An artificial intelligence-based compliance management data processing method and system

Abstract

The application discloses an artificial intelligence-based compliance management data processing system and method, wherein the method comprises the following steps: S1, obtaining an audit influencing factor of a material to be audited, wherein the audit influencing factor comprises basic information of the material to be audited, a risk degree value and / or a first personnel experience value, the basic information of the material to be audited comprises a tender project type to which the material to be audited belongs and a material content amount; and S2, after the audit influencing factor of the material to be audited is formed into first input data, the first input data is input into a deep learning training model for processing to obtain a first audit duration corresponding to the material to be audited. It can be seen that the application improves the accuracy of the determination of the audit duration of the material to be audited, and the risk degree value is introduced into the audit influencing factor, so that the risk prevention and control degree and the compliance of the tender work are further improved.

Description

Technical Field

[0001] This invention relates to compliance data processing technology, and in particular to a compliance management data processing method and system based on artificial intelligence. Background Technology

[0002] In order to make business processes more standardized and effectively achieve compliance risk control, cost reduction and efficiency improvement, and transformation, many companies are now using electronic bidding and procurement systems to realize the digital and information-based construction of bidding and procurement compliance management.

[0003] The current electronic bidding and procurement system mainly performs intelligent detection and identification of bidders' compliance. Once an anomaly is detected, a warning mechanism is immediately triggered. This intelligent compliance detection function mainly includes: 1. Detection and identification of purchase records: If at least two purchase records involve the same IP address, the same contact person's name, and / or the same contact person's phone number, the system will issue a warning to the project management member (e.g., by displaying a warning signal in a pop-up window on the management system interface and / or sending a warning message to the project management member's mobile terminal), so that they should be aware of the situation as soon as possible and implement corresponding corrective measures; 2. Detection and identification of bidding materials: If the uploaded and submitted bidding materials... If the IP addresses, contact names, phone numbers, MAC addresses, and / or hard drive machine codes are identical, or the wording of similar content in the bidding materials (such as historical project performance information, financial information, etc.) is extremely similar, the system will issue a warning to alert the bidding personnel that there may be non-compliant bidding (such as collusion or bid-rigging). 3. The system can identify the equity information of bidders by comparing their corporate credit information, shareholder relationships, and senior management relationships to identify any connections between bidders, such as controlling stakes, joint ownership, or shared senior management, thereby preventing bid-rigging and collusion. It is evident that the current electronic bidding and procurement system primarily focuses on intelligently detecting and identifying the bidding materials uploaded by bidders to achieve efficient and compliant processing. However, the functions of the electronic bidding and procurement system do not adequately consider the efficiency and compliance of the pre-bidding stage, especially the preparation and review of bidding materials. Currently, for the drafting and reviewing stage of bidding materials, the system's main functions are to assign the drafted bidding materials to appropriate reviewers and to compile and store the reviewed and revised materials. Regarding the allocation of bidding material review tasks, to ensure that the review tasks are completed on time and to ensure a reasonable distribution of workload among reviewers, the system generally uses the required review time for the materials to be reviewed as the basis for task allocation. However, the required review time for the materials to be reviewed is currently mainly determined based on the historical work experience of the assigned personnel. Therefore, the determination of this review time is easily influenced by subjective factors, resulting in low accuracy. This can easily lead to low overall review efficiency and increased workload for reviewers. Summary of the Invention

[0004] This application aims to address at least one of the technical problems existing in the prior art. To this end, this application proposes an artificial intelligence-based compliance management data processing method and system, which can improve the accuracy of determining the review time for bidding materials.

[0005] In a first aspect, embodiments of this application provide a compliance management data processing method based on artificial intelligence, the method comprising the following steps:

[0006] S1. Obtain the review influencing factors of the materials to be reviewed, wherein the review influencing factors include the basic information of the materials to be reviewed, the risk level value and / or the experience value of the first person in charge, and the basic information of the materials to be reviewed includes the type of bidding project to which the materials to be reviewed belong and the amount of material content;

[0007] S2. After taking the factors affecting the review of the materials to be reviewed as the first input data, input the first input data into the deep learning training model for processing to obtain the first review time corresponding to the materials to be reviewed.

[0008] In some embodiments, the risk level value is obtained through the following steps:

[0009] Obtain the audit risk level and / or the personnel identity verification security level;

[0010] The corresponding risk level value is determined based on the audit risk level and / or the personnel identity verification security level, wherein the audit risk level is determined based on the severity level of the audit risk triggering event, and the personnel identity verification security level is determined based on the security level of the personnel identity verification event.

[0011] In some embodiments, obtaining the audit risk level specifically includes:

[0012] If an audit risk triggering event is detected, determine whether there is an audit risk triggering event of the highest severity level.

[0013] When a review risk event of the highest severity level is identified, the first level will be used as the review risk level.

[0014] When it is determined that there is no audit risk triggering event with the highest severity level, a second level is determined based on the severity level corresponding to each audit risk triggering event, and the second level is used as the audit risk level.

[0015] In some embodiments, obtaining the security level of personnel identity verification specifically includes:

[0016] Obtain the security level of each user authentication event;

[0017] When a personnel authentication event with the highest security level exists, the third level will be used as the personnel authentication security level.

[0018] When there is no personnel authentication event with the highest security level, the fourth level is determined based on the security level of each personnel authentication event, and the fourth level is used as the personnel authentication security level.

[0019] In some embodiments, the first person's experience value is obtained through the following steps:

[0020] Obtain personnel's historical work information, wherein the historical work information includes the length of service and the first processing score of each historical project material;

[0021] The first experience value is determined based on the first processing score of each historical project material.

