Asset allocation model treatment method based on risk quantification and related equipment

Abstract

The invention discloses an asset allocation model governance method and related equipment based on risk quantification, and the method comprises the steps: firstly configuring a model storage library and a model document library for a to-be-governed model, the model storage library centrally storing the whole life cycle information of the model, and the model document library centrally storing related documents; then, carrying out full-life-cycle workflow management based on the model storage library information, and carrying out stage verification by using a preset model verification tool box in stages of model development, verification, deployment and the like; meanwhile, a quantitative risk score is calculated by using a preset model risk assessment toolbox according to the model performance and the business influence index. And finally, calling a preset analysis report template library based on the model storage library, the document library information, and a verification result and a quantitative risk score in full-life-cycle workflow management, and automatically generating a standardized governance report. The objective of the invention is to solve the problems of management fragmentation, risk non-quantification, life cycle management fracture and low treatment efficiency in the prior art.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of financial asset allocation model, and particularly relates to an asset allocation model governance method based on risk quantification and related equipment. BACKGROUND

[0002] With the continuous development of the financial market and the diversification of financial products, financial institutions increasingly rely on complex mathematical models to optimize asset allocation, such as using risk budgeting models for strategic asset allocation and using factor models for tactical asset allocation. These models analyze and predict vast amounts of market data to provide a basis for investment decisions. With the increasing number of models and the increasing complexity of model structures, financial institutions face some management problems when applying models in practice, which has become a key point restricting their risk control capabilities and decision-making efficiency.

[0003] Specifically, (1) In existing technical practices, different business lines or departments often independently develop and maintain their own asset allocation models, resulting in scattered storage of models and their related resources (such as code, parameters, training data, and documents) in different systems or media. This fragmented management mode makes it difficult for financial institutions to grasp the overall picture of their model assets, and the retrieval, reuse, and auditing of model information become difficult, which also increases the operational risk caused by version confusion or dependency loss. (2) Currently, the evaluation of model risk mainly relies on expert experience, qualitative judgment, or semi-quantitative analysis based on a small number of indicators, lacking a standard and comprehensive quantitative system to comprehensively measure the technical performance defects of models and the potential business impact. This makes it difficult for management to accurately determine the risk level of different models and their trends, resulting in a lack of targeted risk prevention measures. At the same time, the verification and monitoring of models are often limited to the initial development or online stage, lacking a full-life-cycle continuous verification mechanism covering model development, validation, deployment, operation, and update to retirement, which cannot timely capture the performance degradation problems of models in the production environment caused by data drift and concept drift. (3) The governance activities such as compliance check, document maintenance, and report generation of models highly depend on manual operation, with inconsistent and inefficient processes. Especially when providing model governance reports to internal management or external regulatory agencies, it is often necessary to manually collect and organize data from multiple isolated sources on the spot, which not only increases the workload and is prone to errors, but also makes it difficult to ensure the consistency and comparability of report content and format, and cannot meet the requirements of internal governance and external compliance. In summary, the existing technology has problems such as management fragmentation, unquantifiable risk, broken life cycle management, and low governance efficiency in the governance of asset allocation models. SUMMARY

[0004] In view of the problems in the prior art, the application provides a risk quantification-based asset allocation model management method and related equipment, and aims to solve the problems of management fragmentation, unquantifiable risk, broken life cycle management and low management efficiency in the prior art.

[0005] To solve the above technical problems, the application is implemented by the following technical solutions: According to a first aspect of the application, a risk quantification-based asset allocation model management method is provided, comprising the following steps: An asset allocation model to be managed is configured with a model repository and a model document library, wherein the model repository is used to centrally store full life cycle related information of the asset allocation model, and the model document library is used to centrally store documents related to the asset allocation model; Based on the information stored in the model repository, full life cycle workflow management is performed on the asset allocation model, which at least includes: in the development, verification, deployment, operation, update and retirement stages of the asset allocation model, a preconfigured model verification toolbox is used to perform stage-by-stage verification on the asset allocation model; and a preconfigured model risk assessment toolbox is used to calculate a quantified risk score of the asset allocation model based on performance indicators and business impact indicators of the asset allocation model; Based on the information of the model repository and the model document library, the verification results generated in the full life cycle workflow management process and the quantified risk score, a preconfigured analysis report template library is called to automatically generate a standardized management report for the asset allocation model.

