereby improving the efficiency of risk assessment based on resource transfer.

[0158] It should be noted that the non-Company software tools or components appearing in the embodiments of the present application are merely examples and do not represent actual use.

[0159] See also Figure 8 The embodiment of the present application further provides a risk assessment device based on resource transfer, which can implement the above-mentioned risk assessment method based on resource transfer, and the device includes:

[0160] An acquisition module 810 is used to acquire a training sample of a sample object, wherein the training sample includes a sample resource transfer category of the sample object, initial assessment feature data, and a risk category of the sample object, wherein the risk category of the sample object is used to indicate the risk level of resource transfer of the sample object in a scenario corresponding to the sample resource transfer category;

[0161] A data selection module 820 is used to perform feature selection on the initial evaluation feature data based on the sample resource transfer category to determine candidate evaluation feature data;

[0162] An input module 830, used to input candidate assessment feature data into a preset initial risk assessment model, the initial risk assessment model including a data extraction layer, a vectorization layer, a splicing layer and a risk assessment sub-model;

[0163] A data extraction module 840 is used to extract the candidate evaluation feature data based on the data extraction layer to obtain numerical type feature data and label type feature data;

[0164] A vectorization module 850 is used to perform vectorization processing on the label type feature data based on the vectorization layer to obtain a label type feature vector;

[0165] A data splicing module 860 is used to perform data splicing on the numerical type feature data and the label type feature vector based on the splicing layer to obtain sample splicing data;

[0166] Prediction module 870, used to perform risk assessment on sample splicing data based on the risk assessment sub-model to obtain a risk category of the predicted object;

[0167] A parameter adjustment module 880 is used to determine a model loss value based on the risk category of the predicted object and the risk category of the sample object, adjust the parameters of the data selection model and the initial risk assessment model based on the model loss value, and construct a target risk assessment model based on the data selection model and the initial risk assessment model after the parameter adjustment;

[0168] The risk assessment module 890 is used to obtain target assessment feature data of the target object for which resource transfer is requested, and perform risk assessment on the target assessment feature data based on a target risk assessment model.

[0169] The specific implementation of the risk assessment device based on resource transfer is basically the same as the specific implementation of the risk assessment method based on resource transfer described above, and will not be described in detail here.

[0170] The embodiment of the present application also provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the above-mentioned risk assessment method based on resource transfer when executing the computer program. The electronic device can be any smart terminal including a tablet computer, a car computer, etc.

[0171] See also Fig. 9 , Fig. 9 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0172] The processor 910 may be implemented by a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0173] The memory 920 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 920 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 920, and the processor 910 calls and executes the risk assessment method based on resource transfer in the embodiment of this application;

[0174] Input / output interface 930, used to implement information input and output;

[0175] Communication interface 940, used to realize communication interaction between the device and other devices, which can be realized through wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);

[0176] bus 950 , which transmits information between the various components of the device (e.g., processor 910 , memory 920 , input / output interface 930 , and communication interface 940 );

[0177] The processor 910 , the memory 920 , the input / output interface 930 , and the communication interface 940 are connected to each other in communication within the device via a bus 950 .

[0178] An embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned risk assessment method based on resource transfer is implemented.

[0179] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0180] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0181] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0182] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0183] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0184] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0185] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0186] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0187] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0188] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0189] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0190] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A risk assessment method based on resource transfer, characterized in that: The method comprises: Acquire a training sample of a sample object, wherein the training sample includes a sample resource transfer category of the sample object, initial assessment feature data, and a sample object risk category, wherein the sample object risk category is used to indicate a risk level of the sample object in performing resource transfer in a scenario corresponding to the sample resource transfer category; Performing feature selection on the initial evaluation feature data based on the sample resource transfer category to determine candidate evaluation feature data; Inputting the candidate assessment feature data into a preset initial risk assessment model, wherein the initial risk assessment model includes a data extraction layer, a vectorization layer, a splicing layer and a risk assessment sub-model; Extracting the candidate evaluation feature data based on the data extraction layer to obtain numerical feature data and label feature data; Performing vectorization processing on the label type feature data based on the vectorization layer to obtain a label type feature vector; Based on the splicing layer, the numerical type feature data and the label type feature vector are spliced ​​to obtain sample splicing data; Performing risk assessment on the sample splicing data based on the risk assessment sub-model to obtain a risk category of the predicted object; Determine a model loss value based on the predicted object risk category and the sample object risk category, adjust parameters of the data selection model and the initial risk assessment model based on the model loss value, and construct a target risk assessment model based on the data selection model and the initial risk assessment model after parameter adjustment; Target assessment feature data of a target object for which resource transfer is requested is acquired, and risk assessment is performed on the target assessment feature data based on the target risk assessment model.