[0022] The first weighting coefficient is determined based on years of service.

[0023] The first experience value is adjusted using the first weighting coefficient to obtain the second experience value, which is then used as the first personnel experience value.

[0024] The years of service are directly proportional to the first weighting coefficient, and both the first weighting coefficient and the first experience value are directly proportional to the first person's experience value.

[0025] In some embodiments, the first processing score of the historical project materials is obtained through the following steps:

[0026] Obtain the second weight coefficient corresponding to the historical project materials and the second processing score of the historical project materials, wherein the second weight coefficient is determined based on the similarity between the historical project materials and the materials to be reviewed;

[0027] The first processing score is determined based on the second weighting coefficient and the second processing score.

[0028] In some embodiments, the second processing score of the historical project materials is obtained through the following steps:

[0029] After obtaining the number of modifications to historical project materials, a first score is determined based on the number of modifications, wherein the number of modifications and the first score are inversely proportional.

[0030] Obtain a third weighting coefficient and / or a fourth weighting coefficient, wherein the third weighting coefficient is determined based on the modification duration of historical project materials, and the fourth weighting coefficient is determined based on the importance of the modified content;

[0031] The second score is obtained by adjusting the first score based on the third and / or fourth weighting coefficients.

[0032] In some embodiments, the review influencing factors also include the importance value of the materials to be reviewed.

[0033] In some embodiments, the method further includes the following steps:

[0034] S3. Display the review task of the materials to be reviewed on the first interface. The first interface is provided with a first button, which is used to trigger the pop-up and display of the second interface. The second interface is used to display and modify the review influencing factors and the first review time.

[0035] Secondly, embodiments of this application provide an artificial intelligence-based compliance management data processing system, the system comprising:

[0036] The first acquisition unit is used to acquire the review influencing factors of the materials to be reviewed, wherein the review influencing factors include the basic information of the materials to be reviewed, the risk level value and / or the experience value of the first person. The basic information of the materials to be reviewed includes the type of bidding project to which the materials to be reviewed belong and the amount of material content.

[0037] The first processing unit is used to input the review influencing factors of the materials to be reviewed into the first input data, and then input the first input data into the deep learning training model for processing to obtain the first review time corresponding to the materials to be reviewed.

[0038] This application achieves at least one of the following technical effects: The proposed solution uses the factors influencing the review of the materials under review as the first input data, which is then processed by a deep learning training model to obtain the corresponding first review duration. These influencing factors include the basic information of the materials under review, a risk level value, and / or the experience value of the first personnel. The basic information of the materials under review includes the type of bidding project to which the materials belong and the amount of material content. Therefore, this application combines multi-dimensional considerations with a deep learning training model to determine the review duration, achieving high accuracy and improving the efficiency and quality of the review process. Furthermore, the inclusion of a risk level value among the influencing factors further enhances the risk control and compliance of the bidding process. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below.

[0040] Figure 1 This application provides a flowchart illustrating the steps of an artificial intelligence-based compliance management data processing method.

[0041] Figure 2This application provides a flowchart illustrating the steps for determining the risk level value in an artificial intelligence-based compliance management data processing method.

[0042] Figure 3 This application provides a flowchart illustrating the steps for determining the experience value of the first person in an artificial intelligence-based compliance management data processing method.

[0043] Figure 4 This application provides a schematic diagram of the framework of an artificial intelligence-based compliance management data processing system. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0045] To facilitate bidding and procurement management, an increasing number of enterprises are adopting electronic bidding and procurement systems to achieve digital and information-based compliance management, thereby playing a role in compliance risk control, cost reduction, and efficiency improvement. Currently, the main function of commonly used electronic bidding and procurement systems is to intelligently detect and identify the bidding materials uploaded by bidders, enabling efficient and compliant processing of these materials. However, for the early stages of bidding, especially the preparation and review of bidding materials, the system's functions are limited to basic features, such as distributing, compiling, and storing the bidding materials to be reviewed (which may at least include the bidding materials required by the bidding party during the bidding process, such as bidding announcements, bidding documents, and invitations to tender).

[0046] In commonly used electronic bidding and procurement systems, the review of prepared bidding materials is typically based on the required review time, which determines the assigned reviewers and the completion deadline. However, the determination of review time currently relies primarily on the historical experience of management personnel. This is not only susceptible to subjective human factors but also fails to account for other factors that influence review time, resulting in low accuracy and a poor fit with actual conditions. Therefore, this application provides an artificial intelligence-based compliance management data processing system and method that can improve the accuracy of review time determination.

[0047] Reference Figure 1This application provides an artificial intelligence-based compliance management data processing method, which includes the following steps.

[0048] S1. Obtain the review influencing factors of the materials to be reviewed, wherein the review influencing factors include the basic information of the materials to be reviewed, the risk level value and / or the experience value of the first person in charge, and the basic information of the materials to be reviewed includes the type of bidding project to which the materials to be reviewed belong and the amount of material content.

[0049] Specifically, the materials pending review refer to bidding materials compiled by the compilers and awaiting review by the reviewers.