[0006] In a possible implementation manner of the first aspect, the preconfigured model risk assessment toolbox is used to calculate the quantified risk score of the asset allocation model based on the performance indicators and business impact indicators of the asset allocation model, and specifically includes: A model quality score of the asset allocation model is calculated, which is obtained based on performance evaluation indicators of the asset allocation model on a predetermined task; A model impact score of the asset allocation model is calculated, which is obtained by weighted summation of performances of the asset allocation model in terms of business value, technical performance, explainability and usability; The model quality score and the model impact score are multiplied to obtain the quantified risk score.

[0007] In a possible implementation manner of the first aspect, the model impact score is calculated by the following formula:

[0008] wherein, 、 、 、 is a weight coefficient determined by an analytic hierarchy process.

[0009] In a possible implementation of the first aspect, the periodically verifying the asset allocation model by using the preset model verification toolbox comprises accuracy verification in the asset allocation model development stage, wherein the accuracy verification is completed by judging whether an accuracy of the asset allocation model reaches a preset threshold, and the accuracy is calculated based on true positive examples, true negative examples, false positive examples and false negative examples.

[0010] In a possible implementation of the first aspect, the periodically verifying the asset allocation model by using the preset model verification toolbox comprises robustness verification in the asset allocation model development stage, wherein the robustness verification is completed by calculating a proportion of a prediction result of the asset allocation model remaining unchanged under input data with noise, and judging whether the proportion reaches a preset threshold.

[0011] In a possible implementation of the first aspect, the periodically verifying the asset allocation model by using the preset model verification toolbox comprises generalization capability verification in the asset allocation model verification stage, wherein the generalization capability verification is completed by calculating an average loss of the asset allocation model on a verification data set, and judging whether the average loss is lower than a preset threshold.

[0012] In a possible implementation of the first aspect, the standardized governance report comprises at least one of an asset allocation analysis report, an asset allocation target report and a system index report.

[0013] In a possible implementation of the first aspect, the method further comprises: dynamically adjusting a governance strategy or a model parameter of the asset allocation model based on a quantitative risk score and a performance change trend reflected in the standardized governance report.

[0014] According to a second aspect of the present application, a computer device is provided, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the asset allocation model governance method based on risk quantification when executing the computer program.

[0015] According to a third aspect of the present application, a computer readable storage medium is provided, which stores a computer program, wherein the computer program is executable on a processor to implement the asset allocation model governance method based on risk quantification.

[0016] According to a fourth aspect of the present application, there is provided a computer program product which, when executed by a processor, implements the risk quantification-based asset allocation model governance method.

[0017] Compared with the prior art, the present application has at least the following beneficial effects: The risk quantification-based asset allocation model governance method provided by the present application solves the problems of scattered storage and difficult unified control of models and related information by constructing a unified model information management foundation through centralized configuration of a model storage and a model document library for asset allocation models. On this basis, the whole life cycle workflow management covering development, verification, deployment, operation, update and retirement is implemented based on the centrally stored information, and stage-by-stage verification and quantitative risk assessment are performed using a preset toolbox, thereby realizing continuous active monitoring and objective risk measurement of the model state. Finally, a standardized governance report is generated by automatically calling an analysis report template library, and all the aforementioned management, monitoring and assessment results are solidified as structured decision-making basis. The present application systematically changes the model governance from a scattered, passive and qualitative manual management mode to a centralized, active, quantitative and automatic closed-loop management process, thereby solving the problems of fragmented management, unquantifiable risk, broken life cycle management and low governance efficiency in the prior art. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the specific embodiments of the present application, the drawings needed in the specific embodiment description will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0019] Figure 1 A flowchart of the risk quantification-based asset allocation model governance method according to an embodiment of the present application.

