An AI intelligent investment decision optimization method and investment decision optimization system

Abstract

The application relates to the field of artificial intelligence decision-making technology and proposes an AI intelligent investment decision optimization method and an investment decision optimization system, which comprises the following steps: configuring an AI investment decision result set and an AI income item for an AI investment decision maker, configuring an artificial investment decision result set and an artificial income item for a multi-level artificial investment decision maker, and constructing programmatic constraints and threshold constraints of the AI investment decision maker; analyzing the historical adoption rate of the multi-level artificial investment decision maker in the artificial investment decision result set through the artificial income item, outputting the candidate investment decision tendency of the AI investment decision maker in the AI investment decision result set according to the programmatic constraints and the threshold constraints, screening the AI investment decision tendency which meets the candidate investment decision tendency, analyzing the AI adoption rate and the trace effectiveness of the AI investment decision tendency through the AI income item, and adjusting the checking strictness of the AI investment decision maker on the programmatic constraints and the threshold constraints according to the AI adoption rate and the trace effectiveness. The application realizes effective interaction between the AI and the multi-level artificial investment decision subject.

Description

Technical Field

[0001] This invention belongs to the field of artificial intelligence decision-making technology, specifically relating to an AI-powered intelligent investment decision optimization method and an investment decision optimization system. Background Technology

[0002] Multi-level, hierarchical investment decision-making review models are commonly used in large enterprises' infrastructure investment, equity investment, and credit granting scenarios. This model involves bottom-up, hierarchical manual review, with the enterprise's leaders and employees participating as decision-makers in the investment decision-making process. In recent years, with the gradual introduction of artificial intelligence into investment decision-making assistance scenarios, numerous AI-assisted investment decision-making systems have emerged in the industry. In existing solutions, AI models independently generate investment decision recommendations and then deliver them to human reviewers. The drawback of this existing solution is that the AI ​​does not understand the preferences of each level of review, and the output recommendations are prone to conflict with human expectations. When the investment decision recommendations generated by the AI ​​do not conform to the preferences of each level of reviewer, the human decision-makers, who hold the final decision-making power, can directly ignore, reject, or even bypass the investment decision recommendations generated by the AI ​​model. Therefore, there is no effective interaction between the AI ​​and the multi-level human decision-makers.

[0003] For example, patent application CN119962739A discloses a smart heating system optimization scheduling platform that considers human-machine collaboration. This patent application relies on an AI unit to output scheduling strategies, which are then manually verified by a dispatcher. This system still follows a one-way machine-output approach with manual review, making it difficult to resolve conflicts between human and machine opinions. Therefore, existing technologies suffer from the deficiency of ineffective interaction between AI and multi-level human decision-making bodies. Summary of the Invention

[0004] This invention provides an AI-powered intelligent investment decision optimization method and investment decision optimization system, the main purpose of which is to enable effective interaction between AI and multi-level human investment decision-making entities.

[0005] To achieve the above objectives, the present invention provides an AI-powered intelligent decision-making optimization method, comprising: Configure AI voting result sets and AI benefit items for AI voting decision-makers, configure human voting result sets and human benefit items for multi-level human voting decision-makers, and construct the procedural constraints and threshold constraints of the AI ​​voting decision-makers; By analyzing the historical adoption rate of multi-level human voters in the human voting result set through human benefit analysis, and outputting the candidate voting tendency of AI voters in the AI ​​voting result set based on procedural constraints and threshold constraints. Based on historical adoption rates, AI voting tendencies that match candidate voting tendencies are selected from the highest level of human voters to the lowest level of human voters. The AI ​​voting tendency is pushed to human voters at all levels to determine the current voting tendency of human voters at each level. Based on the current voting tendency, the adoption rate and traceability effectiveness of the AI ​​voting tendency are analyzed through AI benefit items. Adjust the stringency of AI voting decisions regarding procedural and threshold constraints based on AI adoption rates and traceability effectiveness to optimize AI-driven intelligent voting.

[0006] Optionally, based on historical adoption rates, AI voting tendencies that match candidate voting tendencies are screened from the highest-level human voter to the lowest-level human voter, including: Based on the matching results between the target strategy and historical input features, the target adoption rate corresponding to the target strategy is selected from the historical adoption rates; Based on the target adoption rate, starting from the highest level of human voter and going down to the lowest level of human voter, the deduced voting tendency of each level of human voter regarding the candidate voting tendency is deduced step by step. When the predicted voting tendencies of all levels of human voters are consistent with the candidate voting tendencies, the candidate voting tendencies regarding the target strategy are taken as the AI ​​voting tendencies of the AI ​​voters.

