Risk assessment method, device, equipment, storage medium and program product

Abstract

This application provides a risk assessment method, apparatus, device, storage medium, and program product, relating to the field of financial technology. The method includes: acquiring multi-dimensional user data associated with a target user; performing feature extraction processing on the multi-dimensional user data to obtain event features; performing statistical calculations on the multi-dimensional user data based on preset calculation rules to obtain global statistical features corresponding to the preset calculation rules; obtaining a comprehensive prediction result corresponding to the multi-dimensional user data based on the event features and the global statistical features; and determining the risk assessment result for the target user based on the comprehensive prediction result. The method of this application achieves the technical effect of improving the reliability of risk assessment.

Description

Technical Field

[0001] This application relates to the field of financial technology, and more particularly to a risk assessment method, apparatus, device, storage medium, and program product. Background Technology

[0002] With the development of the financial sector, the types of products used for financial asset management are constantly increasing. In order to ensure the stability of the financial asset management market and the sustainable development of the industry, it is necessary to establish a sound risk control system to achieve risk assessment and early warning for financial asset management and provide reliable asset management support.

[0003] In existing technologies, when assessing the risk of financial product asset management for users, information is mainly collected based on standardized questionnaires filled out by users themselves. Then, asset management risk assessment is conducted based on a pre-set evaluation system, and corresponding asset management recommendations are generated based on the user's risk assessment results.

[0004] Because existing risk assessment methods rely on periodic questionnaires and lack consideration for real-time objective factors, they suffer from low reliability in risk assessment. Summary of the Invention

[0005] This application provides a risk assessment method, apparatus, equipment, storage medium, and program product to solve the technical problem of low reliability in risk assessment.

[0006] Firstly, this application provides a risk assessment method, including:

[0007] Obtain multi-dimensional user data associated with the target user;

[0008] Feature extraction processing is performed on multi-dimensional user data to obtain event features;

[0009] Statistical calculations are performed on multi-dimensional user data based on preset calculation rules to obtain global statistical features corresponding to the preset calculation rules.

[0010] Based on event characteristics and global statistical characteristics, a comprehensive prediction result corresponding to multi-dimensional user data is obtained;

[0011] The risk assessment results for the target users are determined based on the comprehensive forecast results.

[0012] In one possible implementation, feature extraction processing is performed on multi-dimensional user data to obtain event features, including:

[0013] The event identification results are obtained by processing multi-dimensional user data.

[0014] The event identification results are processed to obtain event features.

[0015] In one possible implementation, multi-dimensional user data is processed to obtain event recognition results, including:

[0016] Based on event recognition rules, multi-dimensional user data is filtered to obtain specific events and the first event data corresponding to the specific events;

[0017] Based on the analysis model, multi-dimensional user data is analyzed to obtain fuzzy events and corresponding second event data.

[0018] The definite events, the first event data, the fuzzy events, and the second event data are identified as the event recognition results.

[0019] In one possible implementation, a comprehensive prediction result corresponding to multi-dimensional user data is obtained based on event characteristics and global statistical characteristics, including:

[0020] Encode the global statistical features to obtain sub-score features;

[0021] By concatenating event features and sub-score features, a feature vector of multi-dimensional user data is obtained.

[0022] The feature vector is input into the multi-objective prediction model to obtain the comprehensive prediction result. The prediction target in the multi-objective prediction model corresponds one-to-one with the calculation rule type in the preset calculation rule. The comprehensive prediction result includes at least the prediction probability corresponding to each prediction target.

[0023] The risk assessment results for the target users are determined based on the comprehensive forecast results, including:

[0024] The risk assessment results are determined based on the predicted probability corresponding to each prediction target; the risk assessment results include at least the risk level of each prediction target and the overall risk level of the target user.

[0025] In one possible implementation, after determining the risk assessment result for the target user based on the comprehensive prediction results, the method further includes:

[0026] A set of constraints is generated based on preset constraint rules, the risk level of each prediction target, and the comprehensive risk level of the target user.

[0027] Obtain user-side characteristics of the target users and the target product library;

[0028] Extract the core features of the products that correspond to the product information in the target product library;

[0029] Based on the set of constraints, user-side features, and core product features, output a product recommendation scheme for the target user.

[0030] In one possible implementation, obtaining the user-side characteristics of the target user includes:

[0031] Based on multi-dimensional user data, acquire the target user's asset data;

[0032] Asset data, risk assessment results, and global statistical characteristics are identified as user-side features.

[0033] Secondly, this application provides a risk assessment device, comprising:

[0034] The acquisition module is used to acquire multi-dimensional user data associated with the target user;

[0035] The first processing module is used to perform feature extraction processing on multi-dimensional user data to obtain event features;

[0036] The second processing module is used to perform statistical calculations on multi-dimensional user data based on preset calculation rules to obtain global statistical features corresponding to the preset calculation rules.

[0037] The third processing module is used to obtain a comprehensive prediction result corresponding to the user's multi-dimensional data based on event characteristics and global statistical characteristics.

[0038] The fourth processing module is used to determine the risk assessment results for the target user based on the comprehensive prediction results.

[0039] In one possible implementation, the first processing module is further configured to:

[0040] The event identification results are obtained by processing multi-dimensional user data.

[0041] The event identification results are processed to obtain event features.

