Automatic evaluation system for enterprise risk management

Abstract

The invention provides an automatic evaluation system for enterprise risk management, and relates to the technical field of automatic evaluation, and the system carries out the desensitization of obtained enterprise evaluation data, and obtains desensitization evaluation data; constructing a behavior feature difference matrix, determining the behavior feature anomaly degree of the target enterprise in the interaction process according to the behavior feature difference matrix, and determining the behavior characterization of each interaction behavior in the target enterprise based on the behavior feature anomaly degree and the behavior feature difference matrix; determining the behavior diversity of each interaction behavior in the target enterprise according to the evaluation data training set, and updating the evaluation data training set into an interaction behavior training set according to the behavior characterization and the behavior diversity of each interaction behavior; according to the technical scheme of the invention, the representativeness and diversity of the training data can be improved when the automatic evaluation model is trained, so that the evaluation capability of the automatic evaluation model is enhanced.

Description

Technical Field

[0001] The present application relates to the technical field of automated assessment, and more specifically, to an automated assessment system for enterprise risk management. Background Art

[0002] The automated assessment system for enterprise risk management is a comprehensive management tool that combines big data analysis, big models and artificial intelligence technologies, and is designed to monitor, analyze and predict the potential risks faced by enterprises in real time from multiple dimensions. The core function of this system is to collect and process multi-source data, use big model algorithms to generate comprehensive assessment results of corporate credit risk, operational risk, market risk and compliance risk, and provide enterprises with quantitative risk scores and response suggestions.

[0003] In an automated evaluation system, an automated evaluation model is usually used to evaluate the target enterprise. The automated evaluation model can automatically learn the complex patterns and features in the data through training with massive data, without the need for a large amount of manual feature engineering. In the prior art, the automated evaluation model requires a large amount of high-quality data for training, but the behavioral data used for evaluation in the enterprise often has an imbalance problem, that is, the samples of abnormal behavior are far less than normal behavior. This imbalance will cause the automated evaluation model to favor normal behavior during the training process, reducing the ability to classify abnormal behavior. In addition, the interactive behavior patterns in different regions and industries may be significantly different, and it is necessary to build a diverse data set for training. Therefore, how to improve the representativeness and diversity of training data when training the automated evaluation model to enhance the evaluation ability of the automated evaluation model has become a difficult problem faced by the industry. Summary of the invention

[0004] The present application provides an automated assessment system for enterprise risk management, which can improve the representativeness and diversity of training data when training an automated assessment model to enhance the assessment capability of the automated assessment model.

[0005] The present application provides an automated assessment system for enterprise risk management, the assessment system comprising: An evaluation data acquisition module is used to obtain the enterprise evaluation data of the target enterprise in the risk management process, desensitize the enterprise evaluation data, and then obtain the desensitized evaluation data of the target enterprise; A representativeness determination module, used to construct a behavior characteristic difference matrix through the interactive behavior data in the desensitized evaluation data, determine the abnormality of the behavior characteristics of the target enterprise in the interactive process according to the behavior characteristic difference matrix, and determine the behavior representativeness of each interactive behavior in the target enterprise based on the behavior characteristic abnormality and the behavior characteristic difference matrix; A training set updating module is used to obtain a given evaluation data training set, determine the behavior diversity of each interactive behavior in the target enterprise based on the evaluation data training set, and update the evaluation data training set into an interactive behavior training set according to the behavior representativeness and behavior diversity of each interactive behavior; The automatic evaluation module is used to train the automatic evaluation model based on random forest using the interactive behavior training set, and generate an evaluation result for the target enterprise according to the trained automatic evaluation model.

[0006] In this embodiment, enterprise evaluation data of the target enterprise during the risk management process is obtained from the management database of the target enterprise.

[0007] In this embodiment, desensitizing the enterprise evaluation data is to encrypt the sensitive information in the enterprise evaluation data.

[0008] 4. The automated assessment system for enterprise risk management according to claim 1, wherein constructing a behavior feature difference matrix using the interactive behavior data in the desensitized assessment data specifically comprises: Obtaining interactive behavior data in the desensitization evaluation data; Determining the difference in interaction characteristics between each interaction behavior in the interaction behavior data; The behavior feature difference matrix is ​​constructed through the interaction feature differences between each interaction behavior.