[0050] Regarding the factors influencing the review process, these refer to factors that affect the review time of the materials to be reviewed. During the research process, it was found that these factors mainly include the following: 1. Basic information of the materials to be reviewed—This mainly includes the type of bidding project to which the materials belong and the amount of material content. Different types of bidding projects have different levels of difficulty. For example, the difficulty of a general contracting project for construction is greater than that of a construction project for general contracting, and the difficulty of a construction project for general contracting is greater than that of a construction supervision project for general contracting. Alternatively, construction-related bidding projects are more difficult than service-related ones. The more difficult the bidding project, the longer the review time required. Therefore, the type of bidding project to which the materials to be reviewed belong essentially represents the difficulty of reviewing the materials. Different bidding project types correspond to different levels of review difficulty, and the more difficult the bidding project, the more difficult the corresponding material review. As for the amount of content in the materials to be reviewed, the more content there is, the longer the review will take. This content amount can be represented by parameters such as the total number of words and / or the number of pages. Thus, the basic information of the materials to be reviewed will affect the review time. 2. Risk Level Value—The risk level value mainly... This is identified in response to security risk events that may arise during the compilation of materials. These security risk events can include network security issues (such as whether the network has been attacked), personnel identity verification issues, and material leaks. When these security risk events occur during the compilation process, the reviewers must verify these security risk situations during the review stage to determine whether there has been a leak of bidding material information. If so (i.e., the higher the severity of the security risk event), the more necessary it is to further determine the content of the leak and make corresponding modifications. Therefore, the higher the risk level, the longer the reviewers need to spend on the review work. It is evident that the degree of security risk events encountered by the compilers during the compilation stage will affect the review time of subsequent reviewers; 3. First-person experience value—The first-person experience value refers to the compilation experience value of the compilers who compiled the materials to be reviewed. If the compilers have a low number of years of experience in compiling bidding materials, and / or their evaluation scores in the process of compiling bidding materials are low, it indicates that the compilers have low experience in this area. This will lead to more revisions during review, and / or reviewers will spend more time reading and understanding the content of the materials. Therefore, it is evident that the experience value of the compilers will also affect the review time.

[0051] S2. After taking the factors affecting the review of the materials to be reviewed as the first input data, input the first input data into the deep learning training model for processing to obtain the first review time corresponding to the materials to be reviewed.

[0052] Specifically, the deep learning training model refers to a pre-trained deep learning model. The training steps for this model may include: using historical review factors affecting bidding materials as input data and historical review durations as output data; then training the deep learning model using both input and output data until the training termination condition is met. The deep learning training model can be implemented using multimodal fusion models, RNNs, LSTMs, or other existing models, depending on the specific circumstances and requirements; no particular limitation is made here.

[0053] As can be seen from the above, the solution of this application embodiment utilizes a combination of multi-dimensional considerations and a deep learning training model in determining the review time of the materials to be reviewed. This can greatly improve the accuracy of the review time determination, thereby improving the accuracy and rationality of the allocation of review tasks. It avoids situations where review tasks cannot be completed within the specified time limit due to unreasonable allocation or / or where the reviewers' tasks are particularly heavy. Moreover, the risk level value is also introduced as a consideration factor among the multi-dimensional considerations. This ensures that the review time not only meets the basic review and modification needs but also meets the needs for risk verification and confirmation, thereby further improving the risk control level and compliance of the bidding materials processing.

[0054] In some embodiments, the risk level value is primarily determined based on the audit risk level and / or the personnel authentication security level. The audit risk level is determined based on the severity level of the audit risk triggering event. This triggering event mainly refers to the first security risk event that occurs during the compilation of materials, such as network attacks, improper operations by the compiler (e.g., photographing the compiled materials with a mobile phone, accessing data using third-party devices), or unauthorized personnel spying on the compilation of bidding materials. These security risk events all pose a risk of information leakage. Therefore, in addition to reviewing the materials, the audit process also requires verification of these security risk events to ensure that the currently reviewed materials have not been leaked. If a leak is found, the content needs to be rectified promptly. Similarly, in addition to security risks that may arise during the writing process, it's also important to consider whether the writers encountered authentication security risks when logging into the system. For example, whether writers used MAC addresses to log in from terminals within a designated area, exceeded the limit for incorrect password attempts, exceeded the limit for incorrect biometric identification attempts, or had their authentication login information spied on by others. These issues can also lead to information leaks. Therefore, if authentication security risks occur during the writing phase, they should be verified during the subsequent review process. It is evident that compared to a writing phase without security risks, a writing phase with security risks requires a longer review period, and the more severe the security risk, the longer the review will take. Therefore, referring to... Figure 2 The risk level value can be obtained through the following steps.

[0055] A1. Obtain the audit risk level and / or the personnel identity verification security level.

[0056] A2. Determine the corresponding risk level value based on the audit risk level and / or the personnel identity verification security level, wherein the audit risk level is determined based on the severity level corresponding to the audit risk triggering event, and the personnel identity verification security level is determined based on the security level corresponding to the personnel identity verification event.

[0057] Specifically, the audit risk level can be directly the severity level corresponding to the audit risk triggering event (i.e., the security risk event that occurs during the compilation of materials), or the audit risk level can be obtained by optimizing and adjusting the severity level. Regarding the personnel identity verification security level, currently, in identity verification methods, if a security risk occurs, the corresponding security level identity verification mechanism is triggered based on the severity of the security risk. The more severe the security risk, the higher the security level of the triggered identity verification mechanism. Therefore, in this embodiment, the triggered personnel identity verification security level is directly used to represent the severity of the security risk that occurs during identity verification. Thus, it can be seen that both the audit risk level and the personnel identity verification security level are directly proportional to the risk level value; the higher the severity level of the audit risk triggering event and the higher the personnel identity verification security level, the greater the risk level value.

[0058] It is evident that by adopting the aforementioned method for determining risk levels, the potential for security risks is taken into account more comprehensively. This not only improves the accuracy of audit duration but also enhances the risk prevention and compliance of the audit process.