[0042] In one possible implementation, the first processing module is further configured to:

[0043] Based on event recognition rules, multi-dimensional user data is filtered to obtain specific events and the first event data corresponding to the specific events;

[0044] Based on the analysis model, multi-dimensional user data is analyzed to obtain fuzzy events and corresponding second event data.

[0045] The definite events, the first event data, the fuzzy events, and the second event data are identified as the event recognition results.

[0046] In one possible implementation, the third processing module is further configured to:

[0047] Encode the global statistical features to obtain sub-score features;

[0048] By concatenating event features and sub-score features, a feature vector of multi-dimensional user data is obtained.

[0049] The feature vector is input into the multi-objective prediction model to obtain the comprehensive prediction result. The prediction target in the multi-objective prediction model corresponds one-to-one with the calculation rule type in the preset calculation rule. The comprehensive prediction result includes at least the prediction probability corresponding to each prediction target.

[0050] The corresponding fourth processing module is also used for:

[0051] The risk assessment results are determined based on the predicted probability corresponding to each prediction target; the risk assessment results include at least the risk level of each prediction target and the overall risk level of the target user.

[0052] In one possible implementation, the fourth processing module is further configured to:

[0053] A set of constraints is generated based on preset constraint rules, the risk level of each prediction target, and the comprehensive risk level of the target user.

[0054] Obtain user-side characteristics of the target users and the target product library;

[0055] Extract the core features of the products that correspond to the product information in the target product library;

[0056] Based on the set of constraints, user-side features, and core product features, output a product recommendation scheme for the target user.

[0057] In one possible implementation, the fourth processing module is further configured to:

[0058] Based on multi-dimensional user data, acquire the target user's asset data;

[0059] Asset data, risk assessment results, and global statistical characteristics are identified as user-side features.

[0060] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor;

[0061] The memory stores the instructions that the computer executes;

[0062] The processor executes computer execution instructions stored in memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0063] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0064] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements various possible implementations of the first or second aspect described above.

[0065] This application provides a risk assessment method, apparatus, equipment, storage medium, and program product. Based on this method, multi-dimensional user data associated with a target user is acquired. Feature extraction processing is performed on the multi-dimensional user data to obtain event features. Statistical calculations are then performed on the multi-dimensional user data based on preset calculation rules to obtain global statistical features. A comprehensive prediction result corresponding to the multi-dimensional user data is obtained based on the event features and global statistical features, and the risk assessment result for the target user is determined based on the comprehensive prediction result. This application achieves comprehensive and accurate data collection through the acquisition of multi-dimensional data; it mines event features related to user behavior or assets from the multi-dimensional user data using feature extraction; it statistically obtains the global statistical features of the multi-dimensional user data using rule calculation; and it combines the two types of features to obtain a corresponding comprehensive prediction result, achieving the goal of accurate prediction through feature fusion. The obtained comprehensive prediction result is used to determine the risk assessment result for the target user, further realizing the generation of accurate risk assessment conclusions. Compared with existing technologies, this method avoids the information delays associated with relying on questionnaires for information collection. At the same time, it achieves comprehensive information collection through multi-dimensional user data, and improves the accuracy of risk assessment by combining the comprehensive prediction results generated from the two features. By using the accurate comprehensive prediction results to generate assessment results, the technical effect of improving the credibility of risk assessment is achieved. Attached Figure Description

[0066] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0067] Figure 1 A system architecture diagram of a risk assessment system provided in this application;

[0068] Figure 2 A flowchart illustrating an embodiment of the risk assessment method provided in this application;

[0069] Figure 3 A flowchart illustrating Embodiment 2 of the risk assessment method provided in this application;

[0070] Figure 4 A schematic diagram of the risk assessment device provided in this application;

[0071] Figure 5 A schematic diagram of the structure of the electronic device provided in this application.

[0072] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0073] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0074] 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, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of the relevant data all comply with relevant laws, regulations, and standards, necessary confidentiality measures have been taken, they do not violate public order and good morals, and corresponding operation portals are provided for users to choose to authorize or refuse.

[0075] It should be noted that the risk assessment methods, devices, equipment, storage media, and program products provided in this application can be used in the fintech field, or in any field other than fintech. The application fields of the risk assessment methods, devices, equipment, storage media, and program products in this application are not limited.

[0076] In existing technologies, when conducting risk assessments for users, it is necessary to use an asset management platform to distribute questionnaires to users, establish basic user profiles based on the user information collected from the questionnaires, and then conduct asset management risk assessments on the user profiles based on a pre-set evaluation system to obtain the corresponding risk assessment results.

[0077] However, when conducting risk assessments on users using existing technologies, the questionnaire survey method has a certain update cycle. When user behavior or asset information changes, if a questionnaire survey and information collection are not conducted at this time, the user's detailed information cannot be synchronized in a timely manner. As a result, the final assessment results do not conform to the current user's actual situation, leading to a technical problem of low reliability in risk assessments in existing technologies.