[0009] In this embodiment, determining the abnormality of the behavior characteristics of the target enterprise during the interaction process according to the behavior characteristic difference matrix specifically includes: Extracting features from the behavioral feature difference matrix to obtain the interaction representation of the target enterprise during the interaction process; The abnormality of the behavior characteristics of the target enterprise during the interaction process is determined according to the interaction characterization quantity and the preset adjustment coefficient.

[0010] In this embodiment, determining the behavior representativeness of each interactive behavior in the target enterprise based on the behavior characteristic abnormality and the behavior characteristic difference matrix specifically includes: For each interactive behavior in the target enterprise, extract all interactive feature differences corresponding to the interactive behavior in the behavioral feature difference matrix; The behavioral representativeness of the interactive behavior is determined by all the interactive feature differences corresponding to the interactive behavior and the abnormality of the behavioral feature, and then the behavioral representativeness of each interactive behavior in the target enterprise is obtained.

[0011] In this embodiment, determining the behavior diversity of each interactive behavior in the target enterprise based on the evaluation data training set specifically includes: For each interactive behavior in the target enterprise, the diversity of the interactive behavior is measured according to the evaluation data training set to obtain the behavioral diversity of the interactive behavior, and then the behavioral diversity of each interactive behavior in the target enterprise is obtained.

[0012] In this embodiment, updating the evaluation data training set into an interactive behavior training set according to the behavioral representation and behavioral diversity of each interactive behavior specifically includes: Get the set adjustment parameters; For each interactive behavior, determining a characteristic contribution coefficient of the interactive behavior according to the behavior representation and behavior diversity corresponding to the interactive behavior and the adjustment parameter, thereby obtaining the characteristic contribution coefficient of each interactive behavior; The evaluation data training set is updated using the feature contribution coefficients of all interactive behaviors to obtain an interactive behavior training set.

[0013] In this embodiment, the evaluation results of the target enterprise generated according to the trained automatic evaluation model specifically include: Preprocess the target enterprise's desensitized assessment data to obtain preprocessed desensitized assessment data; The pre-processed desensitized evaluation data is input into the trained automated evaluation model for evaluation, thereby obtaining the evaluation result of the target enterprise.

[0014] In this embodiment, the evaluation result includes the comprehensive risk score of the target enterprise and the interactive risk scores of each dimension.

[0015] The technical solution provided by the embodiments disclosed in this application has the following beneficial effects: The enterprise evaluation data of the target enterprise in the risk management process is obtained through the evaluation data acquisition module, and the enterprise evaluation data is desensitized to obtain the desensitized evaluation data of the target enterprise; the representativeness determination module constructs a behavior feature difference matrix through the interactive behavior data in the desensitized evaluation data, determines the abnormality of the behavioral characteristics of the target enterprise in the interactive process according to the behavior feature difference matrix, and determines the behavioral representativeness of each interactive behavior in the target enterprise based on the behavioral feature abnormality and the behavior feature difference matrix; the training set update module obtains a given evaluation data training set, determines the behavioral diversity of each interactive behavior in the target enterprise according to the evaluation data training set, and updates the evaluation data training set to an interactive behavior training set according to the behavioral representativeness and behavioral diversity of each interactive behavior; the automatic evaluation module uses the interactive behavior training set to train an automated evaluation model based on random forests, and generates an evaluation result for the target enterprise according to the trained automated evaluation model.

[0016] It can be seen that in this application, firstly, a behavior feature difference matrix is ​​constructed to convert complex interactive behavior differences into a structured matrix form, which helps to more clearly reveal the correlation and pattern between interactive behaviors. Based on the behavior feature difference matrix, the calculation of the abnormality of the behavior features can be made more accurate, and the behavior representativeness quantifies the representation ability of the interactive behavior in the global and local contexts, which can greatly improve the representativeness of the data in the training of the automated evaluation model; then, by measuring the behavior diversity, the automated evaluation model can cover the behavioral characteristics of different dimensions in the target enterprise, avoiding the insufficient generalization ability of the automated evaluation model due to the single data, and by combining the behavior representativeness and behavior diversity to determine the interactive behavior training set, the automated evaluation model can improve its evaluation ability for complex interactive processes; finally, through a diversified and representative interactive behavior training set, the automated evaluation model can more accurately identify and evaluate the risks of the enterprise, have stronger adaptability and robustness, and improve its generalization performance.