[0059] In some embodiments, the risk level value can directly include two variables: the audit risk level and the personnel authentication security level, which are then incorporated as feature parameters into the feature matrix of the first input data. Alternatively, the risk level value can be a single variable, and the final level value obtained by fusing the audit risk level and the personnel authentication security level is assigned to the risk level value. Given that this application utilizes a deep learning training model to process the input feature data to obtain the audit duration, directly including both the audit risk level and the personnel authentication security level in the risk level value to form the first input data not only improves data processing efficiency but also achieves higher accuracy.

[0060] In some embodiments, since at least two audit risk triggering events may occur during the material compilation process, and the severity levels of these at least two audit risk triggering events may be the same or different, the specific implementation steps for obtaining the audit risk level may specifically include the following steps.

[0061] A11. If no audit risk triggering event is detected, set the audit risk level to 0; if an audit risk triggering event is detected, determine whether there is an audit risk triggering event with the highest severity level.

[0062] A12. When it is determined that there is an audit risk triggering event with the highest severity level, the first level will be used as the audit risk level.

[0063] Specifically, if at least one audit risk of the highest severity level occurs among several audit risk trigger events, then the audit risk level is set to Level 1, and Level 1 represents the highest audit risk level. Wherein, if a smaller audit risk level value indicates a higher severity level, then Level 1 has the smallest value, and its value is smaller than Level 2; conversely, if a larger audit risk level value indicates a higher severity level, then Level 1 has the largest value, and its value is larger than Level 2. Further, in this embodiment, the method of using a larger audit risk level value to represent a higher severity level is adopted. This allows the highest severity level of the audit risk trigger event to be directly used as Level 1. For example, if the highest severity level is 5, then the audit risk level can also be set to 5. Of course, the highest severity level can also be adjusted by adding a weighting coefficient (which is an empirical value) to obtain the first level, such as Level 1 = Weighting coefficient k * Highest severity level. This can be selected and set according to actual needs, and no specific limitation is made here.

[0064] A13. When it is determined that there is no audit risk triggering event with the highest severity level, a second level is determined based on the severity level corresponding to each audit risk triggering event, and the second level is used as the audit risk level.

[0065] Specifically, the step of determining the second level based on the severity level corresponding to each audit risk triggering event includes: A131, obtaining the severity levels {y1, y2, y3, ... y} corresponding to several audit risk triggering events. n}, where y1 represents the severity level corresponding to the first audit risk triggering event, and so on; A132, after averaging the severity levels corresponding to several audit risk triggering events, the average severity level Y is obtained. avg =(y1+y2+y 3+ …+y n The severity level is calculated as (n) / n, and then the average severity level is used as the second level, where n represents the total number of severity levels corresponding to the audit risk triggering events. Of course, if a weighting coefficient k1 is introduced in this method of determining the audit risk level based on severity, then for the second level, it should be k1*Y. avg This ensures that the first level is always greater than the second level. Additionally, it should be noted that the setting of risk-triggered events and their corresponding severity levels can be implemented using the existing risk control functions in the electronic bidding and procurement system; no specific limitations are made here.

[0066] In some embodiments, the accuracy of identifying the "material peeping" event among various audit risk triggering events is relatively low. The current identification method involves using a camera (located in front of the display screen, on the same side as the display interface, to capture the area in front of the screen) on the computer terminal used by the compiler to photograph the area in front of the display screen. This allows identification of whether other personnel without material viewing permissions are lingering in that area; if so, material peeping is determined, otherwise, it is determined not to be material peeping. However, this method can lead to situations where a person merely passes through the area without their gaze resting on the display screen, and this is still considered material peeping. Therefore, to further improve the accuracy of identifying the material peeping audit risk triggering event and its corresponding severity level, the audit risk triggering event can be determined through the following steps.

[0067] A01. Acquire a first video image; wherein the first video image is obtained by a camera set in front of the display screen, that is, the camera is mainly used to capture the area in front of the display screen.

[0068] A02. When the display screen shows the content of the materials to be reviewed, the first video image is obtained by performing face recognition.

[0069] A03. After identifying the person's identity in the first face image, obtain the total permissions corresponding to that person's identity.

[0070] Specifically, after processing the face in the first face image using a facial recognition algorithm, the identity of the person corresponding to that face is obtained. Then, the total permissions corresponding to the person's identity are retrieved from a preset database. The facial recognition algorithm can be implemented using existing facial identity verification algorithms, which will not be elaborated upon here.

[0071] A04. If, based on the total permissions, it is determined that the person does not have the first permission, the eye region of the first face image is located and identified; if the eye region is identified, the audit risk trigger event is confirmed and processed.

[0072] Specifically, when someone is peeking at the content of the materials to be reviewed, their face must be facing the display interface. The eye region can then be located from the first facial image using eye features. Conversely, if the eye region cannot be identified from the first facial image, it indicates that the person has not actually viewed the content of the materials displayed on the interface. Therefore, eye region localization can prevent accidental triggering of review security risk events due to a person merely standing in front of the screen without actually viewing the materials, thus greatly improving the accuracy of identifying review risk triggering events.

[0073] Furthermore, to improve the accuracy of identifying audit risk trigger events, pupil localization can be performed on the eye area image. Then, by estimating the direction of the pupil's gaze, it can be determined whether the person's gaze is on the materials to be reviewed displayed on the screen, or even whether their gaze is on the screen itself. This avoids falsely triggering audit risk events due to the person's eyes being directed towards the screen when their gaze is not actually on the materials to be reviewed. Therefore, step A04 specifically includes the following steps.