[0007] Optionally, based on the target adoption rate, starting from the highest-level human voter and going down to the lowest-level human voter, the inferred voting tendency of each level of human voter regarding the candidate is deduced step by step, including: The projected voting tendencies of the highest-level human voters are determined based on the voting tendencies of the candidates. Starting from the predicted voting tendency of the highest-level human voter, the human vote result corresponding to the highest adoption rate among the target adoption rates of the current-level human voter is taken as the predicted voting tendency of the current-level human voter. If the predicted voting tendency of the current level of human voters is consistent with the voting tendency of the corresponding candidate, then it continues to be passed down to the next level; If the predicted voting tendency of the current level of human voter is inconsistent with the corresponding candidate voting tendency, the transmission to the next level will stop, so as to realize the process of reversing the predicted voting tendency of each level of human voter for the candidate voting tendency.

[0008] Optionally, configure the AI ​​investment decision set and AI benefit items for the AI ​​investment decision-maker, including: An AI voting result set is established from the confirmation and rejection results; An AI benefit item is established by combining the AI ​​adoption rate item and the traceability validity item. The AI ​​adoption rate item is the percentage of levels where the AI ​​voting results are consistent with the human voting results at each level out of the total number of human voting levels. The traceability validity item is the percentage of levels where the human voting results are not nonexistent in response to the AI ​​voting results out of the total number of human voting levels.

[0009] Optionally, configure a set of human voting results and human benefit items for multi-level human voting decision-makers, including: A set of manual voting results is established based on the confirmed results, rejected results, and no results; A human benefit item is established based on the human adoption rate item, which is the percentage of times that the human voting results at each level are consistent with the AI ​​voting results out of the total number of times voting is participated in at each level.

[0010] Optionally, the procedural constraints and threshold constraints of the AI ​​voter are constructed, including: In the voting process of the AI ​​voter, procedural constraints are determined by the migration conditions between the voting node and its adjacent nodes; Threshold constraints are determined by quantifying the threshold range corresponding to the indicators.

[0011] Optionally, the historical adoption rate of multi-level human voters within the human voting result set can be analyzed through human benefit analysis, including: For each type of historical input feature, the historical adoption rate of each level of human voter in the human vote result set is calculated through the human benefit item; The historical adoption rate includes the confirmed adoption rate and the rejected adoption rate.

[0012] Optionally, based on procedural constraints and threshold constraints, output the candidate voting tendencies of AI voters within the AI ​​voting result set, including: For different investment decisions on the same project, select target strategies that meet procedural constraints and threshold constraints; Generate candidate voting tendencies that are confirmed at each level regarding the target strategy.

[0013] Optionally, the stringency of AI voter verification regarding procedural and threshold constraints can be adjusted based on AI adoption rates and retrospective effectiveness, including: When the AI ​​adoption rate is lower than the first preset threshold, the verification strictness of procedural constraints and threshold constraints is increased. When the traceability validity is lower than the second preset threshold, the verification strictness of procedural constraints and threshold constraints is reduced; When the AI ​​adoption rate is not lower than the first preset threshold and the traceability effectiveness is not lower than the second preset threshold, the strictness of the verification of procedural constraints and threshold constraints is maintained.

[0014] To achieve the above objectives, the present invention provides an AI-powered intelligent investment decision optimization system, comprising: The constraint construction module is used to configure the AI ​​voting result set and AI benefit item for AI voting decision-makers, configure the human voting result set and human benefit item for multi-level human voting decision-makers, and construct the procedural constraints and threshold constraints of the AI ​​voting decision-makers; The candidate output module is used to analyze the historical adoption rate of multi-level human voters in the human voting result set through human benefit items, and output the candidate voting tendency of AI voters in the AI ​​voting result set based on procedural constraints and threshold constraints. The bias screening module is used to screen AI voting biases that match candidate voting biases, based on historical adoption rates, starting from the highest level of human voters and ending at the lowest level of human voters. The effectiveness analysis module is used to push the AI ​​voting tendency to human voters at each level to determine the current voting tendency of human voters at each level. Based on the current voting tendency, the AI ​​adoption rate and traceability effectiveness of the AI ​​voting tendency are analyzed through AI benefit items. The voting optimization module is used to adjust the strictness of AI voting on procedural and threshold constraints based on the AI ​​adoption rate and traceability effectiveness, so as to optimize AI intelligent voting.