[0078] To address the aforementioned technical issues, this application proposes the following technical concept: utilizing multi-dimensional user-related data for risk assessment. Specifically: acquiring multi-dimensional user data associated with the target user to achieve multi-dimensional information collection; extracting features from the multi-dimensional user data to obtain event features, thereby mining user behavior and specific event attributes related to assets; performing statistical calculations on the multi-dimensional user data based on preset calculation rules to obtain global statistical features, thereby constructing macro-level characteristics of user behavior and assets; using event features and global statistical features to determine the comprehensive prediction results corresponding to the multi-dimensional user data, thereby achieving the goal of integrating macro-level features and micro-level event features for accurate prediction; and determining the risk assessment results for the target user based on the comprehensive prediction results, thereby generating accurate risk assessment conclusions and improving the credibility of risk assessment.

[0079] Figure 1 This application provides a system architecture diagram for a risk assessment system, which is a computer device. Figure 1 As shown, the above framework includes at least one of a data acquisition device 101, a processing device 102, and a display device 103.

[0080] It is understood that the structures illustrated in the embodiments of this application do not constitute a specific limitation on the system architecture described above. In other feasible embodiments of this application, the architecture may include more or fewer components than illustrated, or combine some components, or split some components, or arrange different components, which can be determined according to the actual application scenario and is not limited here. Figure 1 The components shown can be implemented in hardware, software, or a combination of both.

[0081] In the specific implementation process, the data acquisition device 101 may include an input / output interface or a communication interface, and the data acquisition device 101 can be connected to the processing device through the input / output interface or the communication interface.

[0082] The processing device 102 can acquire multi-dimensional user data associated with the target user; perform feature extraction processing on the multi-dimensional user data to obtain event features; perform statistical calculations on the multi-dimensional user data based on preset calculation rules to obtain global statistical features corresponding to the preset calculation rules; obtain a comprehensive prediction result corresponding to the multi-dimensional user data based on the event features and global statistical features; and determine the risk assessment result of the target user based on the comprehensive prediction result.

[0083] The display device 103 can also be a touch screen or the screen of a terminal device, used to receive user commands while displaying the above-mentioned content, so as to realize interaction with the user.

[0084] It should be understood that the aforementioned processing device can be implemented by a processor reading instructions from memory and executing those instructions, or it can be implemented by a chip circuit.

[0085] Furthermore, the network architecture and business scenarios described in the embodiments of this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided in the embodiments of this application. As those skilled in the art will know, with the evolution of network architecture and the emergence of new business scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0086] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.

[0087] Figure 2 A flowchart illustrating an embodiment of the risk assessment method provided in this application is shown below. Figure 2 As shown, the method includes:

[0088] S201. Obtain multi-dimensional user data associated with the target user.

[0089] In this step, the multi-dimensional user data associated with the target user includes user behavior data and user asset data. User behavior data includes transaction behavior data, click behavior data, browsing behavior data, etc. The acquired multi-dimensional user data refers to data within a preset time period. The preset time period is a pre-defined statistical window, which can be data from the past year or the past six months. For example, the multi-dimensional user data could be user behavior data from the past year, or user asset change data from the past year, etc.

[0090] For example, the collected multi-dimensional user data targets data collected from target users within the asset management platform. The corresponding user behavior data includes transaction behavior data (transfer, consumption, and asset management); click behavior data (button and product clicks within the platform); and browsing behavior data (page dwell time and content type viewed). Correspondingly, user asset data includes changes in account balance, market value of holdings, amount of liabilities, and asset-liability ratio.

[0091] Optionally, one possible approach to obtaining multi-dimensional user data associated with the target user is as follows: Based on the user's authorized usage information, collect multi-dimensional user data from the database within the platform currently used by the user; perform data cleaning on the collected multi-dimensional user data, removing invalid data and supplementing missing data; and perform anonymization processing on sensitive data. This anonymization processing can involve encrypting and processing sensitive information such as the user's ID number and mobile phone number to obtain the corresponding anonymized data.

[0092] For example, to assess the risk of a bank's credit card user, it is necessary to collect multi-dimensional user data for that user over the past year. The corresponding behavioral data could include: the target user's monthly transaction frequency, transaction amount per transaction, number of clicks on installment payment services, and open rate of overdue reminder SMS messages. The corresponding asset data could include: the target user's monthly credit card bill amount, repayment amount, changes in savings account balance, and other loan liabilities.

[0093] S202. Perform feature extraction processing on multi-dimensional user data to obtain event features.

[0094] In this step, event features refer to the characteristics of events related to user behavior or assets within the user's multi-dimensional data. An event refers to a key behavioral or asset change that is meaningful for risk assessment. For example, if a target user's daily redemption amount from a wealth management product on a certain day exceeds 50% of the target user's total assets on that day, this information can be identified as a large redemption event.

[0095] Alternatively, one possible approach to extracting event features from multi-dimensional user data is as follows:

[0096] S2021. Process user multi-dimensional data to obtain event recognition results.

[0097] In this step, processing user multi-dimensional data refers to using pre-defined event recognition rules and pre-trained analysis models to identify events and obtain different events and their corresponding event data.

[0098] Alternatively, one possible way to obtain the event recognition result is as follows:

[0099] a1. Based on event recognition rules, filter user multi-dimensional data to obtain definite events and the first event data corresponding to the definite events.

[0100] In this step, the event identification rule refers to predefined, hard rules that can be directly judged. Events that conform to the rules are defined events, and the corresponding raw data is the first event data. Defined events include the event type corresponding to each event, and the first event data consists of the occurrence time of that type of event and specific event information.

