Decision result interpretation system and method based on financial risk control user portrait

Abstract

The invention designs a decision result interpretation system and method based on a financial risk control user portrait, and the method comprises the steps: collecting user data, and constructing a user portrait; inputting the preprocessed user portrait data and other risk indexes into a decision model, outputting a risk assessment result, and storing the risk assessment result; in combination with key information in the user portrait data, mapping a relationship between a risk assessment result and an input feature, and using an explanatory algorithm to explain the risk decision model; the decision interpretation result is displayed through a chart, a flow chart and a key index mode; and according to the user feedback and the interpretation effect, optimizing the interpretation algorithm and the risk decision model to form a closed-loop feedback mechanism. According to the method, the key information is extracted by using the user portrait, the key information and the important variables in the decision model are mapped to form a decision result explanation basis, and an explanatory algorithm is introduced to explain the black box risk decision model, so that the transparency of the decision result is realized.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of financial risk control, and particularly relates to a decision result explanation system and method based on user portrait of financial risk control. BACKGROUND

[0002] Financial risk control decisions are usually based on a large amount of user data and transaction records, and risk assessment and decision results are obtained after data collection, preprocessing and training of machine learning models (such as reinforcement learning).

[0003] The existing technical solutions mainly include the following: 1. Risk control decision system based on black box model: using reinforcement learning and deep learning for risk prediction, the model has good effect, but the decision process is not transparent, and lacks effective explanation mechanism. 2. Traditional user portrait construction method: static portrait is established by user historical data, behavior data and transaction records to assist risk assessment, but it is not combined with decision process for explanation. 3. Partial explanation algorithm attempt: model explanation methods such as LIME and SHAP can locally explain the black box model, but they are not closely combined with user portrait data in the financial risk control system, and the explanation results are difficult to fully reflect the internal logic of risk decision. Overall, the existing technical solutions have achieved certain results in decision accuracy and strategy optimization, but there are deficiencies in decision result explanation, user portrait and model key feature mapping and visual display.

[0004] Although the current system can improve the accuracy and efficiency of risk decision and perform well in risk judgment, the decision process often lacks transparency. Although the black box system based on reinforcement learning and deep model can improve the accuracy of risk identification, it cannot intuitively explain the model output, which makes it difficult for users and regulatory authorities to be convinced. The existing system ignores the explainability of decision results. Although the user portrait is constructed, the key features in the user portrait are not organically combined with the risk decision results. The internal relationship between the user portrait information and the decision results is difficult to intuitively explain, which often makes it difficult for regulatory authorities and business personnel to understand why the model gives a certain risk rating, and further leads to understanding barriers in risk review, regulatory compliance and business communication.

[0005] Therefore, in-depth mining of user portrait and explanation of key features of decision model have become urgent problems to be solved in the field of financial risk control. SUMMARY

[0006] In order to solve the above problems, the application designs a decision result explanation system and method based on user portrait of financial risk control, so as to construct an explanation module by using user portrait data and key features in the decision model, realize visual explanation of the decision result, and improve the transparency and decision trust of the system.

[0007] The application discloses a decision result explanation system based on a financial risk control user portrait, and relates to the technical field of user portrait. The data collection module collects user data and transmits the collected user data to the user portrait construction module; the user portrait construction module constructs a user portrait based on the collected user data and transmits user portrait data to the risk decision module. The risk decision module outputs a risk assessment result based on the user portrait data and stores the risk assessment result. The explanation module uses an explanatory algorithm to explain the risk decision model based on the risk assessment result and the user portrait, and transmits a decision explanation result to the interactive visual display module for display. The application discloses a decision result explanation method based on a financial risk control user portrait, and relates to the technical field of user portrait. Step S1, collecting user data and constructing a user portrait. Step S2, inputting preprocessed user portrait data and other risk indicators into a decision model, outputting a risk assessment result, and storing the risk assessment result. Step S3, combining key information in the user portrait data, mapping the relationship between the risk assessment result and input features, and using an explanatory algorithm to explain the risk decision model. Step S4, displaying a decision explanation result in the form of a chart, a flowchart or a key indicator. Step S5, optimizing the explanatory algorithm and the risk decision model according to user feedback and explanation effect, and forming a closed-loop feedback mechanism.