[0015] Compared to the problems described in the background art, this invention configures AI decision-makers with AI decision-making result sets and AI payoff items, thus positioning them as game participants with independent decision-making abilities and payoff objectives. This makes the AI ​​decision-maker and multi-level human decision-makers form equal game entities, laying the groundwork for the subsequent output of independent decision-making tendencies by the AI. Furthermore, this invention configures multi-level human decision-makers with human decision-making result sets and human payoff items, positioning them as game participants with independent decision-making abilities and payoff objectives, thus forming equal game entities with the AI ​​decision-maker. Further, this invention constructs procedural and threshold constraints for the AI ​​decision-maker, enabling it to automatically filter input strategies based on these constraints. Finally, this invention analyzes the human payoff items within the human decision-making result sets of multi-level human decision-makers. The invention quantifies the decision preferences of human voters at various levels by using historical adoption rates. Furthermore, it outputs the candidate voting tendencies of AI voters within the AI ​​voting result set based on procedural and threshold constraints, thus filtering voting strategies within these constraints. Further, based on historical adoption rates, the invention filters AI voting tendencies that match candidate tendencies from the highest to the lowest level of human voters, using a top-down reverse deduction of multi-level human voting preferences to prioritize voting schemes more easily passed through each level of review. This reduces instances where humans ignore, reject, or bypass AI suggestions. Furthermore, the invention analyzes the AI ​​adoption rate and traceability effectiveness of AI voting tendencies based on current voting tendencies and AI benefit items, assessing the suitability and verifiability of AI voting tendencies in multi-level voting reviews, providing data for continuous optimization of AI voting. Therefore, this invention enables effective interaction between AI and multi-level human voting entities. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating an AI-powered intelligent decision-making optimization method according to an embodiment of the present invention. Figure 2 A schematic diagram illustrating the derivation of an AI intelligent decision optimization method according to an embodiment of the present invention; Figure 3 This is a schematic diagram of manual voting in an AI intelligent voting optimization method according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the entire AI voting process of an AI intelligent voting optimization method provided in an embodiment of the present invention; Figure 5 This is a functional block diagram of an AI-powered intelligent investment decision optimization system provided in an embodiment of the present invention; Figure 6A schematic diagram of a computer device for an AI intelligent decision optimization method according to an embodiment of the present invention; The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0018] This application provides an AI-powered intelligent decision-making optimization method. The executing entity of this AI-powered intelligent decision-making optimization method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the AI-powered intelligent decision-making optimization method can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster.

[0019] Reference Figure 1 The diagram shown is a flowchart illustrating an AI-powered intelligent voting decision optimization method according to an embodiment of the present invention. In this embodiment, the AI-powered intelligent voting decision optimization method includes: S1. Configure AI voting result set and AI benefit item for AI voter, configure human voting result set and human benefit item for multi-level human voter, and construct the procedural constraints and threshold constraints of the AI ​​voter.

[0020] This invention configures AI decision-making result sets and AI payoff items for AI decision-makers, thereby configuring AI decision-makers as game participants with independent decision-making capabilities and payoff objectives. This makes AI decision-makers and multi-level human decision-makers form equal game subjects, thus laying the groundwork for AI to output independent decision-making tendencies in the future.

[0021] The AI ​​investment decision-maker is an AI investment decision-making software system deployed on a computer. The AI ​​investment decision result set refers to the set of all possible investment decisions that the AI ​​investment decision-maker can output when making investment decisions. The AI ​​benefit term is a function used to calculate the AI ​​adoption rate and traceability validity. The AI ​​adoption rate is used to characterize the degree of recognition of the AI ​​investment decision-maker's AI investment decision results among human investment decision-makers at all levels. The traceability validity is used to characterize the understandability and verifiability of the AI ​​investment decision-maker's investment decision basis among human investment decision-makers at all levels.