[0101] For example, if an event identification rule is established that a single transaction amount greater than or equal to 50,000, then a large expenditure event is identified. If three transaction records that meet the rule are extracted by traversing the user's multi-dimensional data, then three clear events are identified. The type of clear event is a large expenditure event, and the first event data corresponding to the clear event is the transaction time, transaction amount, and merchant type of these three transactions.

[0102] a2. Based on the analysis model, analyze the user's multi-dimensional data to obtain fuzzy events and the second event data corresponding to the fuzzy events.

[0103] In this step, fuzzy events refer to implicit risk events that cannot be directly determined by hard rules, and the corresponding raw data is the second event data. The analysis model is a machine learning model or a statistical model, used to uncover potential patterns in the user's multi-dimensional data. Fuzzy events include the event type corresponding to each event, as well as the confidence level for inferences made about that event type, and the corresponding second event data includes the raw data for that event type.

[0104] Optionally, the analytical model can be obtained by acquiring historical risk assessment data of the target user platform, training the model based on the historical risk assessment data, and obtaining an analytical model that can be used for fuzzy event recognition.

[0105] Accordingly, the method for analyzing multi-dimensional user data based on the analytical model to obtain fuzzy event and secondary event data can be as follows: Input the target user's multi-dimensional user data for the past year into the analytical model. If it is identified that the target user has made irregular monthly savings and has a high frequency of browsing housing loan products within the past year, then a home purchase preparation event is determined, and the confidence level of this event is set at 80%. Simultaneously, the time and amount of savings and the time of browsing are extracted as the secondary event data. The corresponding fuzzy event type is "home purchase preparation event," with an estimated confidence level of 80%.

[0106] Optionally, when there are high-confidence, high-frequency fuzzy events of the same type in the output of the analysis model, new rules can be defined for the fuzzy events, and the new rules can be updated into the event recognition rules.

[0107] a3. Define the explicit events, the first event data, the fuzzy events, and the second event data as the event identification results.

[0108] In this step, explicit event data, first-time data, ambiguous event data, and second-time event data are merged to form a complete event identification result. Event merging can be done by: deduplication, removing duplicate events, and storing data in a structured manner based on time. The event identification result includes: event identifier, event type, and event data fields.

[0109] In steps a1-a3 of this embodiment, a recognition scheme is formed by filtering explicit events and corresponding data based on event recognition rules, obtaining fuzzy events and corresponding data based on analysis models, and then integrating the two types of results to obtain the event recognition results. This scheme takes into account both the accurate filtering of explicit events driven by rules and the in-depth mining of fuzzy events driven by models. It ensures the accuracy of the recognition of known typical events and can also capture potential atypical events, enriching the coverage of event recognition and making the event features extracted later more comprehensive.

[0110] S2022. Perform feature processing on the event identification results to obtain event features.

[0111] In this step, feature processing refers to converting the structured event recognition results into numerically computable feature vectors for subsequent model input.

[0112] Alternatively, the feature processing can be performed as follows:

[0113] b1. Perform event coding for categorized events, specifically by converting the event type into one-hot coding.

[0114] b2. Quantify numerical events, specifically by standardizing the key parameters of the event to the [0,1] range, or by unifying the units of the key parameters.

[0115] b3. Aggregate time-series events, specifically by calculating statistical values ​​for multiple events of the same type.

[0116] For example, 3 overdue payments with an average overdue period of 5 days are converted into features [number of overdue payments = 3, average number of overdue days = 5, overdue label = 1]; small-amount multiple overseas transactions are converted into features [number of overseas transactions = 8, average amount per transaction = 0.2 million yuan, implicit risk label = 1].

[0117] In steps S2021 to S2022 of this embodiment, multi-dimensional user data is first processed to obtain event identification results, and then the event identification results are characterized to obtain event features. By processing in stages, the blindness of direct feature extraction is avoided, providing more accurate micro-feature support for subsequent risk assessment.

[0118] S203. Perform statistical calculations on multi-dimensional user data based on preset calculation rules to obtain global statistical features corresponding to the preset calculation rules.

[0119] In this step, global statistical features are macro-level statistical indicators of the target user's full data, used to reflect the overall behavioral or asset trends of the target user, complementing the event features obtained in step S202 above. Preset calculation rules are statistical formulas or indicator definitions developed based on business needs.

[0120] For example, the preset calculation rules include: behavioral rules, asset rules, and risk rules; among which, behavioral rules include: calculation of total consumption amount in the past year, calculation of average monthly transaction frequency, and calculation of average browsing time of financial products; asset rules include: calculation of average monthly growth rate of total assets, calculation of emergency fund coverage ratio, calculation of load ratio, and calculation of monthly volatility of savings account; risk rules include: calculation of delinquency rate, calculation of the proportion of defaults to total transaction frequency, etc.

[0121] For example, the preset calculation rules are as follows: the formula for calculating the average monthly consumption amount is: total consumption amount in the past year / 12; the formula for calculating the debt-to-asset ratio is: liabilities / total assets; the formula for calculating the emergency fund coverage ratio is: liquid assets / average monthly expenditure; after calculating the data of a certain user, the global statistical characteristics are obtained as follows: [average monthly consumption: 12,000 yuan, debt-to-asset ratio: 0.35, emergency fund coverage ratio: 1.5].

[0122] S204. Based on event characteristics and global statistical characteristics, obtain comprehensive prediction results corresponding to multi-dimensional user data.