[0017] In summary, the technical solution adopted in this application can improve the representativeness and diversity of training data when training the automated evaluation model, so as to enhance the evaluation capability of the automated evaluation model. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0019] Figure 1 It is a module structure diagram of an automated assessment system for enterprise risk management provided in accordance with the present application; Figure 2 It is a flowchart of determining the abnormality of the behavioral characteristics of the target enterprise during the interaction process provided by this application; Figure 3 It is a schematic diagram of the process of determining the interactive behavior training set provided by this application. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0021] The embodiment of the present application provides an automated assessment system for enterprise risk management, the core of which is to obtain the enterprise assessment data of the target enterprise in the risk management process through the assessment data acquisition module, desensitize the enterprise assessment data, and then obtain the desensitized assessment data of the target enterprise; the representativeness determination module constructs a behavior feature difference matrix through the interactive behavior data in the desensitized assessment data, determines the abnormality of the behavior features of the target enterprise in the interactive process according to the behavior feature difference matrix, and determines the behavioral representativeness of each interactive behavior in the target enterprise based on the abnormality of the behavior features and the behavior feature difference matrix; the training set update module obtains a given assessment data training set, determines the behavioral diversity of each interactive behavior in the target enterprise according to the assessment data training set, and updates the assessment data training set to an interactive behavior training set according to the behavioral representativeness and behavioral diversity of each interactive behavior; the automatic assessment module uses the interactive behavior training set to train the automated assessment model based on random forests, and generates an assessment result for the target enterprise according to the trained automated assessment model. The present application adopts the above technical solution to improve the representativeness and diversity of the training data when training the automated assessment model, so as to enhance the assessment ability of the automated assessment model.

[0022] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods. Figure 1 As shown, this figure is a module structure diagram of an automated assessment system for enterprise risk management according to this embodiment of the present application. The assessment system includes: an assessment data acquisition module 100, a representativeness determination module 200, a training set update module 300 and an automatic assessment module 400, which are respectively described as follows: The evaluation data acquisition module 100 is used to acquire the enterprise evaluation data of the target enterprise in the risk management process, perform desensitization processing on the enterprise evaluation data, and then obtain the desensitized evaluation data of the target enterprise.

[0023] In specific implementation, the enterprise assessment data of the target enterprise in the risk management process can be obtained in the management database of the target enterprise; it should be noted that in the present application, the enterprise assessment data includes the interactive behavior data, financial data, operational data, market data and compliance data of the target enterprise, and the enterprise assessment data is used for automated assessment in enterprise risk management, wherein the interactive behavior data includes multiple interactive behaviors and interactive tables corresponding to the interactive behaviors.

[0024] In this embodiment, desensitizing the enterprise evaluation data is to encrypt the sensitive information in the enterprise evaluation data, and then obtain the desensitized evaluation data of the target enterprise; in specific implementation, first, according to the content of the enterprise evaluation data, it is clarified which information belongs to sensitive data and needs to be encrypted, for example, enterprise identification information (company name, registration number, taxpayer identification number, etc.), financial data (specific transaction amount, account balance, credit score, etc.) and personnel information (privacy information involving key persons in charge of the enterprise, number of employees, etc.); then, symmetric encryption can be used to encrypt the sensitive information in the enterprise evaluation data to obtain the desensitized evaluation data of the target enterprise; it should be noted that in financial data processing, symmetric encryption is often used for encryption of large-scale data because of its fast encryption and decryption speed.

[0025] The representativeness determination module 200 is used to construct a behavior characteristic difference matrix through the interactive behavior data in the desensitized evaluation data, determine the abnormality of the behavior characteristics of the target enterprise during the interaction process according to the behavior characteristic difference matrix, and determine the behavioral representativeness of each interactive behavior in the target enterprise based on the behavior characteristic abnormality and the behavior characteristic difference matrix.

[0026] In this embodiment, the behavior feature difference matrix is ​​constructed by the interactive behavior data in the desensitization evaluation data in the following manner, namely: Obtaining interactive behavior data in the desensitization evaluation data; Determining the difference in interaction characteristics between each interaction behavior in the interaction behavior data; The behavior feature difference matrix is ​​constructed through the interaction feature differences between each interaction behavior.