2. The method according to claim 1, characterized in that: The initial evaluation feature data includes initial evaluation feature sub-data corresponding to the object evaluation feature, and the feature selection of the initial evaluation feature data based on the sample resource transfer category to determine candidate evaluation feature data includes: Performing feature evaluation on the initial evaluation feature sub-data and the sample resource transfer category based on a preset feature evaluation model to obtain a feature importance evaluation value of the object evaluation feature, wherein the feature importance evaluation value is used to characterize the importance of the object evaluation feature in the risk assessment scenario corresponding to the sample resource transfer category; The initial evaluation feature sub-data is selected based on the feature importance evaluation value and the preset feature quantity to determine the candidate evaluation feature data.

3. The method according to claim 2, characterized in that The step of selecting the initial evaluation feature sub-data based on the feature importance evaluation value and the preset feature quantity to determine the candidate evaluation feature data includes: Obtaining feature preference weights of the object evaluation features and historical resource transfer scores of the sample object; determining a target feature weight for the object assessment feature based on the feature preference weight and the historical resource transfer score; Performing weighted calculation on the target feature weight and the feature importance evaluation value to obtain a target evaluation value; The initial evaluation feature sub-data is selected based on the target evaluation value and the preset feature quantity to determine the candidate evaluation feature data.

4. The method according to claim 1, characterized in that: The risk assessment sub-model includes a preset number of hidden layers and a normalization layer. The risk assessment of the sample splicing data is performed based on the risk assessment sub-model to obtain the risk category of the predicted object, including: Extracting data from the sample spliced ​​data based on the preset number of hidden layers to obtain sample hidden data; Normalizing the sample hidden data based on the normalization layer to obtain a sample risk assessment probability; The object category is divided based on the preset category threshold and the sample risk assessment probability to obtain the predicted object risk category.

5. The method according to any one of claims 1 to 4, characterized in that: The step of obtaining a training sample of a sample object includes: Acquire an initial sample of initial objects, wherein the initial sample includes a risk category, a resource transfer status, and a resource transfer type of the initial object; Selecting the initial object based on a preset time point, the risk category of the initial object, the resource transfer status, and the resource transfer type to determine the sample object; Performing data selection on the resource transfer feature data of the sample object to obtain valid feature data; Performing data preprocessing on the effective feature data to obtain the initial evaluation feature data of the sample object; The training samples are constructed based on the initial evaluation feature data.

6. The method according to claim 5, characterized in that Data selection is performed on the resource transfer feature data of the sample object to obtain valid feature data, including: Performing data missing rate detection on the resource transfer feature data of the sample object to obtain the feature data missing rate; If the missing rate of the characteristic data is less than or equal to a preset missing rate threshold, determining a data filling value of the corresponding resource transfer characteristic data; Performing data filling on the resource transfer characteristic data based on the data filling value to obtain filled characteristic data; Performing correlation detection on the filled feature data to obtain feature correlation; Data selection is performed on the resource transfer feature data based on the feature correlation to obtain the valid feature data.

7. The method according to claim 5, characterized in that The performing data preprocessing on the effective feature data to obtain the initial evaluation feature data of the sample object includes: Performing chi-square binning on the valid feature data to obtain initial binning; Perform bin data change detection based on the initial binned data to obtain bin data change distribution; Adjusting the initial binning based on the distribution of binning data changes to obtain target binning; The effective feature data of the target bin is subjected to variable conversion based on a preset variable conversion function to obtain the initial evaluation feature data.

8. A risk assessment device based on resource transfer, characterized in that: The device comprises: An acquisition module, used to acquire a training sample of a sample object, wherein the training sample includes a sample resource transfer category of the sample object, initial assessment feature data, and a sample object risk category, wherein the sample object risk category is used to indicate a risk level of resource transfer by the sample object in a scenario corresponding to the sample resource transfer category; A data selection module, configured to perform feature selection on the initial evaluation feature data based on the sample resource transfer category to determine candidate evaluation feature data; An input module, used for inputting the candidate assessment feature data into a preset initial risk assessment model, wherein the initial risk assessment model includes a data extraction layer, a vectorization layer, a splicing layer and a risk assessment sub-model; A data extraction module, used for extracting the candidate evaluation feature data based on the data extraction layer to obtain numerical type feature data and label type feature data; A vectorization module, used for performing vectorization processing on the label type feature data based on the vectorization layer to obtain a label type feature vector; A data splicing module, used for performing data splicing on the numerical type feature data and the label type feature vector based on the splicing layer to obtain sample splicing data; A prediction module, used to perform risk assessment on the sample splicing data based on the risk assessment sub-model to obtain a risk category of a predicted object; A parameter adjustment module, used to determine a model loss value based on the predicted object risk category and the sample object risk category, perform parameter adjustment on the data selection model and the initial risk assessment model based on the model loss value, and construct a target risk assessment model based on the data selection model and the initial risk assessment model after parameter adjustment; The risk assessment module is used to obtain target assessment feature data of a target object for requesting resource transfer, and perform risk assessment on the target assessment feature data based on the target risk assessment model.

9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images