[0020] Figure 2 A detailed schematic diagram of the risk quantification-based asset allocation model governance method according to an embodiment of the present application. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions of the present application will be described clearly and completely below with reference to the drawings. Obviously, the described embodiments are some embodiments of the present application, but not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0022] For example, Figure 1 andFigure 2 As shown, the embodiment of the present application provides a risk quantification-based asset allocation model governance method, which specifically comprises the following steps: S1, configuring a model repository and a model document library for the asset allocation model to be governed, wherein the model repository is used to centrally store the full life cycle related information of the asset allocation model, and the model document library is used to centrally store the documents related to the asset allocation model.

[0023] Specifically, the model repository is implemented by a relational database, which centrally stores various types of information of the asset allocation model in the full life cycle, including model basic information, model input and output information, metadata, model accuracy indicators, model stability indicators, model process related data, model compliance and audit information, and model association information, etc. The model document library is built based on a document management system, which is used to centrally store the design specifications, technical specifications, API interface documents, user operation manuals, and historical change records of the asset allocation model. In addition, the model document library also records the change history of the model, including the date, change content, change personnel and change reason of each change, as well as the reason, date and archive information of the model retirement, throughout the various stages and processes in the full life cycle of the model, providing comprehensive information for model governance.

[0024] S2, based on the information stored in the model repository, performing full life cycle workflow management on the asset allocation model, which at least includes: using a pre-configured model verification toolbox to perform stage-by-stage verification on the asset allocation model during the development, verification, deployment, operation, update and retirement stages of the asset allocation model; and using a pre-configured model risk assessment toolbox to calculate the quantitative risk score of the asset allocation model based on the performance indicators and business impact indicators of the asset allocation model.

[0025] That is, based on the information in the model repository, the full life cycle workflow management is implemented on the asset allocation model. The full life cycle workflow management covers the complete process of the asset allocation model from development, verification, deployment, operation, update to retirement. In this process, the system integrates the model verification toolbox and the model risk assessment toolbox.

[0026] Specifically, at each key stage of the model life cycle, the system automatically calls the corresponding tool in the model verification toolbox for periodic verification. This includes but is not limited to: performance and robustness verification in the development stage; generalization ability verification in the verification stage; data drift real-time monitoring in the deployment stage; performance degradation measurement in the running stage; version backtracking verification in the update stage; and replacement effectiveness verification in the decommissioning stage. At the same time, throughout the life cycle, regular interpretive checks such as compliance and fairness are also conducted. All checks are based on pre-set quantitative formulas and thresholds for judgment, and the results are recorded.

[0027] At the same time, the system will periodically call the model risk assessment toolbox, for example, the model risk assessment toolbox is called every month. The model risk assessment toolbox obtains the performance indicators of the model from the model repository, and combines business impact analysis to calculate the model risk score, and obtains a standardized quantitative risk score, for example, 85 points.

[0028] S3, based on the information of the model repository, the model document library, and the verification results generated in the whole life cycle workflow management process and the quantitative risk score, call the pre-set analysis report template library to automatically generate a standardized governance report for the asset allocation model.

[0029] That is, the governance system automatically calls the pre-defined report template in the analysis report template library according to all the information generated in the above process, including the basic information in the model repository, the link of the model document library, the results of the verification at each stage, and the quantitative risk score calculated. The system fills the relevant data and text into the template to automatically generate a standard format governance report. For example, the report content covers model overview, life cycle status at each stage, risk score and interpretation, problems found and suggestions, etc. without manual data collection and preparation.

[0030] Through the method of the embodiment, centralized storage of asset allocation models, full life cycle automatic monitoring and verification, risk quantification and evaluation, and automatic generation of governance reports are realized, effectively solving the problems of model management fragmentation, invisible risk, and low governance efficiency.