[0027] In specific implementation, first, the interactive behavior data in the desensitization evaluation data can be obtained; then, the interactive feature differences between each interactive behavior in the interactive behavior data can be determined, that is, for each interactive behavior in the interactive behavior data, the feature distance between the interactive behavior and other interactive behaviors can be calculated. The feature distance can be represented by the Euclidean distance, which is not limited here, so that the feature distance is used as the interactive feature difference between the interactive behavior and other interactive behaviors. The interactive feature difference represents the degree of feature difference between the interactive behaviors. The interactive feature difference between each interactive behavior in the interactive behavior data can be obtained in the above manner; finally, a behavioral feature difference matrix is ​​constructed through the interactive feature differences between each interactive behavior, that is, the interactive feature differences between each interactive behavior are arranged according to the corresponding relationship, and then a matrix is ​​obtained, and the matrix is ​​used as the behavioral feature difference matrix.

[0028] Preferably, in this embodiment, the abnormality of the behavior characteristics of the target enterprise during the interaction process is determined according to the behavior characteristic difference matrix, referring to Figure 2 The figure is a schematic diagram of a process for determining the abnormality of the behavior characteristics of the target enterprise during the interaction process in some embodiments of the present application. In this embodiment, determining the abnormality of the behavior characteristics of the target enterprise during the interaction process can be achieved by using the following steps: First, in step S21, feature extraction is performed on the behavior feature difference matrix to obtain the interaction representation of the target enterprise during the interaction process; Then, in step S22, the abnormality of the behavior characteristics of the target enterprise during the interaction process is determined according to the interaction characterization amount and the preset adjustment coefficient.

[0029] In the specific implementation, first, feature extraction can be performed on the behavior feature difference matrix to obtain the interaction characterization amount of the target enterprise in the interaction process, that is, the mean of all interaction feature differences in the behavior feature difference matrix is ​​used as the interaction characterization amount, and the interaction characterization amount is a quantity used to characterize the overall difference level of the interaction behavior. In other embodiments, the median of all interaction feature differences can also be used as the interaction characterization amount, which will not be repeated here; finally, the abnormality of the behavior characteristics of the target enterprise in the interaction process is determined according to the interaction characterization amount and the preset adjustment coefficient, and the square root of the product of the fraud feature characterization amount and the preset adjustment coefficient can be taken, so that the result is used as the abnormality of the behavior characteristics of the target enterprise in the interaction process, and the abnormality of the behavior characteristics represents the overall abnormality degree of the target enterprise in the interaction process.

[0030] In this embodiment, the behavior representativeness of each interactive behavior in the target enterprise can be determined based on the behavior characteristic abnormality and the behavior characteristic difference matrix in the following manner, namely: For each interactive behavior in the target enterprise, extract all interactive feature differences corresponding to the interactive behavior in the behavioral feature difference matrix; The behavioral representativeness of the interactive behavior is determined by all the interactive feature differences corresponding to the interactive behavior and the abnormality of the behavioral feature, and then the behavioral representativeness of each interactive behavior in the target enterprise is obtained.

[0031] In specific implementation, first, for each interactive behavior in the target enterprise, all interactive feature differences corresponding to the interactive behavior are extracted from the behavioral feature difference matrix; then, the behavioral representativeness of the interactive behavior can be determined by all interactive feature differences corresponding to the interactive behavior and the behavioral feature abnormality. The behavioral representativeness indicates the ability of the corresponding interactive behavior to represent other interactive behaviors in the target enterprise. In actual implementation, the behavioral representativeness can be determined by the following formula: in, represents the behavioral representation of interaction behavior a, n represents the total number of interaction feature differences corresponding to interaction behavior a, represents the difference of the ith interaction feature corresponding to the interaction behavior a, It represents the abnormality of behavioral characteristics. Through the above method, the behavioral representativeness of each interactive behavior in the target enterprise can be obtained.