[0022] In one embodiment of the present invention, configuring an AI voting result set and an AI benefit item for AI voters includes: establishing an AI voting result set from confirmation and rejection results; and establishing an AI benefit item from an AI adoption rate item and a traceability validity item. The AI ​​adoption rate item is the percentage of levels where the AI ​​voting result is consistent with the human voting result at each level out of the total number of human voting levels. The traceability validity item is the percentage of levels where a human voting result that is not nonexistent is made in response to the AI ​​voting result out of the total number of human voting levels.

[0023] It should be noted that when multiple levels of human voters issue a decision that is not "no" in response to the AI ​​voting result, it indicates that the multiple levels of human voters have either confirmed or rejected the decision. This is because the set of human voting results includes confirmed results, rejected results, and no result. No result indicates that the human voters cannot understand the voting basis of the AI ​​voting result or that the human voters believe that the voting basis of the AI ​​voting result is insufficient. In short, when multiple levels of human voters issue a decision of "no result," it indicates that there is a problem with the voting basis of the AI ​​voting result.

[0024] Furthermore, in this embodiment of the invention, by configuring a set of human decision-making results and human benefit items for multi-level human decision-makers, the multi-level human decision-makers are configured as game participants with independent decision-making capabilities and benefit objectives, thereby forming a game subject of equal level with AI decision-makers.

[0025] The aforementioned multi-level human decision-makers refer to multiple human decision-making entities arranged in a hierarchical order from low to high levels according to the decision-making and approval process. Low-level human decision-makers are located at the beginning of the approval chain, while high-level human decision-makers are located at the end of the approval chain. For example, the heads and employees of a large enterprise serve as decision-makers, corresponding to the high-level and low-level levels respectively in terms of hierarchical order.

[0026] In one embodiment of the present invention, configuring a set of human voting results and a human benefit item for multi-level human voters includes: establishing a set of human voting results from confirmed results, rejected results, and no results; and establishing a human benefit item from a human adoption rate item, wherein the human adoption rate item is the percentage of times that the human voting results at each level are consistent with the AI ​​voting results out of the total number of times each level participates in voting.

[0027] The human adoption rate is a function used to calculate the human adoption rate, which is used to characterize the degree of recognition of the human voting results of each level of human voters among AI voters.

[0028] It should be noted that in the manual acceptance rate item, each level will have a corresponding manual acceptance rate. For example, the manual acceptance rate for a lower level is the percentage of times that the manual voting result and the AI ​​voting result are consistent in the historical voting process of that level out of the total number of voting sessions at each level. In addition, when the manual voting result and the AI ​​voting result are consistent, there are two situations: consistent confirmation result and consistent rejection result. Therefore, there will be a confirmed acceptance rate and a rejection acceptance rate.

[0029] Furthermore, this embodiment of the invention constructs procedural constraints and threshold constraints for the AI ​​voter, enabling the AI ​​voter to automatically filter input strategies based on these constraints.

[0030] The procedural constraints are used to limit the legal execution path of the voting strategy in the voting process, and the threshold constraints are used to limit the legal numerical boundaries of the voting strategy in the voting process.

[0031] In one embodiment of the present invention, the construction of the procedural constraints and threshold constraints of the AI ​​voter includes: determining procedural constraints through the migration conditions between the voter node and adjacent nodes in the voting process of the AI ​​voter; and determining threshold constraints through the threshold range corresponding to the quantification index.

[0032] The investment decision-making process refers to a series of processes required for the AI ​​investment decision-maker to make investment decisions after receiving the investment decision strategy. The investment decision-making node refers to each node involved in the investment decision-making process, such as the material integrity check node, the quantitative indicator verification node, and the investment tendency output node. The material integrity check node is used by the AI ​​investment decision-maker to verify whether the investment decision strategy is complete; the quantitative indicator verification node is used by the AI ​​investment decision-maker to verify whether the quantitative indicators of the investment decision strategy are within legal numerical boundaries; and the investment tendency output node is used by the AI ​​investment decision-maker to generate its own investment tendency. Adjacent nodes refer to investment decision nodes adjacent to the investment decision node. For example, in the material integrity check process, the investment decision node is the project initiation basic data verification node, and the adjacent node is the feasibility study report document verification node. The migration condition refers to two adjacent investment decision nodes. Between these nodes, the system allows for the fulfillment of prerequisite judgment rules that must be met before a project can be transferred from the current decision-making node to the next adjacent node. For example, if the decision-making node is the project initiation basic data verification node and the adjacent node is the feasibility study report verification node, the migration conditions are that all project initiation basic materials, such as project approval, project site selection, and investment scale description, are uploaded completely and the files can be parsed normally. The quantitative indicators refer to quantifiable and numerical business parameters in the infrastructure investment project decision-making process, such as the total investment amount, capital ratio, and asset-liability ratio. The threshold range refers to the upper and lower limits of the legal range of the quantitative indicators. It should be noted that the lower limit of the threshold range is the conventional lower limit threshold uniformly implemented in the infrastructure investment industry and is also a common standard for investment decision review across the industry. The upper limit of the threshold range is the upper limit threshold set independently by large enterprise groups based on their own investment risk management requirements, capital costs, and business strategies.