[0031] In one implementation manner, the quantitative risk score of the asset allocation model is calculated based on the performance indicators and business impact indicators of the asset allocation model using the pre-set model risk assessment toolbox, specifically including: The model quality score of the asset allocation model is calculated first, which is based on the performance evaluation index of the asset allocation model on a predetermined task. Then the model impact score of the asset allocation model is calculated, which is obtained by weighted summation of the performance of the asset allocation model in business value, technical performance, explainability and usability, specifically:

[0032] wherein, , , , is the weight coefficient determined by the analytic hierarchy process. Finally, the model quality score and the model impact score are multiplied to obtain the quantitative risk score.

[0033] Specifically, the model quality score reflects the technical performance of the model. The standard and formula for evaluating model quality depend on the specific task (such as classification, regression, clustering, etc.). Reference can be made to the formula commonly used for evaluating model quality in common tasks, such as precision calculation, recall rate, F1 Score, AUC-ROC, mean square error, mean absolute error, R-Squared, silhouette coefficient, ARI, etc. These formulas can be flexibly selected according to the specific task for evaluating model quality. For example, the system obtains the confusion matrix data of the model in the test set from the model repository in recent days, selects the commonly used F1 score in classification tasks as the performance evaluation index, calculates the F1 score as 0.82, and takes this value as the model quality score. For other tasks such as regression and clustering, mean square error, silhouette coefficient, etc. can be selected accordingly as the basis for calculation.

[0034] The model impact score comprehensively reflects the importance of the model in the business level. For example, the analytic hierarchy process (AHP) is used to compare and evaluate the multiple aspects of model impact, and after matrix consistency test, four aspects and weight coefficients are determined, specifically: business value corresponds to =0.4, technical performance corresponds to =0.3, explainability corresponds to =0.2, and usability / fairness corresponds to =0.1. Then the experts score the performance of the model in the above four aspects on a 100-point scale: business value 90 points, technical performance 80 points, explainability 70 points, and usability / fairness 85 points.

[0035] According to the formula, the model impact score = 0.4 × 90 + 0.3 × 80 + 0.2 × 70 + 0.1 × 85 = 82.5. Multiplying the two scores together: 0.82 × 82.5 = 67.65. This score intuitively quantifies the overall risk level of the model after comprehensively considering technical performance and business impact. A lower score indicates that key attention and intervention are needed.

[0036] In one possible implementation, the asset allocation model is periodically validated using a pre-built model validation toolkit, including an accuracy validation performed during the asset allocation model development phase. The accuracy validation is performed by determining whether the accuracy of the asset allocation model reaches a preset threshold, and the accuracy is calculated based on true positives, true negatives, false positives, and false negatives.

[0037] Specifically, after the model development is completed, accuracy verification needs to be performed using the reserved test dataset. The system calls the accuracy verification module in the model verification toolbox. The accuracy verification module runs the model to predict on the test set and compares it with the true labels, counting the number of true positives (TP), true negatives (TN), false positives (FP), and false negatives (FN).

[0038] Subsequently, the accuracy verification module uses the formula... Calculate the accuracy of the model. Assume the calculated accuracy is 94.5%. The system's pre-set accuracy is determined by a threshold. The accuracy rate is 92%. The accuracy verification module executes the judgment condition: If the accuracy rate is 94.5% ≥ 92%, the condition is met, the verification passes, and the result is recorded in the accuracy verification item of the model development stage in the model repository, with a status of "passed". If the accuracy rate does not reach the threshold, the verification fails, the system will issue an alarm, prevent the model from entering the next stage, and prompt the developers to optimize the model.

[0039] In one possible implementation, the use of a pre-built model validation toolkit to perform phased validation of the asset allocation model includes robustness validation during the asset allocation model development phase. The robustness validation is performed by calculating the proportion of the asset allocation model's prediction results that remain unchanged under noisy input data and determining whether the proportion reaches a preset threshold.