[0074] A041. If, based on the total permissions, it is determined that the person does not have the first permission, eye region localization and recognition are performed on the first face image.

[0075] A042. If the eye region is identified, the pupil region is obtained from the eye region.

[0076] A043. After calculating the pupil gaze direction in the pupil region using a gaze direction estimation algorithm, the pupil gaze direction vector is obtained.

[0077] A044. Based on the coordinate transformation mapping relationship between the eye parameter coordinate system and the display screen coordinate system, the pupil gaze direction vector in the eye parameter coordinate system is converted into the first coordinate vector in the display screen coordinate system. Then, based on the position of the first coordinate vector in the display screen coordinate system, it is determined whether the person's pupil gaze falls on the page of the material to be reviewed.

[0078] A045. If it is determined that the person's pupils are focused on the page of the material to be reviewed, then a review risk trigger event is identified.

[0079] The severity level of a risk-triggered event can be determined through the following steps.

[0080] A05. If a security risk event is identified during the review process, the currently displayed page content is obtained, wherein the currently displayed page content refers to the content of the materials to be reviewed currently displayed on the display interface.

[0081] A06. Identify the importance level of the page content and determine the severity level of the audit risk trigger event based on the identified importance level. The higher the identified importance level, the higher the severity level.

[0082] Specifically, since technical specifications and financial information are more important than procedural information, and key indicators and parameters are more important than procedural / routine indicators and parameters, the negative impact of leaking technical specifications and financial information before the announcement, or leaking key indicators and parameters before the announcement, will be greater than that of procedural information. Therefore, in the case of the same audit risk triggering event, if the material being leaked is more important, the severity level of the event should be higher.

[0083] It is evident that using the above method to determine audit risk triggering events and their corresponding severity levels is more accurate and can greatly reduce false triggers, thereby avoiding the increase of ineffective work by staff and improving the overall processing efficiency.

[0084] Additionally, it should be noted that the image localization and recognition of the pupil region mentioned above can be achieved using existing algorithms such as ellipse fitting and / or CNN convolutional neural networks; similarly, the gaze direction estimation algorithm can be achieved using the gaze vector calculation algorithm, which will not be elaborated here.

[0085] In some embodiments, given the above-described method of obtaining the audit risk level, similarly, the step of obtaining the personnel identity verification security level may specifically include the following steps.

[0086] A14. Obtain the security level of each personnel authentication event.

[0087] A15. When a personnel authentication event with the highest security level exists, the third level shall be used as the personnel authentication security level.

[0088] A16. When there is no personnel authentication event with the highest security level, determine the fourth level based on the security level of each personnel authentication event, and use the fourth level as the personnel authentication security level.

[0089] Specifically, if there are 5 security levels for personnel authentication events, with 5 representing the highest security level, then when a personnel authentication event with the highest security level exists, 5 will be used as the personnel authentication security level. Conversely, the average security level S will be obtained by averaging the security levels of several personnel authentication events, such as 2, 3, 3, and 4. avg =(s1+s2+s3+ …+s m The security level is calculated as s1 / m, and then the average security level is used as the fourth level. Here, s1 represents the security level of the first personnel authentication event, and so on, with m representing the total number of security levels for personnel authentication events. Of course, if a weighting coefficient k2 is introduced in this method of determining the personnel authentication security level based on the security level, then for the fourth level, it should be k2*Savg, thus ensuring that the third level is greater than the fourth level. It should also be noted that the setting of personnel authentication events and their corresponding security levels can be achieved through the existing authentication security mechanisms in the electronic bidding and procurement system, which will not be elaborated on here.

[0090] It is evident that by adopting the aforementioned methods for obtaining audit risk levels and personnel identity verification security levels, the accuracy of risk level determination can be further improved, providing precise data support for determining subsequent audit durations.

[0091] In some embodiments, refer to Figure 3 The experience value of the first person is obtained through the following steps.

[0092] B1. Obtain the personnel's historical work information, wherein the historical work information includes the length of service and the first processing score of each historical project material.

[0093] Specifically, the years of service mainly refer to the years of experience related to the processing of bidding materials, while the first processing score of historical project materials is mainly used to characterize the quality of the work done by the compilers in each previous processing of bidding materials.

[0094] B2. Determine the first experience value based on the first processing score of each historical project material.

[0095] Specifically, in this embodiment, the first empirical value is calculated as follows: First empirical value = (p1 + p2 + p3 + ... + p j ) / j, where p represents the first processing score of the historical project materials, p i Let p1 represent the first processing score of the i-th historical project material, and so on; j represents the total number of processing scores for historical project materials. In other words, step B2 specifically involves: calculating the average of the first processing scores of several historical project materials to obtain the average first score, and then using the average first score as the first empirical value.

[0096] B3. Determine the first weighting coefficient based on the length of service, wherein the length of service and the first weighting coefficient are directly proportional.

[0097] B4. After adjusting the first experience value using the first weighting coefficient, a second experience value is obtained, and the second experience value is used as the first personnel experience value; both the first weighting coefficient and the first experience value are directly proportional to the first personnel experience value.

[0098] Specifically, to improve the accuracy of the first experience value and more accurately reflect the experience level of the compiler, it is preferable to use a weighting coefficient corresponding to the years of service to adjust the first experience value. That is, for the second experience value, it is specifically: Second experience value = First weighting coefficient k3 * First experience value. In this case, the second experience value is the first experience value of the compiler.