[0123] In this step, the comprehensive prediction result is obtained by integrating local event characteristics and macro-level global statistical characteristics, and outputting a comprehensive prediction result of user risk through a combination of models or rules.

[0124] Optionally, there are two specific implementation methods: obtaining comprehensive prediction results using a rule-driven approach or obtaining comprehensive prediction results using a model-driven approach.

[0125] The rule-driven approach can be as follows: pre-set feature weights; perform weighted summation on each feature according to the weights to obtain a comprehensive prediction score, which is then used as the comprehensive prediction result.

[0126] The model-driven approach can be to concatenate event features and global features into a fused feature vector, input it into a pre-trained machine learning model, and the model outputs a comprehensive prediction probability.

[0127] S205. Determine the risk assessment results for the target users based on the comprehensive forecast results.

[0128] In this step, determining the risk assessment results refers to mapping the comprehensive forecast results to a risk level that is understandable to the business; it is a transformation step from data prediction to business conclusions. The corresponding risk level is determined based on the predicted probabilities in the comprehensive forecast results.

[0129] For example, when the comprehensive prediction result is a single-objective prediction: the mapping relationship between the single objective and the probability and risk level is determined as follows: [0,0.3), low risk; [0.3,0.6), medium risk; [0.6,1.0], high risk.

[0130] It should be noted that in the following Figure 3 In the illustrated embodiment, steps S204 and S205 are explained in further detail, and will not be elaborated upon here.

[0131] The risk assessment method provided in this embodiment acquires multi-dimensional user data associated with the target user. Feature extraction is performed on the multi-dimensional user data to obtain event features; statistical calculations are then performed on the multi-dimensional user data based on preset calculation rules to obtain global statistical features. A comprehensive prediction result corresponding to the multi-dimensional user data is obtained based on the event features and global statistical features, and the risk assessment result for the target user is determined based on the comprehensive prediction result. This application achieves comprehensive and accurate data collection through the acquisition of multi-dimensional data; it mines event features related to user behavior or assets from the multi-dimensional user data using feature extraction; it statistically obtains global statistical features of the multi-dimensional user data using rule calculation; and it combines the two features to obtain a corresponding comprehensive prediction result, achieving the purpose of accurate prediction through feature fusion. The obtained comprehensive prediction result is used to determine the risk assessment result for the target user, further realizing the generation of accurate risk assessment conclusions. Compared with existing technologies, this avoids the information lag inherent in relying on questionnaire surveys for information collection. Simultaneously, it achieves comprehensive information collection through multi-dimensional user data, improves the accuracy of risk assessment by combining the comprehensive prediction results generated from the two features, and enhances the credibility of risk assessment by using the accurate comprehensive prediction results to generate assessment results.

[0132] Figure 3 The flowchart of Embodiment 2 of the risk assessment method provided in this application is shown below. Figure 3 As shown, the method includes:

[0133] S301. Encode the global statistical features to obtain sub-score features.

[0134] In this step, sub-score features refer to the dimensional encoding of global statistical features, which breaks down a single global feature into sub-scores corresponding to the types of preset calculation rules. Each sub-score corresponds to a specific business objective, such as repayment ability, consumption ability, and default probability.

[0135] Alternatively, the sub-fraction features can be calculated in the following ways:

[0136] c1. Split global statistical features according to preset calculation rule types.

[0137] For example, global statistical features can be broken down into three categories: repayment ability features, consumption ability features, and default risk features.

[0138] c2. Encode each type of feature independently.

[0139] For example, the debt-to-equity ratio can be transformed into a repayment ability sub-score, and the average monthly consumption amount can be transformed into a consumption ability sub-score, through linear transformation.

[0140] c3. Integrate all sub-fractions into a sub-fraction feature vector.

[0141] For example, the global statistical features [average monthly consumption: 12,000 yuan, debt-to-asset ratio: 0.35, delinquency rate: 0.02] are broken down into: repayment ability sub-score: 0.7, indicating a low debt-to-asset ratio and strong repayment ability; consumption ability sub-score: 0.6, indicating moderate average monthly consumption; and default risk sub-score: 0.2, indicating a low delinquency rate. The resulting sub-score features are [0.7, 0.6, 0.2].

[0142] S302. Concatenate the event features and sub-score features to obtain the feature vector of the user's multi-dimensional data.

[0143] In this step, the feature vector of user multi-dimensional data refers to the feature concatenation of local event features and sub-dimensional sub-score features to form a complete feature vector containing user key events and sub-dimensional capabilities, which serves as the input to the multi-objective prediction model.

[0144] For example, if there are event features [3,5,1] and sub-score features [0.7,0.6,0.2], then the fused feature vector is [3,5,1,0.7,0.6,0.2].

[0145] S303. Input the feature vector into the multi-objective prediction model to obtain the comprehensive prediction result.

[0146] In this step, the multi-objective prediction model is a model that can output prediction results for multiple business objectives simultaneously. The prediction objectives in the multi-objective prediction model correspond one-to-one with the calculation rule types in the preset calculation rules, and the comprehensive prediction result includes at least the prediction probability corresponding to each prediction objective.

[0147] Alternatively, one possible way to obtain the comprehensive prediction results is as follows:

[0148] d1. Construct a multi-objective prediction model.