[0032] It should be noted that constructing a behavioral feature difference matrix converts complex interactive behavior differences into a structured matrix form, which helps to more clearly reveal the correlation and patterns between interactive behaviors, especially the clustering effect of abnormal behaviors. Based on the behavioral feature difference matrix, the calculation of behavioral feature abnormality can be made more accurate, and the behavioral representativeness quantifies the representation ability of interactive behaviors in global and local contexts, which can provide more comprehensive and accurate analysis for enterprise risk management, and can also greatly improve the representativeness of data in automated assessment model training.

[0033] The training set updating module 300 is used to obtain a given evaluation data training set, determine the behavior diversity of each interactive behavior in the target enterprise based on the evaluation data training set, and update the evaluation data training set to an interactive behavior training set according to the behavior representativeness and behavior diversity of each interactive behavior.

[0034] In specific implementation, a given evaluation data training set can be obtained, which contains a large amount of evaluation data for training an automated evaluation model. The evaluation data is divided into labeled data and unlabeled data. The labeled data is samples that have been labeled as abnormal interaction behaviors or non-abnormal interaction behaviors, and the unlabeled data is unlabeled samples, which may contain unidentified abnormal interaction behaviors. In the automated evaluation of target enterprise risk management, a large amount of unlabeled data and a small amount of labeled data can be used to improve the performance of automated evaluation.

[0035] In this embodiment, the behavior diversity of each interactive behavior in the target enterprise can be determined based on the evaluation data training set in the following manner, namely: For each interactive behavior in the target enterprise, the diversity of the interactive behavior is measured according to the evaluation data training set to obtain the behavioral diversity of the interactive behavior, and then the behavioral diversity of each interactive behavior in the target enterprise is obtained.

[0036] In specific implementation, for each interactive behavior in the target enterprise, the diversity of the interactive behavior can be measured according to the evaluation data training set to obtain the behavioral diversity of the interactive behavior. The behavioral diversity indicates the degree of sample difference between the corresponding interactive behavior and the evaluation data in the evaluation data training set. If the behavioral diversity is higher, the sample difference is higher, indicating that if the interactive behavior is added to the evaluation data training set, the sample diversity of the evaluation data training set will be improved, which is helpful to enhance the evaluation ability of the automated evaluation model. In actual implementation, the behavioral diversity can be determined according to the following formula: in, represents the behavioral diversity of interactive behavior a, a represents interactive behavior, M represents the evaluation data training set, represents the j-th evaluation data in the evaluation data training set M. The behavior diversity of each interactive behavior in the target enterprise can be obtained through the above method.

[0037] Preferably, in this embodiment, the evaluation data training set is updated to an interactive behavior training set according to the behavioral representation and behavioral diversity of each interactive behavior, referring to Figure 3 The figure is a schematic diagram of a process for determining an interactive behavior training set in some embodiments of the present application. In this embodiment, the determination can be implemented by the following steps: First, in step S31, the set adjustment parameters are obtained; Then, in step S32, for each interactive behavior, the characteristic contribution coefficient of the interactive behavior is determined according to the behavior representation and behavior diversity corresponding to the interactive behavior and the adjustment parameter, thereby obtaining the characteristic contribution coefficient of each interactive behavior; Finally, in step S33, the evaluation data training set is updated using the feature contribution coefficients of all interactive behaviors to obtain an interactive behavior training set.

[0038] In the specific implementation, first, the set adjustment parameter can be obtained. The adjustment parameter is a parameter used to adjust the ratio between behavior characterization and behavior diversity. In actual implementation, it can be preset according to historical experimental data. The value of the adjustment parameter in this application is 0.53. Other values ​​can also be taken in actual implementation, which are not limited here; then, for each interactive behavior, the characteristic contribution coefficient of the interactive behavior is determined according to the behavioral characterization and behavioral diversity corresponding to the interactive behavior and the adjustment parameter, that is, the product of the adjustment parameter and the behavioral characterization of the interactive behavior is calculated, the difference between the adjustment parameter and 1 is calculated, and the product of the difference and the behavioral diversity of the interactive behavior is calculated. The sum of the above two products can be used as the coefficient of contribution. The feature contribution coefficient of the interactive behavior indicates the degree of contribution of the interactive behavior to the automated evaluation model training when it is added to the evaluation data training set. The larger the feature contribution coefficient, the higher the contribution. The updated evaluation data training set has a higher diversity. The feature contribution coefficient of each interactive behavior can be obtained in the above manner. Finally, the feature contribution coefficients of all interactive behaviors can be used to update the evaluation data training set, that is, the F interactive behaviors with the largest feature contribution coefficients are added to the evaluation data training set. The value of F can be set experimentally to obtain the interactive behavior training set, which is a data set used for subsequent training of the automated evaluation model.