[0008] Preferably, the step S1 comprises the following steps. Step S11, collecting user basic information, behavior data and historical credit data by calling various data interfaces. Step S12, performing cleaning, denoising, format conversion and standardization processing on the collected data. Step S13, constructing a user portrait based on the collected data by using clustering, statistical analysis and feature extraction methods. Preferably, the user portrait comprises user basic features, behavior preferences and credit status.

[0009] Preferably, the step S2 comprises the following steps. Step S21, inputting preprocessed user portrait data and other risk indicators into a decision model. Step S22, training the decision model by using a reinforcement learning algorithm and outputting a risk assessment result. Step S23, store the risk assessment result and the corresponding user portrait data in the database.

[0010] Preferably, the step S3 comprises: Step S31, extract the main user features affecting the risk assessment result according to the weights and importance indexes obtained in the process of training the decision model with the reinforcement learning algorithm in step S22; Step S32, perform local interpretation on a single risk assessment result using LIME and SHAP algorithms, and calculate the contribution of each user feature to the risk assessment result score; Step S33, map the extracted key user features to the risk assessment result, and generate a detailed explanation report.

[0011] Preferably, the step S4 comprises: Step S41, generate charts, flowcharts and key indicators using a visualization engine; Step S42, visually present the risk assessment result using user interaction operations, and collect user feedback on the explanation report.

[0012] The advantages and effects of the present application are as follows: The decision result explanation system and method based on financial risk control user portrait of the present application comprises the following steps: step S1, collect user data and build a user portrait; step S2, input the preprocessed user portrait data and other risk indicators into a decision model, output a risk assessment result, and store the risk assessment result; step S3, map the relationship between the decision result and the input features based on the key information in the user portrait, and explain the risk decision model; step S4, display the decision explanation result through charts, flowcharts and key indicators; step S5, optimize the explanation algorithm and the decision model according to user feedback and explanation effect, and form a closed-loop feedback mechanism. Through the above design, the present application can extract key features from user portrait data and map them to important variables in the risk decision model to form the basis for decision result explanation, and introduce LIME and SHAP algorithms to explain the black-box risk decision model and realize the transparency of the decision result. Finally, the decision explanation result is visually displayed through charts, flowcharts and other means, supporting user interaction query, and improving system transparency and regulatory compliance.

[0013] The above description is only a summary of the technical solutions of the present application. In order to more clearly understand the technical means of the present application, the contents of the specification can be implemented, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following will be described in detail with the preferred embodiments of the present application and with the help of the accompanying drawings.

[0014] The above and other objects, advantages and features of the present application will become more apparent by describing in detail the embodiments thereof with reference to the attached drawings in which: BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative effort. In all the drawings, similar elements or parts are generally identified by similar reference signs. In the drawings, the elements or parts are not necessarily drawn according to the actual proportions.

[0016] Figure 1 A flow chart of a decision result explanation method based on a financial risk control user portrait designed by the present application. DETAILED DESCRIPTION

[0017] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, not all the embodiments. In the following description, specific details such as specific configurations and components are provided only to help a comprehensive understanding of the embodiments of the present application. Therefore, those skilled in the art should understand that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. In addition, in order to be clear and concise, the description of known functions and structures is omitted in the embodiments.

[0018] It should be understood that the "one embodiment" or "the embodiment" mentioned throughout the specification means that the specific features, structures or characteristics related to the embodiment are included in at least one embodiment of the present application. Therefore, "one embodiment" or "the embodiment" appearing throughout the specification does not necessarily mean the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner.

[0019] In addition, the reference numerals and / or letters can be repeated in different examples of the present application. Such repetition is for the purpose of simplification and clarity, and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed.