[0040] Specifically, after the model has completed training, the robustness verification module in the verification toolbox is activated. The robustness verification module processes the input data for each sample in the test set. Artificially adding a tiny bit of random noise Generate noisy samples Then, the model was used to test the original samples. and noisy samples Make predictions.

[0041] Robustness is calculated by: . Where, is the total number of test samples, is an indicator function that takes value 1 when the condition in the bracket is true (i.e. the predicted results are the same), otherwise 0. This formula calculates the proportion of samples whose prediction results remain unchanged under noise interference.

[0042] For example, suppose that after adding noise to 1000 test samples, the prediction results of 950 samples do not change, then = 950 / 1000 = 0.95. The system preset robustness threshold is 0.90. The robustness check module performs the judgment: 2, i.e. 0.95 ≥ 0.90. The condition is true, and the robustness check passes. It ensures that the model is not sensitive to small perturbations of input data, and enhances its stability in the actual volatile market environment.

[0043] In one implementation, the periodic checking of the asset allocation model by the preset model checking toolbox includes a generalization ability check performed in the asset allocation model verification stage, wherein the generalization ability check is completed by calculating the average loss of the asset allocation model on the verification data set and judging whether the average loss is lower than a preset threshold.

[0044] Specifically, the system calls the generalization ability check module in the checking toolbox. The generalization ability check module uses a verification data set that is the same distribution as the training data but completely independent. For each sample in the verification set, the generalization ability check module records the predicted value of the model and the true label value .

[0045] The generalization error is measured by calculating the average loss of the model on the verification set: . Where, is the loss function.

[0046] For example, suppose that the calculated generalization error on the verification set is 0.05. The system preset generalization ability threshold is 0.08. The generalization ability check module performs the judgment: 3, i.e. 0.05 ≤ 0.08. The condition is true, and the generalization ability check passes. It indicates that the model has good prediction ability for unseen data and has low risk of overfitting, and can enter the deployment stage.

[0047] Real-time monitoring and data drift detection is performed during the deployment phase, i.e., after the model goes online, the system continuously monitors the distribution changes of the input data. Kullback-Leibler divergence is used to measure the difference between the training data distribution and the real-time production data distribution . .

[0048] For example, assuming the calculated KL divergence is 0.02, and the preset drift threshold is 0.05. The verification condition is: 4, i.e., 0.02 ≤ 0.05, the condition is true, and no significant data drift is detected.

[0049] Performance degradation measurement is performed during the running phase. During the model's running period, its key performance indicators such as AUC, F1, etc. are calculated regularly and compared with the initial performance. The degree of performance decline is: . Assuming the initial performance is 0.90 and the current performance is 0.87, then: = (0.90 - 0.87) / 0.90 ≈ 0.033. The preset degradation threshold is 0.05. The verification condition is: , i.e., 0.033 ≤ 0.05, the condition is true, and the performance degradation is within an acceptable range.

[0050] Version rollback verification is performed during the update phase. When the model is upgraded to a new version, it is necessary to ensure that the performance of the new version is not worse than that of the old version. The performance change is: . Assuming the performance of the updated model is 0.91 and the performance before updating is 0.89, then = (0.91 - 0.89) / 0.91 ≈ 0.022. The verification condition is: , i.e., 0.022 ≥ 0, the condition is true, and the version upgrade is effective.

[0051] When model A is replaced by model B, the system's overall performance should not be affected. The verification condition is: . By comparing the volatility of the portfolio constructed by the investment recommendations output by the model before and after the replacement within a week, it is confirmed that the volatility under the new model is flat or lower, and the verification is passed.

[0052] In addition, compliance and fairness verification is performed, and the prediction fairness of the model for different attribute groups is checked regularly. The fairness indicator is: . Assuming the F1 of group 1 is 0.85 and the F1 of group 2 is 0.83, then ≈ 1.024. The requirement As close to 1 as possible, i.e. the performance difference between different test groups should be as small as possible to ensure fairness, the preset threshold is [0.95, 1.05]. 1.024 is in this interval, and the fairness check passes.