[0039] It should be noted that by measuring behavioral diversity, the automated evaluation model can cover behavioral characteristics of different dimensions in the target enterprise, avoiding the insufficient generalization ability of the automated evaluation model due to single data. By combining behavioral representativeness and behavioral diversity to determine the interactive behavior training set, the automated evaluation model's ability to evaluate complex interactive processes can be improved.

[0040] The automatic evaluation module 400 is used to train the automatic evaluation model based on random forest using the interactive behavior training set, and generate an evaluation result for the target enterprise according to the trained automatic evaluation model.

[0041] In specific implementation, the interactive behavior training set can be used to train the random forest-based automated evaluation model, that is, the random forest algorithm is used to iteratively train the automated evaluation model based on the interactive behavior training set to improve the performance of the automated evaluation model.

[0042] In this embodiment, the evaluation results of the target enterprise can be generated according to the trained automatic evaluation model in the following manner, namely: Preprocess the target enterprise's desensitized assessment data to obtain preprocessed desensitized assessment data; The pre-processed desensitized evaluation data is input into the trained automated evaluation model for evaluation, thereby obtaining the evaluation result of the target enterprise.

[0043] In specific implementation, first, the purpose of preprocessing is to convert the desensitized evaluation data into a format suitable for input into the automated evaluation model, and to ensure the quality and consistency of the desensitized evaluation data. The preprocessing in this application is to remove missing values ​​and outliers in the desensitized evaluation data and then standardize the desensitized evaluation data, which will not be repeated here; then, the preprocessed desensitized evaluation data is input into the trained automated evaluation model, and the automated evaluation model evaluates the target enterprise based on the input data according to the learned features and patterns, thereby obtaining an evaluation result of the target enterprise, which includes a comprehensive risk score of the target enterprise and interactive risk scores of each dimension, wherein the comprehensive risk score is a score of the overall risk level of the target enterprise generated by comprehensively all interactive behaviors, and the risk scores of each dimension are risk level scores of the target enterprise in different dimensions, such as credit risk, operational risk, market risk, compliance risk, etc.

[0044] It should be noted that through a diverse and representative interactive behavior training set, the automated assessment model can more accurately identify and assess corporate risks, has stronger adaptability and robustness, and improves its generalization performance.

[0045] It can be seen that in this application, firstly, a behavior feature difference matrix is ​​constructed to convert complex interactive behavior differences into a structured matrix form, which helps to more clearly reveal the correlation and pattern between interactive behaviors. Based on the behavior feature difference matrix, the calculation of the abnormality of the behavior features can be made more accurate, and the behavior representativeness quantifies the representation ability of the interactive behavior in the global and local contexts, which can greatly improve the representativeness of the data in the training of the automated evaluation model; then, by measuring the behavior diversity, the automated evaluation model can cover the behavioral characteristics of different dimensions in the target enterprise, avoiding the insufficient generalization ability of the automated evaluation model due to the single data, and by combining the behavior representativeness and behavior diversity to determine the interactive behavior training set, the automated evaluation model can improve its evaluation ability for complex interactive processes; finally, through a diversified and representative interactive behavior training set, the automated evaluation model can more accurately identify and evaluate the risks of the enterprise, have stronger adaptability and robustness, and improve its generalization performance.

[0046] In summary, the technical solution adopted in this application can improve the representativeness and diversity of training data when training the automated evaluation model, so as to enhance the evaluation capability of the automated evaluation model.