[0033] S2. Analyze the historical adoption rate of multi-level human voters in the human voting result set through human benefit analysis, and output the candidate voting tendency of AI voters in the AI ​​voting result set based on procedural constraints and threshold constraints.

[0034] This invention analyzes the historical adoption rate of multi-level human voters within the human voting result set through human benefit analysis, in order to quantify the decision preferences of human voters at each level.

[0035] In one embodiment of the present invention, the step of analyzing the historical adoption rate of multi-level human voters in the human voting result set through human benefit items includes: calculating the historical adoption rate of each level of human voters in the human voting result set through human benefit items for each type of historical input feature; the historical adoption rate includes the confirmed adoption rate and the rejected adoption rate.

[0036] The historical input features refer to the different voting strategies received for the same project in historical periods before the current period. The current period starts from the execution of the AI ​​voter's candidate voting tendency in the AI ​​voting result set based on procedural constraints and threshold constraints.

[0037] It should be noted that after each type of historical input feature is input into the AI ​​voter, the AI ​​voter gives the voting result. Then, human voters at each level, from low to high, give their voting results. Therefore, each type of historical input feature will have a confirmed acceptance rate and a rejected acceptance rate at each level.

[0038] Furthermore, embodiments of the present invention output the candidate voting tendencies of AI voters within the AI ​​voting result set based on procedural constraints and threshold constraints, thereby filtering voting strategies within the procedural constraints and threshold constraints.

[0039] In one embodiment of the present invention, the step of outputting the candidate voting tendency of AI voters in the AI ​​voting result set based on procedural constraints and threshold constraints includes: for different voting strategies of the same project, screening a first target strategy that meets the procedural constraints and threshold constraints, and generating a candidate voting tendency for confirmation of the first target strategy; for different voting strategies of the same project, screening a second target strategy that does not meet the procedural constraints and threshold constraints, and generating a candidate voting tendency for rejection of the second target strategy.

[0040] The same project refers to a single investment decision review object, such as a single infrastructure investment, equity investment, or credit line project. The investment decision strategy is an investment review plan formulated for the project, including review materials and quantitative values. The review materials are used to pass the procedural constraint test, and the quantitative values ​​are used to pass the threshold constraint test. The candidate investment decision tendency is the investment decision tendency of each target strategy that is confirmed or rejected by the AI ​​investment decision-maker.

[0041] S3. Based on historical adoption rates, select AI voting tendencies that match the candidate voting tendencies, starting from the highest level of human voters and ending at the lowest level.

[0042] Based on historical adoption rates, this invention filters AI voting preferences that match candidate voting tendencies, starting from the highest-level human voter and ending at the lowest-level human voter. By reverse-engineering multi-level human voting preferences from top to bottom, it prioritizes the generation of voting schemes that are more likely to pass the step-by-step review, thereby reducing the situation where humans ignore, reject, or bypass AI suggestions.

[0043] In one embodiment of the present invention, the step of screening AI voting tendencies that match candidate voting tendencies based on historical adoption rates, starting from the highest-level human voter to the lowest-level human voter, includes: selecting the target adoption rate corresponding to the target strategy from the historical adoption rates based on the matching results between the target strategy and historical input features; based on the target adoption rate, progressively deducing the inferred voting tendencies of each level of human voter for the candidate voting tendencies, starting from the highest-level human voter to the lowest-level human voter; and when the inferred voting tendencies of all levels of human voter are consistent with the candidate voting tendencies, the candidate voting tendencies regarding the target strategy are taken as the AI ​​voting tendencies of the AI ​​voter.

[0044] The target strategy includes a first target strategy and a second target strategy, and the target adoption rate includes both the confirmed adoption rate and the rejected adoption rate.