[0053] In an implementation manner, the standardized governance report includes at least one of an asset allocation analysis report, an asset allocation target report, and a system index report.

[0054] Specifically, all the above check results, risk scores, and model information are summarized into an analysis report template library to automatically generate the asset allocation analysis report, the asset allocation target report, and the system index report.

[0055] The asset allocation analysis report focuses on analyzing the asset allocation field and deeply analyzes the composition, distribution, and related performance of the asset portfolio. It covers detailed analysis of various assets such as stocks, bonds, and funds, including the performance characteristics and mutual relationships of assets in different market environments, the rationality and effectiveness of allocation strategies, and the evaluation of the managed model according to the governance method of the present application. The report also analyzes the strategic assets, tactical assets, and portfolio adjustment, providing clear decision-making basis for investors. At the same time, the report also focuses on the influence of macroeconomic situation, industry dynamics, and policy changes on asset allocation, and predicts future market trends.

[0056] The asset allocation target report guides and optimizes the construction of the investment portfolio. By analyzing market trends, risk assessment, asset liquidity, and portfolio volatility, it helps investors set clear investment goals. The report contains macro analysis, model observability improvement, model effectiveness improvement, etc. to ensure accurate investment decisions and asset preservation and appreciation. At the same time, it also involves the model governance process to ensure compliance, robustness, and generalization ability of the investment model, and measures the governance results through risk control and performance check. Finally, the report provides optimization suggestions for strategic and tactical asset allocation to achieve the balance between asset appreciation and risk control.

[0057] The system index report is an important tool for measuring and monitoring model performance. Through a series of key indicators such as accuracy, robustness, generalization ability, and service ability, it evaluates the model's performance in actual operation. The report covers model stability, effectiveness check, performance degradation, and blood analysis, etc. to ensure that the model remains efficient and reliable in a changing market environment. In addition, the system index report also includes real-time monitoring and feedback mechanism of the model, as well as model version control and update to cope with performance degradation and market changes. Through the comprehensive analysis of these indicators, it helps governance managers and business managers make data-based decisions to improve the overall governance capability of the model.

[0058] The analysis report template library can greatly improve work efficiency, effectively reduce the occurrence of errors, and improve report quality. For financial institutions, it can improve the quality and efficiency of model governance report work, and also help promote regulatory compliance.

[0059] In an implementable manner, the method further comprises: dynamically adjusting the governance strategy or model parameter of the asset allocation model based on the quantitative risk score and performance change trend reflected in the standardized governance report.

[0060] For example, assuming that a macro factor tactical asset allocation model shows that its data drift index continues to rise close to the threshold and the quantitative risk score decreases from 75 to 68 in the system indicator report of the last two consecutive quarters. The governance system triggers the dynamic adjustment mechanism. The system automatically issues a warning, suggesting that the model may have decreased adaptability due to market style changes. The system can perform one or more of the following dynamic adjustments based on pre-set rules or management instructions: Automatically increase the monitoring frequency of the model from quarterly verification to monthly verification; at the same time, automatically adjust the risk level of the model from low risk attention to medium risk attention in the governance board, and trigger the process to require the model owner to submit a special analysis report.

[0061] Under the premise of obtaining authorization and ensuring secure isolation, the system can automatically start a model adaptive fine-tuning process. Using recent market data, the system automatically fine-tunes some hyperparameters of the model, such as learning rate and regularization coefficient, in a small range, and performs backtesting and verification in a simulation environment. If the model with the new parameters passes all the verifications and the risk score is improved, a report is generated to apply for replacing the old version of the model.

[0062] Through this embodiment, model governance is no longer static and rigid, but can be adaptively adjusted according to the actual running performance of the model, the change trend of the quantitative risk score, and the market environment.

[0063] In an implementable manner, the full life cycle related information stored in the model repository includes at least one of model basic information, model input and output information, metadata, model accuracy indicators, model stability indicators, model process related data, model compliance and audit information, and model association information.