[0149] In this step, the multi-objective prediction model can employ multi-output neural networks, multi-objective gradient boosting trees, or other similar models.

[0150] d2. Determine the prediction target, which must correspond one-to-one with the preset calculation rule type.

[0151] For example, the preset calculation rule types include: repayment ability, consumption ability, and default risk; then the model prediction targets include: repayment ability probability, consumption ability probability, and default risk probability.

[0152] d3. Input the concatenated feature vector into the multi-objective prediction model. The multi-objective prediction model outputs the prediction probability of each object, which is the comprehensive prediction result.

[0153] For example, after the feature vector is input into the multi-objective model, the output comprehensive prediction result is: [Probability of repayment ability: 0.85, probability of consumption ability: 0.7, probability of default risk: 0.15].

[0154] Optionally, after obtaining the comprehensive prediction results, the accuracy of the multi-objective prediction model can be calculated using a pre-prepared validation dataset. When the accuracy is detected to be lower than a preset threshold, it can be determined which object's prediction accuracy is below the preset threshold; the global statistical features corresponding to that object can be identified, and then a validity analysis can be performed on these global statistical features. Based on the validity analysis results, a rule adjustment prompt can be generated. The rule adjustment prompt is used to drive the adjustment of the rules for calculating the global statistical features in the preset calculation rules.

[0155] S304. Determine the risk assessment result based on the predicted probability corresponding to each predicted target.

[0156] In this step, the risk assessment results should include at least the risk level of each predicted target and the overall risk level of the target user.

[0157] Alternatively, one possible way to determine the risk assessment results is as follows:

[0158] e1. Develop risk level standards for sub-targets.

[0159] For example: A default risk probability range of [0, 0.2) corresponds to a default risk level of R1; a repayment ability probability range of [0.8, 1.0] corresponds to a repayment ability level of R4. A default risk probability range of [0.2, 0.5) corresponds to a default risk level of R2; a repayment ability probability range of [0.5, 0.8) corresponds to a repayment ability level of R2. A default risk probability range of [0.5, 1.0] corresponds to a default risk level of R4; a repayment ability probability range of [0, 0.5) corresponds to a repayment ability level of R1.

[0160] e2. Match the probability of each predicted target to the corresponding level to obtain the risk level of the sub-target.

[0161] e3. Formulate rules for calculating comprehensive risk levels.

[0162] For example, the comprehensive risk level calculation rule can be: default risk weight 0.5, repayment ability weight 0.3, consumption ability weight 0.2, comprehensive risk level = 0.5 × default level + 0.3 × repayment ability level + 0.2 × consumption ability level.

[0163] In steps S301-S304 of this embodiment, a three-step process—global statistical feature encoding, concatenation of event features and sub-score features, and input of the multi-objective prediction model output—ensures a one-to-one correspondence between the model's prediction targets and calculation rule types, and that the comprehensive prediction result includes prediction probabilities. Simultaneously, risk levels are determined based on the prediction probabilities of each prediction target, and the risk assessment result includes the risk levels of each target and the comprehensive risk level. Through feature encoding and concatenation, effective fusion of micro-level event features and macro-level global statistical features is achieved, enhancing the representational power of feature vectors. The design of the multi-objective prediction model strengthens the correlation between the comprehensive prediction result and global statistical rules, allowing the output prediction probabilities to directly support risk level classification. By clarifying the two-layer structure of the risk assessment result, the risk assessment conclusions become more detailed and instructive.

[0164] S305. Generate a set of constraints based on preset constraint rules, the risk level of each prediction target, and the comprehensive risk level of the target user.

[0165] In this step, the set of constraints forms the rule boundary for product recommendations. It is based on the user's risk level and is used to filter out products that do not meet the user's risk tolerance. The preset constraint rules are risks defined by the business side and also serve as product matching rules.

[0166] Alternatively, one possible implementation for generating the set of constraints is:

[0167] f1. Extract preset constraint rules from the preset constraint rule library, for example:

[0168] Rule 1: If the overall risk level is high (R4), it is prohibited to recommend high-leverage financial products.

[0169] Rule 2: For borrowers with a repayment ability rating of weak R1, it is prohibited to recommend large consumer loan products.

[0170] Rule 3: For default risk level of low R1, recommendations for credit card installment plans and low-risk financial products are permitted.

[0171] f2. Match the user's risk level with the preset constraint rules and extract the applicable constraint conditions.

[0172] f3. Store the constraints in a structured manner to form a set of constraints.

[0173] For example, if the target user's overall risk level is low risk R1, strong repayment ability R4, and low default risk R1, then the generated set of constraints can be: [Allow recommendations for low-risk financial products, allow recommendations for credit card installment plans, prohibit recommendations for high-leverage asset management products].

[0174] S306. Obtain the user-side characteristics of the target users and the target product library.

[0175] In this step, user-side features refer to the core user characteristics used for product matching, which integrate user asset data, risk assessment results, and global statistical features. The target product library refers to the collection of all products available on the platform to which the target user belongs.

[0176] Alternatively, one possible implementation for obtaining user-side features is as follows:

[0177] S3061. Obtain the target user's asset data based on multi-dimensional user data.

[0178] In this step, acquiring asset data refers to extracting asset-related fields from multi-dimensional user data, such as total assets, savings balance, and holdings in asset management products.

[0179] S3062 defines asset data, risk assessment results, and global statistical characteristics as user-side features.