[0045] Optionally, the matching process between the target strategy and historical input features based on the matching results is as follows: the review materials and values ​​included in the target strategy are compared with the review materials and values ​​corresponding to the historical input features in each historical period; when the similarity between the review materials and values ​​of the target strategy and the review materials and values ​​corresponding to a certain historical input feature is the largest, the target strategy is determined to match the historical input feature. It should be noted that when calculating the similarity between the two, the review materials and values ​​can be first vectorized, and then the similarity between the vectors can be calculated using existing conventional similarity algorithms.

[0046] It should be noted that the above process of screening AI voting tendencies that match candidate voting tendencies, based on historical adoption rates and starting from the highest-level human voter to the lowest-level human voter, can be simply described as follows: for each target strategy, find its corresponding target adoption rate, and deduce the inferred voting tendency. When the results given by each level of human voter in the inferred voting tendency are consistent with the candidate voting tendency of the AI ​​voter, it is considered that each level of human voter approves of the AI ​​voter.

[0047] In another embodiment of the present invention, the step of progressively deducing the inferred voting tendency of each level of human voter based on the target adoption rate, starting from the highest-level human voter and proceeding down to the lowest-level human voter, includes: determining the inferred voting tendency of the highest-level human voter based on the candidate voting tendency; starting from the inferred voting tendency of the highest-level human voter, taking the human voter result corresponding to the maximum adoption rate among the target adoption rates of the current-level human voter as the inferred voting tendency of the current-level human voter; if the inferred voting tendency of the current-level human voter is consistent with the corresponding candidate voting tendency, then continuing to pass it down to the next level; if the inferred voting tendency of the current-level human voter is inconsistent with the corresponding candidate voting tendency, then stopping the downward transmission, thereby realizing the process of progressively deducing the inferred voting tendency of each level of human voter based on the candidate voting tendency.

[0048] It should be noted that during the process of deducing the voting tendencies of each level of human voter regarding the candidate's voting tendency, the human voter will only give two possibilities: confirmation or rejection, and will not give no result. This is because this is the stage where the AI ​​voter is making deductions. It is assumed that the human voter at each level understands that the basis for the AI ​​voting result is sufficient, and the human voter believes that the basis for the AI ​​voting result is sufficient. Only when the AI ​​voting tendency is subsequently pushed to the human voter at each level and the human voter at each level makes the actual vote will there be a situation where no result occurs.

[0049] It should also be noted that if the predicted voting tendency of the current level human voter is inconsistent with the corresponding candidate voting tendency, it does not meet the screening requirement of "when the predicted voting tendency of all levels of human voters is consistent with the candidate voting tendency, the candidate voting tendency regarding the target strategy shall be taken as the AI ​​voting tendency of the AI ​​voter". Therefore, it stops passing down. However, in the actual human vote in the subsequent S4, regardless of whether the current level confirms or rejects, it continues to pass up to the next level.

[0050] For example, the process of determining the inferred voting tendency of the highest-level human voter based on the candidate voting tendency is as follows: if the candidate voting tendency is confirmed, initially, the inferred voting tendency of the highest-level human voter is also set to confirmed.

[0051] For example, starting from the inferred voting tendency of the highest-level human voter, the human vote result corresponding to the maximum adoption rate among the target adoption rates of the current-level human voter is taken as the inferred voting tendency of the current-level human voter: Assuming there are three levels in total: low-level, medium-level, and high-level, the highest-level human voter is confirmation, the current-level human voter is medium-level human voter, and the target adoption rate of the medium-level human voter includes the confirmation adoption rate and the rejection adoption rate. When the confirmation adoption rate is greater than the rejection adoption rate, confirmation will be taken as the inferred voting tendency of the current-level human voter; otherwise, rejection will be taken as the inferred voting tendency of the current-level human voter. If the confirmation adoption rate is equal to the rejection adoption rate, it is assumed to be consistent with the candidate voting tendency.

[0052] Reference Figure 2 The diagram shown is a schematic representation of an AI-powered intelligent decision-making optimization method according to an embodiment of the present invention. Figure 2 In the process, starting from the generation of candidate deductions by the highest-level human voter, the deduction results corresponding to the maximum adoption rate of the middle-level and lower-level human voters are verified in turn. The deduction ends if the deduction results of any level are inconsistent.