[0064] This embodiment details the specific composition of the "full life cycle related information" stored in the model repository in embodiment 1. Taking a multi-factor stock selection model under governance as an example, the model repository of the model stores the following eight types of information in a structured manner, forming a complete model digital archive: Regarding the full life cycle related information, this embodiment explains as follows: The model basic information includes model name and version, model source, model purpose, model type, model structure, model configuration, and model size.

[0065] The model input and output information includes input feature variables, weight, parameter information, and output variables.

[0066] The metadata includes dataset information involved in the process from training to testing and deploying the model.

[0067] The model accuracy indicators include accuracy, precision, recall, and F1 score.

[0068] The model stability indicators include variance and robustness test information.

[0069] The model process-related data includes training data, metadata information, training algorithm, and computing resources.

[0070] The model compliance and audit information includes compliance awareness, audit history, and results.

[0071] The model correlation information includes model dependency, environment information, model users, and permission information.

[0072] The present application fundamentally improves the management level and risk control ability of financial institutions for asset allocation models. The method provides standardized management processes and centralized storage and documentation systems, solving the problems of scattered model management and lack of unified view, improving management efficiency and standardization. On this basis, by covering the whole life cycle workflow management of model development, verification, deployment, operation, update and retirement, and integrating model checking and risk assessment toolboxes, the model state is continuously monitored and quantitatively evaluated. The quantified model risk score can objectively reveal the technical risk and business impact of the model. Further, the embedded explainability requirement and automatic reporting mechanism enhance the transparency and credibility of the model, meeting the compliance requirements. The method has dynamic adjustment capability, which can optimize the governance strategy and the model itself according to market changes and risk assessment results, thereby closely integrating risk management and business decision-making.

[0073] In another embodiment of the present application, a computer device is provided, which comprises a processor and a memory, the memory is configured to store a computer program, the computer program comprises program instructions, and the processor is configured to execute the program instructions stored in the computer storage medium. The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, and are suitable for implementing one or more instructions, and are particularly suitable for loading and executing one or more instructions in the computer storage medium to implement a corresponding method process or a corresponding function; the processor in the embodiments of the present application can be used for the operation of the asset allocation model management method based on risk quantification.

[0074] In another embodiment of the present application, the present application further provides a storage medium, specifically a computer readable storage medium (Memory), which is a memory device in a computer device, and is configured to store programs and data. It can be understood that the computer readable storage medium herein can include an internal storage medium in the computer device, and of course can also include an expansion storage medium supported by the computer device. The computer readable storage medium provides a storage space, and the storage space stores an operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and the instructions can be one or more computer programs (including program codes). It should be noted that the computer readable storage medium herein can be a random access memory (RAM), and can also be a non-volatile memory, such as at least one disk memory. One or more instructions stored in the computer readable storage medium can be loaded and executed by the processor to implement the corresponding steps of the asset allocation model management method based on risk quantification in the above embodiments.

[0075] Those skilled in the art will appreciate that embodiments of the present application can be readily used as a method, a system or a computer program product. Accordingly, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage etc.) embodying computer readable program code.

[0076] The present application is described in reference to the flowchart and / or block diagrams of the method, apparatus (system) and computer program product according to embodiments of the application. It will be understood that each block of the flowchart and / or block diagrams, and combinations of blocks in the flowchart and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processor or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 means for carrying out each of the

[0077] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 means for carrying out each of the

[0078] The computer program instructions can also be loaded into a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the function specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 means for carrying out each of the

[0079] The present application also provides a computer program product, which is used for executing the risk-quantitative-based asset allocation model management method described above. Since the computer program product provided by the present application belongs to the same inventive concept as the risk-quantitative-based asset allocation model management method described above, the computer program product provided by the present application has all the advantages of the risk-quantitative-based asset allocation model management method described above, and thus the beneficial effects of the computer program product provided by the present application will not be described here.