[0099] In some embodiments, because there are discrepancies between the project type, details, and other information of the processed historical project materials and the materials to be reviewed, in order to further improve the accuracy of the first experience value, the processing score corresponding to the historical project materials that are more similar to the materials to be reviewed is given a larger weight. That is, the more similar the historical project materials are to the materials to be reviewed, the more their processing score reflects the experience level of the compiler in compiling the materials to be reviewed. In view of this, the first processing score for historical project materials can be obtained through the following steps.

[0100] C1. Obtain the second weight coefficient corresponding to the historical project materials and the second processing score of the historical project materials, wherein the second weight coefficient is determined based on the similarity between the historical project materials and the materials to be reviewed.

[0101] C2. Based on the second weighting coefficient and the second processing score, the first processing score is determined.

[0102] Specifically, the second processing score of the historical project material is the original processing score. Therefore, the higher the similarity between the historical project material and the material to be reviewed, the higher the influence of the corresponding original processing score should be. Thus, there is a direct proportional relationship between the similarity and the second weighting coefficient; the greater the similarity, the larger the second weighting coefficient. Therefore, the first processing score = second weighting coefficient k4 * second processing score. The processed score, i.e., the first processing score, is then used as the final required processing score. Furthermore, the similarity between materials can be achieved using existing algorithms such as those based on TF-IDF and cosine similarity, or those based on BERT semantic embedding, which will not be elaborated upon here.

[0103] Given the introduction of the aforementioned similarity, step B2 may specifically include the following steps.

[0104] B21. Obtain the similarity between the materials of each historical project and the materials to be reviewed. At this point, several similarity scores are obtained.

[0105] B22. Select the first similarity from several similarities, wherein the first similarity is a similarity greater than or equal to a first threshold, the second processing score of the historical project material corresponding to the first similarity is the third processing score, and the second similarity is a similarity less than the first threshold.

[0106] B23. When at least one first similarity exists, the second processing score of the historical project material corresponding to the second similarity is deleted, i.e., the third processing score is retained. Then, the first empirical value is determined based on the retained third processing score. That is, after averaging several third processing scores, the average of the second scores is obtained, and the average of the second scores is used as the first empirical value. In this case, the retained third processing score can be understood as the first processing score. It can be seen that in determining the first empirical value, the processing score corresponding to the historical project material that is more similar to the material to be reviewed is selected as the basis, while the processing score corresponding to the historical project material that is less similar to the material to be reviewed is not considered.

[0107] B24. When no first similarity exists, the first processing score is determined based on the second weighting coefficient and the second processing score. Then, the first empirical value is determined based on the first processing score of each historical project material. The second weighting coefficient and the second processing score are as described above and will not be elaborated further here.

[0108] It is evident that using the above method to determine the first experience value better reflects the experience level of the compilers in handling the materials to be reviewed, thus providing more accurate data support for determining the subsequent review time.

[0109] In some embodiments, to further improve the accuracy of the first person's experience value, the second processing score of the historical project material is obtained through the following steps.

[0110] E1. After obtaining the number of revisions to historical project materials, a first score is determined based on the number of revisions. The number of revisions and the first score are inversely proportional. That is, the number of revisions refers to the total number of revisions made to the project materials during the review process from submission to completion. Therefore, the more revisions, the lower the first score.

[0111] E2. Obtain the third weighting coefficient and / or the fourth weighting coefficient, wherein the third weighting coefficient is determined based on the modification duration of historical project materials, and the fourth weighting coefficient is determined based on the importance of the modified content.

[0112] Specifically, the revision duration for historical project materials refers to the revision time required for a single draft. This revision duration can be determined as follows: the moment the draft is received by the editor is taken as the start time, and the moment it is submitted to the reviewer is taken as the end time. The time between the start and end times is then considered the revision duration. Furthermore, to improve the accuracy of this revision duration determination, the moment the draft is first opened by the editor can be taken as the start time. Therefore, the revision duration for a historical project material can include T1, T2, T3, ..., T... l Where l represents the total number of revisions to the historical project materials, and T1 represents the time spent on the first revision, i.e., the first revision duration, T2, T3, ..., T l And so on. The method for determining the third weighting coefficient k5 is as follows: after averaging the modification time of several historical project materials, the average time is obtained, and then the third weighting coefficient k5 is determined based on the average time. The average time and the third weighting coefficient are inversely proportional, that is, the shorter the average modification time, the larger the third weighting coefficient.

[0113] Regarding the importance of revisions, reviewers typically mark necessary changes in materials with annotations and / or revisions. Therefore, by identifying these annotations and / or revisions in the submitted draft, the importance of the content can be determined based on its type. For example, technical specifications and financial information are considered more important than procedural information. Thus, in the review process of historical project materials, several importance levels for revisions exist: V1, V2, V3, ..., V... p Where p represents the total number of importance levels for the modified content, V1 represents the importance level of the first modified content, and V2, V3, ..., V... p And so on. The fourth weighting coefficient k6 is determined as follows: after averaging the importance levels of several modifications to the historical project materials, the average importance level is obtained. Then, the fourth weighting coefficient k6 is determined based on the average importance level. The average importance level and the fourth weighting coefficient k6 are inversely proportional, that is, the higher the average importance level of the modifications, the lower the fourth weighting coefficient k6.

[0114] E3. After adjusting the first score value according to the third weighting coefficient and / or the fourth weighting coefficient, a second score value is obtained. The third weighting coefficient, the fourth weighting coefficient, and the first score value are all directly proportional to the second score value. In this embodiment, the second score value is specifically: k5 * k6 * the first score value. Alternatively, empirical fixed values ​​can be introduced into the calculation formula of the second score value for fine-tuning to make it more accurate. This can be set and determined according to actual needs and is not specifically limited here.