[0180] In this step, user-side characteristics are determined by integrating asset data, risk assessment results, and global statistical features. For example, user-side characteristics could be: [Total assets = 500,000 yuan, savings balance = 100,000 yuan, overall risk level = low, average monthly consumption = 12,000 yuan].

[0181] In steps S3061-S3062 of this embodiment, asset data is obtained based on multi-dimensional user data, and then the asset data, risk assessment results, and global statistical features are integrated into user-side features. This avoids the limitations of single-dimensional information, provides more comprehensive user profile support for product recommendations, and further improves the accuracy of the recommendation scheme.

[0182] S307. Extract the core features of the products that correspond to the product information in the target product library.

[0183] In this step, the core features of the product are the key attributes of the product, which are used to match with user-side features and reflect the core information of the product, such as risk, benefits, and target audience.

[0184] Alternatively, one possible way to extract the core features of a product is as follows:

[0185] g1. Determine the core feature dimensions of the product.

[0186] For example, the core features of a product can be: product type (wealth management, loan, or installment product); product risk level (low R1, medium R2, high R3, high R4); minimum investment amount or loan amount; expected return or interest rate; and target audience tags.

[0187] g2. Traverse the target product library, extract the core features of each product and standardize them.

[0188] g3. Store the core features of the product as feature vectors.

[0189] For example, the core characteristics of a low-risk investment product can be: [Product type: Investment product, Risk level: R1, Minimum investment amount: 10,000 yuan, Expected return: 3.5%].

[0190] S308. Based on the set of constraints, user-side features, and core product features, output a product recommendation scheme for the target user.

[0191] The purpose of this step is to output a personalized product recommendation solution based on the matching degree between user-side features and core product features, combined with the filtering rules of the constraint set.

[0192] Alternatively, one possible implementation of generating product recommendation schemes is as follows:

[0193] h1. Feature matching calculation: Calculate the similarity between user-side features and each product's core features. The higher the similarity, the higher the matching degree.

[0194] h2. Constraint filtering: Remove products that do not meet the set of constraints.

[0195] h3. Sorting and Recommendation: Sort products from highest to lowest matching degree, select the top N products to generate a recommendation plan, and attach the reasons for the recommendation.

[0196] For example, a user's characteristics are low risk (R1), total assets of 500,000 yuan, and savings balance of 100,000 yuan. After matching and filtering, the following recommended solutions are output: Option A: Low-risk wealth management, minimum investment of 10,000 yuan, expected return of 3.5%; Reason for recommendation: Matches risk level, low minimum investment amount. Option B: Credit card installment plan, fee rate of 0.5% / period; Reason for recommendation: Strong repayment ability, suitable for optimizing cash flow.

[0197] Optionally, the product recommendation scheme in this step can also be generated using an optimization model. The optimization objective is determined based on the target user's platform and multi-dimensional user data. The optimization is then performed to find the optimal solution. For example, if the target user's platform is an asset management platform and the target user frequently browses financial products, the optimization objective can be set as maximizing financial returns. Then, the target user's user-side characteristics and the obtained core product characteristics are input into the optimization model for further optimization. Simultaneously, the constraints in the constraint set are used as constraints for model optimization, and the final model output is the product recommendation scheme.

[0198] In steps S305-S308 of this embodiment, a set of constraint conditions is generated based on preset constraint rules, each target risk level, and the overall risk level. By acquiring user-side characteristics and the target product library, and extracting core product features, a product recommendation scheme is finally output. This directly links the risk assessment results with the product recommendations, limits the recommendation scope through the set of constraint conditions, and achieves precise matching by combining user-side characteristics and core product features. This avoids a disconnect between risk assessment results and asset management advice, improving the applicability and rationality of the recommendation scheme.

[0199] Figure 4 A schematic diagram of the risk assessment device provided in this application is shown below. Figure 4 As shown, the risk assessment device provided in this embodiment includes:

[0200] The acquisition module 401 is used to acquire multi-dimensional user data associated with the target user.

[0201] The first processing module 402 is used to perform feature extraction processing on multi-dimensional user data to obtain event features.

[0202] The second processing module 403 is used to perform statistical calculations on the user's multi-dimensional data based on preset calculation rules to obtain the global statistical features corresponding to the preset calculation rules.

[0203] The third processing module 404 is used to obtain a comprehensive prediction result corresponding to the user's multi-dimensional data based on event characteristics and global statistical characteristics.

[0204] The fourth processing module 405 is used to determine the risk assessment results of the target user based on the comprehensive prediction results.

[0205] Optionally, in one possible implementation, the first processing module 402 is further configured to:

[0206] The event recognition results are obtained by processing multi-dimensional user data.

[0207] The event identification results are processed to obtain event features.

[0208] Optionally, in one possible implementation, the first processing module 402 is further configured to:

[0209] Based on event recognition rules, multi-dimensional user data is filtered to obtain definite events and the first event data corresponding to the definite events.

[0210] Based on the analysis model, multi-dimensional user data is analyzed to obtain fuzzy events and corresponding second event data.

[0211] The definite events, the first event data, the fuzzy events, and the second event data are identified as the event recognition results.

[0212] Alternatively, in one possible implementation, the third processing module 404 is further configured to:

[0213] The global statistical features are encoded to obtain sub-score features.