[0047] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0048] A person skilled in the art may understand that all or part of the steps in the various methods of the above embodiments may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, the storage medium including a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically-erasable programmable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0049] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

Claims

1. An automated assessment system for enterprise risk management, characterized in that: The evaluation system comprises: An evaluation data acquisition module is used to obtain the enterprise evaluation data of the target enterprise in the risk management process, desensitize the enterprise evaluation data, and then obtain the desensitized evaluation data of the target enterprise; A representativeness determination module, used to construct a behavior characteristic difference matrix through the interactive behavior data in the desensitized evaluation data, determine the abnormality of the behavior characteristics of the target enterprise in the interactive process according to the behavior characteristic difference matrix, and determine the behavior representativeness of each interactive behavior in the target enterprise based on the behavior characteristic abnormality and the behavior characteristic difference matrix; A training set updating module is used to obtain a given evaluation data training set, determine the behavior diversity of each interactive behavior in the target enterprise based on the evaluation data training set, and update the evaluation data training set into an interactive behavior training set according to the behavior representativeness and behavior diversity of each interactive behavior; The automatic evaluation module is used to train the automatic evaluation model based on random forest using the interactive behavior training set, and generate an evaluation result for the target enterprise according to the trained automatic evaluation model.

2. An automated assessment system for enterprise risk management as claimed in claim 1, characterized in that: Obtain enterprise assessment data of the target enterprise in the risk management process from the management database of the target enterprise.

3. The automated assessment system for enterprise risk management according to claim 1, characterized in that: Desensitizing the enterprise evaluation data is to encrypt the sensitive information in the enterprise evaluation data.

4. The automated assessment system for enterprise risk management according to claim 1, characterized in that: Constructing a behavior feature difference matrix through the interactive behavior data in the desensitization evaluation data specifically includes: Obtaining interactive behavior data in the desensitization evaluation data; Determining the difference in interaction characteristics between each interaction behavior in the interaction behavior data; The behavior feature difference matrix is ​​constructed through the interaction feature differences between each interaction behavior.

5. The automated assessment system for enterprise risk management according to claim 1, characterized in that: Determining the abnormality of the behavior characteristics of the target enterprise during the interaction process according to the behavior characteristic difference matrix specifically includes: Extracting features from the behavioral feature difference matrix to obtain the interaction representation of the target enterprise during the interaction process; The abnormality of the behavior characteristics of the target enterprise during the interaction process is determined according to the interaction characterization quantity and the preset adjustment coefficient.

6. The automated assessment system for enterprise risk management according to claim 1, characterized in that: Determining the behavioral representation of each interactive behavior in the target enterprise based on the behavioral characteristic abnormality and the behavioral characteristic difference matrix specifically includes: For each interactive behavior in the target enterprise, extract all interactive feature differences corresponding to the interactive behavior in the behavioral feature difference matrix; The behavioral representativeness of the interactive behavior is determined by all the interactive feature differences corresponding to the interactive behavior and the abnormality of the behavioral feature, and then the behavioral representativeness of each interactive behavior in the target enterprise is obtained.

7. The automated assessment system for enterprise risk management according to claim 1, characterized in that: Determining the behavior diversity of each interactive behavior in the target enterprise based on the evaluation data training set specifically includes: For each interactive behavior in the target enterprise, the diversity of the interactive behavior is measured according to the evaluation data training set to obtain the behavioral diversity of the interactive behavior, and then the behavioral diversity of each interactive behavior in the target enterprise is obtained.

8. The automated assessment system for enterprise risk management according to claim 1, characterized in that: Updating the evaluation data training set into an interactive behavior training set according to the behavioral representation and behavioral diversity of each interactive behavior specifically includes: Get the set adjustment parameters; For each interactive behavior, determining a characteristic contribution coefficient of the interactive behavior according to the behavior representation and behavior diversity corresponding to the interactive behavior and the adjustment parameter, thereby obtaining the characteristic contribution coefficient of each interactive behavior; The evaluation data training set is updated using the feature contribution coefficients of all interactive behaviors to obtain an interactive behavior training set.

9. The automated assessment system for enterprise risk management according to claim 1, characterized in that: The evaluation results of the target enterprise generated based on the trained automatic evaluation model include: Preprocess the target enterprise's desensitized assessment data to obtain preprocessed desensitized assessment data; The pre-processed desensitized evaluation data is input into the trained automated evaluation model for evaluation, thereby obtaining the evaluation result of the target enterprise.

10. An automated assessment system for enterprise risk management according to claim 1, characterized in that: The assessment results include the comprehensive risk score of the target enterprise and the interactive risk scores of each dimension.

Images