[0020] The term "and / or", used in the present document, only describes the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B can mean that A exists alone, B exists alone, and A and B exist simultaneously. The term " / and" in the present document describes another association relationship of the associated objects, which means that there can be two relationships, for example, A / and B can mean that A exists alone and A and B exist simultaneously. In addition, the character " / " in the present document generally represents an "or" relationship between the associated objects before and after it.

[0021] The term "at least one" in the present document only describes the association relationship of the associated objects, which means that there can be three relationships, for example, at least one of A and B can mean that A exists alone, A and B exist simultaneously, and B exists alone.

[0022] It should also be noted that, in the present document, relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion. Embodiments

[0023] A decision result explanation system based on financial risk control user portrait includes a data acquisition module, a user portrait construction module, a risk decision module, an explanation module, and an interactive visual display module. The data acquisition module acquires user data and transmits the acquired user data to the user portrait construction module. The user portrait construction module constructs a user portrait based on the acquired user data and transmits the user portrait data to the risk decision module. The risk decision module outputs a risk assessment result based on the user portrait data and stores the risk assessment result. The explanation module uses an explanatory algorithm to explain the risk decision model based on the risk assessment result and the user portrait, and transmits the decision explanation result to the interactive visual display module for display.

[0024] Please refer to Figure 1 Based on the decision result explanation system, the present embodiment further provides a decision result explanation method based on financial risk control user portrait, which includes the following steps: Step S1, acquiring user data and constructing a user portrait; Step S2, inputting the preprocessed user portrait data and other risk indicators into the decision model, outputting a risk assessment result, and storing the risk assessment result; Step S3, combine the key information in the user portrait data, map the relationship between the risk assessment result and the input features, and use the explanatory algorithm to explain the risk decision model; Step S4, display the decision explanation result through charts, flowcharts, and key indicators; Step S5, optimize the explanatory algorithm and risk decision model according to user feedback and explanation effect, form a closed-loop feedback mechanism; establish a feedback collection interface and a reward mechanism to realize dynamic adjustment of model parameters and strategy optimization Further, the step S1 comprises: Step S11, call each data interface to collect user basic information, behavior data and historical credit data; Further, the data collection module collects user data from multiple data sources, including transaction data, behavior data, credit records, etc. Step S12, use ETL technology to clean, denoise, format convert and standardize the collected data; Step S13, based on the collected data, use clustering, statistical analysis and feature extraction methods to build a user portrait; Further, the user portrait includes user basic characteristics, behavior preferences and credit status.

[0025] Further, the step S2 comprises: Step S21, input the preprocessed user portrait data and other risk indicators into the decision model; Step S22, use reinforcement learning algorithm to train the decision model and output the risk assessment result, the decision model is a black box model constructed by using deep learning and reinforcement learning technology, and the decision strategy is continuously optimized through feedback mechanism; Step S23, store the risk assessment result and the corresponding user portrait data in the database.

[0026] Further, the step S3 comprises: Step S31, according to the weight and importance index obtained in the process of training the decision model by using the reinforcement learning algorithm in step S22, extract the main user features affecting the risk assessment result; Step S32, use LIME and SHAP algorithms to perform local explanation on a single risk assessment result, and calculate the contribution of each user feature to the risk assessment result score; Step S33, map the extracted key user features with the risk assessment result, and generate a detailed explanation report.

[0027] Further, the step S4 comprises: Step S41, generating a chart, a flowchart and key indicators by using a visualization engine; Step S42, visually presenting the risk assessment result by using user interaction operation, and collecting user feedback on the interpretation report. The closed-loop feedback mechanism ensures that the interpretation effect and the decision model can be continuously self-improved and iteratively optimized.