[0214] By concatenating event features and sub-score features, a feature vector of multi-dimensional user data is obtained.

[0215] The feature vector is input into the multi-objective prediction model to obtain the comprehensive prediction result. The prediction target in the multi-objective prediction model corresponds one-to-one with the calculation rule type in the preset calculation rule. The comprehensive prediction result includes at least the prediction probability corresponding to each prediction target.

[0216] The corresponding fourth processing module 405 is also used for:

[0217] The risk assessment results are determined based on the predicted probability corresponding to each prediction target; the risk assessment results include at least the risk level of each prediction target and the overall risk level of the target user.

[0218] Alternatively, in one possible implementation, the fourth processing module 405 is further configured to:

[0219] A set of constraints is generated based on preset constraint rules, the risk level of each prediction target, and the comprehensive risk level of the target user.

[0220] Obtain user-side characteristics of the target users and the target product library.

[0221] Extract the core features of the products that correspond to the product information in the target product library.

[0222] Based on the set of constraints, user-side features, and core product features, output a product recommendation scheme for the target user.

[0223] Alternatively, in one possible implementation, the fourth processing module 405 is further configured to:

[0224] Based on multi-dimensional user data, obtain the target user's asset data.

[0225] Asset data, risk assessment results, and global statistical characteristics are identified as user-side features.

[0226] The risk assessment device provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0227] Figure 5 A schematic diagram of the structure of the electronic device provided in this application. Figure 5 As shown, the electronic device provided in this embodiment includes at least one processor 501 and a memory 502. Optionally, the device further includes a communication component 503. The processor 501, memory 502, and communication component 503 are connected via a bus 504.

[0228] In a specific implementation, at least one processor 501 executes computer execution instructions stored in memory 502, causing at least one processor 501 to perform the above-described method.

[0229] The specific implementation process of processor 501 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0230] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0231] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0232] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0233] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0234] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0235] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0236] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0237] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0238] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0239] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0240] If the functionality is implemented as 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 this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0241] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0242] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A risk assessment method, characterized by, The method includes: Obtain multi-dimensional user data associated with the target user; Feature extraction processing is performed on the user's multi-dimensional data to obtain event features; Statistical calculations are performed on the user's multi-dimensional data based on preset calculation rules to obtain global statistical features corresponding to the preset calculation rules; Based on the event characteristics and the global statistical characteristics, a comprehensive prediction result corresponding to the user's multi-dimensional data is obtained; The risk assessment result for the target user is determined based on the comprehensive prediction results.

2. The method of claim 1, wherein, The process of extracting features from the user's multi-dimensional data to obtain event features includes: The multi-dimensional user data is processed to obtain event recognition results; The event identification results are subjected to feature processing to obtain the event features.

3. The method of claim 2, wherein, The process of processing the user's multi-dimensional data to obtain event recognition results includes: Based on event recognition rules, the user's multi-dimensional data is filtered to obtain specific events and the first event data corresponding to the specific events; Based on the analysis model, the user's multi-dimensional data is analyzed to obtain fuzzy events and the second event data corresponding to the fuzzy events; The explicit event, the first event data, the fuzzy event, and the second event data are determined as the event identification result.

4. The method of claim 1, wherein, The step of obtaining a comprehensive prediction result corresponding to the user's multi-dimensional data based on the event characteristics and the global statistical characteristics includes: The global statistical features are encoded to obtain sub-score features; The event features and the sub-score features are concatenated to obtain the feature vector of the user's multi-dimensional data; The feature vector is input into a multi-objective prediction model to obtain the comprehensive prediction result; the prediction targets in the multi-objective prediction model correspond one-to-one with the calculation rule types in the preset calculation rules, and the comprehensive prediction result includes at least the prediction probability corresponding to each prediction target; The step of determining the risk assessment result of the target user based on the comprehensive prediction result includes: The risk assessment result is determined based on the predicted probability corresponding to each of the predicted targets; the risk assessment result includes at least the risk level of each of the predicted targets and the comprehensive risk level of the target user.

5. The method of claim 4, wherein, After determining the risk assessment result of the target user based on the comprehensive prediction result, the method further includes: A set of constraints is generated based on preset constraint rules, the risk level of each prediction target, and the comprehensive risk level of the target user. Obtain the user-side characteristics of the target user and the target product library; Extract the core features of the products that correspond to the product information in the target product library; Based on the set of constraints, the user-side features, and the core product features, a product recommendation scheme for the target user is output.

6. The method according to claim 5, characterized in that, The acquisition of the user-side features of the target user includes: Based on the multi-dimensional user data, obtain the target user's asset data; The asset data, the risk assessment results, and the global statistical features are determined as the user-side features.

7. A risk assessment device, characterized in that, include: The acquisition module is used to acquire multi-dimensional user data associated with the target user; The first processing module is used to perform feature extraction processing on the user's multi-dimensional data to obtain event features; The second processing module is used to perform statistical calculations on the user's multi-dimensional data based on preset calculation rules to obtain global statistical features corresponding to the preset calculation rules. The third processing module is used to obtain a comprehensive prediction result corresponding to the user's multi-dimensional data based on the event characteristics and the global statistical characteristics. The fourth processing module is used to determine the risk assessment result of the target user based on the comprehensive prediction result.

8. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method of any one of claims 1 to 6.

Images