[0115] In some embodiments, the review influencing factors also include a importance value for the materials to be reviewed. This importance value is generally set manually, primarily for cases where the materials belong to the same type of bidding project and other factors are similar; in such cases, management personnel need to determine it.

[0116] In some embodiments, given that the review influencing factors and / or the first review duration may be subject to strong human intervention before the review task of the materials to be reviewed is assigned, a function to modify these parameters is provided on the first interface used to display the review task of the materials to be reviewed. Therefore, the method of this application may further include the following steps: S3, displaying the review task of the materials to be reviewed on the first interface, wherein the first interface is provided with a first button, which is used to trigger the pop-up and display of a second interface, which is used to display and modify the review influencing factors and the first review duration. This facilitates staff in modifying and determining the review influencing factors and / or the first review duration. Of course, this modification function requires permission verification before it can be activated, to prevent unauthorized staff from modifying it, thereby improving the security and reliability of system operation.

[0117] Reference Figure 4 This application provides an artificial intelligence-based compliance management data processing system, which includes:

[0118] The first acquisition unit is used to acquire the review influencing factors of the materials to be reviewed, wherein the review influencing factors include the basic information of the materials to be reviewed, the risk level value and / or the experience value of the first person. The basic information of the materials to be reviewed includes the type of bidding project to which the materials to be reviewed belong and the amount of material content.

[0119] The first processing unit is used to input the review influencing factors of the materials to be reviewed into the first input data, and then input the first input data into the deep learning training model for processing to obtain the first review time corresponding to the materials to be reviewed.

[0120] The units of the above system embodiment correspond one-to-one with the steps of the method embodiment. Therefore, the implementation principle and beneficial effects of this system embodiment are the same as those of the above method embodiment, and will not be repeated here.

[0121] Furthermore, embodiments of this application also provide a computer-readable storage medium storing a computer program, which is executed by a processor to implement the steps of the above-described method embodiments.

[0122] The number of processors mentioned in the above storage medium embodiments and system embodiments can be at least one, capable of executing at least any of the steps in the above method embodiments. When the number is at least two, the at least two processors can communicate with each other, not limited to wired or wireless communication connections, and the at least one processor can communicate with various smart terminal devices. Furthermore, the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0123] Finally, it should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0124] Note that the above are merely preferred embodiments and the technical principles employed in this application. Those skilled in the art will understand that this application is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of this application. Therefore, although this application has been described in detail through the above embodiments, this application is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of this application, the scope of which is determined by the scope of the appended claims.

Claims

1. A compliance management data processing method based on artificial intelligence, characterized in that, The method includes the following steps: S1. Obtain the review influencing factors of the materials to be reviewed, wherein the review influencing factors include the basic information of the materials to be reviewed, the risk level value, and the experience value of the first person in charge. The basic information of the materials to be reviewed includes the type of bidding project to which the materials to be reviewed belong and the amount of material content. The risk level value is obtained through the following steps: obtaining the audit risk level and the personnel identity verification security level; determining the corresponding risk level value based on the audit risk level and the personnel identity verification security level, wherein the audit risk level is determined based on the severity level corresponding to the audit risk triggering event, and the personnel identity verification security level is determined based on the security level corresponding to the personnel identity verification event; The audit risk triggering events and their corresponding severity levels are determined through the following steps: A01. Acquire a first video image; wherein the first video image is captured by a camera positioned in front of the display screen; A02. When the display screen shows the content of the materials to be reviewed, perform facial recognition on the first video image to obtain a first face image; A03. After identifying the person's identity in the first face image, obtain the total permissions corresponding to that person's identity; A041. If it is determined from the total permissions that the person's identity does not have the first permission, perform eye region localization and recognition on the first face image; A042. If the eye region is identified, obtain the pupil region from the eye region; A043. After calculating the pupil gaze direction using a gaze direction estimation algorithm, obtain the pupil gaze direction vector; A0 44. Based on the coordinate transformation mapping relationship between the eye parameter coordinate system and the display screen coordinate system, the pupil gaze direction vector in the eye parameter coordinate system is converted into a first coordinate vector in the display screen coordinate system. Then, based on the position of the first coordinate vector in the display screen coordinate system, it is determined whether the person's pupil gaze falls on the page of the material to be reviewed; A045. If it is determined that the person's pupil gaze falls on the page of the material to be reviewed, then a review risk triggering event is identified; A05. If a review security risk event is identified, the currently displayed page content is obtained, wherein the currently displayed page content refers to the content of the material to be reviewed currently displayed on the display interface; A06. The importance level of the page content is identified, and the severity level corresponding to the review risk triggering event is determined based on the identified importance level, wherein the higher the identified importance level, the higher the severity level; The first personnel experience value is obtained through the following steps: obtaining the personnel's historical work information, wherein the historical work information includes years of service and the first processing score of each historical project material; the first processing score of the historical project material is obtained through the following steps: obtaining the second weighting coefficient and the second processing score of the historical project material, wherein the second weighting coefficient is determined based on the similarity between the historical project material and the material to be reviewed; and determining the first processing score based on the second weighting coefficient and the second processing score, wherein the first processing score = second weighting coefficient * second processing score; the second processing score of the historical project material is obtained through the following steps: obtaining the number of modifications to the historical project material, and determining the first score value based on the number of modifications, wherein the number of modifications and the first score value are inversely proportional; obtaining the third... The system employs a weighting coefficient and a fourth weighting coefficient. The third weighting coefficient is determined based on the modification duration of historical project materials, and the fourth weighting coefficient is determined based on the importance of the modified content. After adjusting the first score based on the third and fourth weighting coefficients, a second score is obtained, specifically: third weighting coefficient * fourth weighting coefficient * first score. The average of the first processed scores for several historical project materials is calculated to obtain the first average score, which is then used as the first experience value. A first weighting coefficient is determined based on years of service, where years of service are directly proportional to the first weighting coefficient. The first experience value is then adjusted using the first weighting coefficient to obtain a second experience value, which is used as the first personnel experience value, where the second experience value = first weighting coefficient * first experience value. S2. After taking the factors affecting the review of the materials to be reviewed as the first input data, input the first input data into the deep learning training model for processing to obtain the first review time corresponding to the materials to be reviewed.