[0028] The application designs a decision result interpretation method based on financial risk control user portrait, which comprises the following steps: S1, collecting user data and constructing a user portrait; S2, inputting the preprocessed user portrait data and other risk indicators into a decision model, outputting a risk assessment result, and storing the risk assessment result; S3, combining the key information in the user portrait, mapping the relationship between the decision result and the input features, and interpreting the risk decision model; S4, displaying the decision interpretation result through a chart, a flowchart and key indicators; S5, optimizing the interpretation algorithm and the decision model according to user feedback and interpretation effect, and forming a closed-loop feedback mechanism. Through the above design, the application can extract key features from user portrait data and map them with important variables in the risk decision model to form the basis for interpreting the decision result, and introduce LIME and SHAP algorithms to interpret the black-box risk decision model and realize the transparency of the decision result. Finally, the decision interpretation result is visually displayed through a chart, a flowchart and other methods, user interaction query is supported, and the system transparency and regulatory compliance are improved.

[0029] The above only describes the preferred embodiments of the application, and does not limit the protection scope of the application. For those skilled in the art, the application can have various changes and variations. Any changes, modifications, replacements, integrations and parameter changes made to these embodiments within the spirit and principles of the application, which can realize the same functions without departing from the principles and spirit of the application, fall within the protection scope of the application.

Claims

1. A decision result interpretation system based on financial risk control user profiles, characterized in that, It includes a data acquisition module, a user profile building module, a risk decision-making module, an explanation module, and an interactive visualization module; The data acquisition module collects user data and transmits the collected user data to the user profile building module. The user profile building module builds user profiles based on the collected user data and transmits the user profile data to the risk decision-making module; The risk decision-making module outputs risk assessment results based on user profile data and stores the risk assessment results; The explanation module uses an interpretive algorithm to explain the risk decision-making model based on risk assessment results and user profiles. The decision interpretation results are then transmitted to the interactive visualization module for display.

2. A method for interpreting decision results based on financial risk control user profiles, characterized in that, Includes the following steps: Step S1: Collect user data and build user profiles; Step S2: Input the preprocessed user profile data and other risk indicators into the decision model, output the risk assessment results, and store the risk assessment results; Step S3: Combine key information from user profile data to map the relationship between risk assessment results and input features, and use interpretive algorithms to interpret the risk decision model; Step S4: Present the decision interpretation results using charts, flowcharts, and key indicators; Step S5: Based on user feedback and the effectiveness of the explanation, optimize the explanatory algorithm and risk decision-making model to form a closed-loop feedback mechanism.

3. The method for interpreting decision results based on financial risk control user profiles according to claim 2, characterized in that, Step S1 includes: Step S11: Call each data interface to collect user basic information, behavioral data, and historical credit data; Step S12: Clean, denoise, convert formats, and standardize the collected data; Step S13: Based on the collected user data, construct user profiles using clustering, statistical analysis, and feature extraction methods.

4. The method for interpreting decision results based on financial risk control user profiles according to claim 3, characterized in that, The user profile data includes basic user characteristics, behavioral preferences, and credit status.

5. The method for interpreting decision results based on financial risk control user profiles according to claim 4, characterized in that, Step S2 includes: Step S21: Input the preprocessed user profile data and other risk indicators into the decision model; Step S22: Train the decision model using a reinforcement learning algorithm and output the risk assessment results; Step S23: Store the risk assessment results and corresponding user profile data into the database.

6. The method for interpreting decision results based on financial risk control user profiles according to claim 5, characterized in that, Step S3 includes: Step S31: Based on the weights and importance indicators obtained in step S22 during the training of the decision model using reinforcement learning algorithms, extract the main user features that affect the risk assessment results. Step S32: Use LIME and SHAP algorithms to perform local interpretation of individual risk assessment results and calculate the contribution of each user characteristic to the risk assessment result score. Step S33: Map the extracted key user characteristics to the risk assessment results and generate a detailed explanatory report.

7. The method for interpreting decision results based on financial risk control user profiles according to claim 6, characterized in that, Step S4 includes: Step S41: Use a visualization engine to generate charts, flowcharts, and key metrics; Step S42: Present the risk assessment results intuitively using user interaction and collect user feedback on the explanation report.

Images