[0080] In the present application, the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, different embodiments or examples described in the present specification and the features of different embodiments or examples can be combined and combined by those skilled in the art without contradiction.

[0081] Finally, it should be noted that the above-described embodiments are only specific embodiments of the present application, which are used to illustrate the technical solutions of the present application, but not to limit the present application. The protection scope of the present application is not limited thereto. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily think of changes to the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some technical features thereof, within the technical scope disclosed by the present application. Such modifications, changes or replacements do not cause the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be covered within the protection scope of the present application.

Claims

1. A governance method for an asset allocation model based on risk quantification, characterized in that, include: Configure a model repository and a model document repository for the asset allocation model to be governed, wherein the model repository is used to centrally store information related to the entire lifecycle of the asset allocation model, and the model document repository is used to centrally store documents related to the asset allocation model; Based on the information stored in the model repository, the asset configuration model is managed through a full lifecycle workflow. The full lifecycle workflow management includes at least the following: during the development, verification, deployment, operation, update, and decommissioning stages of the asset configuration model, the asset configuration model is periodically verified using a pre-built model verification toolbox; and, using a pre-built model risk assessment toolbox, a quantitative risk score of the asset configuration model is calculated based on the performance indicators and business impact indicators of the asset configuration model. Based on the information from the model repository and model document repository, as well as the verification results and quantitative risk scores generated during the full lifecycle workflow management process, a standardized governance report for the asset allocation model is automatically generated by calling a pre-set analysis report template library.

2. The asset allocation model governance method based on risk quantification according to claim 1, characterized in that, The process involves using a pre-built model risk assessment toolkit to calculate a quantitative risk score for the asset allocation model based on its performance indicators and business impact indicators. This includes: Calculate the model quality score of the asset allocation model, which is obtained based on the performance evaluation index of the asset allocation model on the predetermined task; The model impact score of the asset allocation model is calculated by weighting and summing the performance of the asset allocation model in terms of business value, technical performance, interpretability, and usability. The quantitative risk score is obtained by multiplying the model quality score by the model influence score.

3. The asset allocation model governance method based on risk quantification according to claim 2, characterized in that, The model influence score is calculated using the following formula: in, , , , These are the weight coefficients determined using the analytic hierarchy process (AHP).

4. The asset allocation model governance method based on risk quantification according to claim 1, characterized in that, The provision of a pre-built model validation toolkit for periodic validation of the asset allocation model includes accuracy validation during the development phase of the asset allocation model. The accuracy validation is performed by determining whether the accuracy of the asset allocation model reaches a preset threshold. The accuracy is calculated based on true positives, true negatives, false positives, and false negatives.

5. The asset allocation model governance method based on risk quantification according to claim 1, characterized in that, The provision of a pre-built model validation toolkit for periodic validation of the asset allocation model includes robustness validation during the asset allocation model development phase. The robustness validation is performed by calculating the proportion of the asset allocation model's prediction results that remain unchanged under noisy input data and determining whether the proportion reaches a preset threshold.

6. The asset allocation model governance method based on risk quantification according to claim 1, characterized in that, The provision of a pre-built model validation toolkit for periodic validation of the asset allocation model includes a generalization capability validation performed during the asset allocation model validation phase. The generalization capability validation is performed by calculating the average loss of the asset allocation model on the validation dataset and determining whether the average loss is lower than a preset threshold.

7. The asset allocation model governance method based on risk quantification according to claim 1, characterized in that, The standardized governance report includes at least one of the following: an asset allocation analysis report, an asset allocation target report, and a system indicator report.

8. The asset allocation model governance method based on risk quantification according to claim 1, characterized in that, The method further includes: Based on the quantitative risk score and performance change trend reflected in the standardized governance report, the governance strategy or model parameters of the asset allocation model are dynamically adjusted.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements a risk quantification-based asset allocation model governance method as described in any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements a risk quantification-based asset allocation model governance method as described in any one of claims 1 to 8.

Images