2. The method as described in claim 1, characterized in that, The acquisition of the audit risk level specifically includes: If an audit risk triggering event is detected, determine whether there is an audit risk triggering event of the highest severity level. When a review risk event of the highest severity level is identified, the first level will be used as the review risk level. When it is determined that there is no audit risk triggering event of the highest severity level, a second level is determined based on the severity level corresponding to each audit risk triggering event, and the second level is used as the audit risk level.

3. The method as described in claim 1, characterized in that, The security level for obtaining personnel identity verification specifically includes: Obtain the security level of each user authentication event; When a personnel authentication event with the highest security level exists, the third level will be used as the personnel authentication security level. When there is no personnel authentication event with the highest security level, the fourth level is determined based on the security level of each personnel authentication event, and the fourth level is used as the personnel authentication security level.

4. The method as described in claim 1, characterized in that, The factors influencing the review also include the importance value of the materials to be reviewed.

5. The method according to any one of claims 1-4, characterized in that, The method also includes the following steps: S3. Display the review task of the materials to be reviewed on the first interface. The first interface is provided with a first button, which is used to trigger the pop-up and display of the second interface. The second interface is used to display and modify the review influencing factors and the first review time.

6. A compliance management data processing system based on artificial intelligence, characterized in that, The system includes: The first acquisition unit is used to acquire the review influencing factors of the materials to be reviewed. The review influencing factors include the basic information of the materials to be reviewed, the risk level value, and the experience value of the first person. The basic information of the materials to be reviewed includes the type of bidding project to which the materials to be reviewed belong and the amount of material content. The risk level value is obtained through the following steps: obtaining the audit risk level and the personnel identity verification security level; determining the corresponding risk level value based on the audit risk level and the personnel identity verification security level, wherein the audit risk level is determined based on the severity level corresponding to the audit risk triggering event, and the personnel identity verification security level is determined based on the security level corresponding to the personnel identity verification event; The audit risk triggering events and their corresponding severity levels are determined through the following steps: acquiring a first video image; wherein the first video image is captured by a camera positioned in front of the display screen; when the display screen shows the content of the materials to be audited, performing facial recognition on the first video image to obtain a first face image; performing personnel identification on the face in the first face image to obtain the total permissions corresponding to the personnel identity; if it is determined based on the total permissions that the personnel identity does not have the first permission, performing eye region localization and identification on the first face image; if the eye region is identified, obtaining the pupil region from the eye region; after calculating the pupil gaze direction using a gaze direction estimation algorithm, obtaining the pupil gaze direction vector; and then, based on the eye parameters... The coordinate transformation mapping relationship between the coordinate system and the display screen coordinate system converts the pupil gaze direction vector in the eye parameter coordinate system into a first coordinate vector in the display screen coordinate system. Then, based on the position of the first coordinate vector in the display screen coordinate system, it is determined whether the person's pupil gaze is on the page of the material to be reviewed. If it is determined that the person's pupil gaze is on the page of the material to be reviewed, a review risk trigger event is identified. If a review security risk event is identified, the currently displayed page content is obtained, where the currently displayed page content refers to the content of the material to be reviewed currently displayed on the display interface. The importance level of the page content is identified, and the severity level corresponding to the review risk trigger event is determined based on the identified importance level, where the higher the identified importance level, the higher the severity level. The first personnel experience value is obtained through the following steps: obtaining the personnel's historical work information, wherein the historical work information includes years of service and the first processing score of each historical project material; the first processing score of the historical project material is obtained through the following steps: obtaining the second weighting coefficient and the second processing score of the historical project material, wherein the second weighting coefficient is determined based on the similarity between the historical project material and the material to be reviewed; and determining the first processing score based on the second weighting coefficient and the second processing score, wherein the first processing score = second weighting coefficient * second processing score; the second processing score of the historical project material is obtained through the following steps: obtaining the number of modifications to the historical project material, and determining the first score value based on the number of modifications, wherein the number of modifications and the first score value are inversely proportional; obtaining the third... The system employs a weighting coefficient and a fourth weighting coefficient. The third weighting coefficient is determined based on the modification duration of historical project materials, and the fourth weighting coefficient is determined based on the importance of the modified content. After adjusting the first score based on the third and fourth weighting coefficients, a second score is obtained, specifically: third weighting coefficient * fourth weighting coefficient * first score. The average of the first processed scores for several historical project materials is calculated to obtain the first average score, which is then used as the first experience value. A first weighting coefficient is determined based on years of service, where years of service are directly proportional to the first weighting coefficient. The first experience value is then adjusted using the first weighting coefficient to obtain a second experience value, which is used as the first personnel experience value, where the second experience value = first weighting coefficient * first experience value. The first processing unit is used to input the review influencing factors of the materials to be reviewed into the first input data, and then input the first input data into the deep learning training model for processing to obtain the first review time corresponding to the materials to be reviewed.

Images