[0053] S4. Push the AI ​​voting tendency to human voters at each level to determine the current voting tendency of human voters at each level. Based on the current voting tendency, analyze the AI ​​adoption rate and traceability effectiveness of the AI ​​voting tendency through AI benefit items.

[0054] This invention, by analyzing the AI ​​adoption rate and traceability of AI voting preferences based on current voting preferences and AI benefit items, assesses the suitability and verifiability of AI voting preferences in multi-level voting review, providing data support for continuous optimization of AI voting decisions.

[0055] It should be noted that the process of pushing the AI's voting tendency to human voters at each level to determine the current voting tendency made by human voters at each level is different from the deduction in S3. The voting in S4 is actually carried out by humans, while S3 is a deduction performed by AI without human voting. Furthermore, the current voting tendency includes the voting tendency given at each level.

[0056] Reference Figure 3 The diagram shown illustrates a manual voting process within an AI-powered intelligent voting optimization method according to an embodiment of the present invention. Figure 3 In the process, the AI ​​voting tendency flows sequentially to lower, middle and senior human voters for review, with each level having the option to confirm, reject or not make a judgment.

[0057] Reference Figure 4 The image shown is a schematic diagram of the entire AI voting decision process according to an embodiment of the present invention. Figure 4 In the process, the voting data upload server inputs the voting strategy to the AI ​​voter, the database provides historical data support to the AI ​​voter, the AI ​​voter outputs the AI ​​voting tendency to the human voting server, and the human voting results are fed back to the procedural constraint and threshold constraint module, which further influences the AI ​​voter.

[0058] S5. Adjust the verification strictness of AI decision-makers regarding procedural constraints and threshold constraints based on AI adoption rate and traceability effectiveness to optimize AI intelligent decision-making.

[0059] The embodiments of the present invention adjust the verification strictness of AI voters regarding procedural constraints and threshold constraints based on the AI ​​adoption rate and traceability effectiveness, so as to update the voting strategy screening criteria of AI voters and further adapt to the preferences of multi-level human voting, thereby enabling effective interaction between humans and machines.

[0060] In one embodiment of the present invention, adjusting the verification strictness of procedural constraints and threshold constraints by AI voters based on AI adoption rate and traceability effectiveness includes: increasing the verification strictness of procedural constraints and threshold constraints when the AI ​​adoption rate is lower than a first preset threshold; decreasing the verification strictness of procedural constraints and threshold constraints when traceability effectiveness is lower than a second preset threshold; and maintaining the verification strictness of procedural constraints and threshold constraints when the AI ​​adoption rate is not lower than the first preset threshold and the traceability effectiveness is not lower than the second preset threshold.

[0061] It should be noted that the low adoption rate of AI indicates poor consistency between AI's decision-making results and those of human decision-making at various levels. Therefore, the stringency of verification needs to be increased to reduce misjudgments. For example, in the scenario described above, where "the decision-making node is the verification node for basic project approval documents, the adjacent node is the verification node for feasibility study reports, the migration conditions are the complete uploading of all basic project approval materials such as project approval documents, project site selection, and investment scale descriptions, and the documents can be parsed normally; quantitative indicators refer to quantifiable and numerical business parameters in the infrastructure investment project decision-making process, such as the total project investment, capital ratio, and asset-liability ratio; and the threshold range refers to the upper and lower limits of the legal range of quantitative indicators," the AI ​​decision-maker will only provide confirmation when all approval materials and quantitative data are confirmed.

[0062] In the above scenario, all basic project approval materials, including project initiation approval documents, project site selection, and investment scale descriptions, are available in both PDF and WORD versions. To improve the strictness of procedural constraint verification, versions that frequently contain garbled characters, such as the WORD version, need to be removed. Furthermore, to improve threshold constraints, the lower limit of the threshold constraints needs to be increased. Regarding the first preset threshold, it can be initially set at 50%, and then increased periodically. This is because, with the advancement of AI voting technology, the requirements for AI adoption rates are becoming increasingly stringent, thus requiring an increase in the first preset threshold over time. Finally, regarding the judgment criteria, when the AI ​​voting tendency is rejection or confirmation, the criteria used in verifying procedural and threshold constraints should be identified. For example, if the AI's voting tendency is to reject, while most human voters tend to approve, resulting in a low adoption rate for the AI's voting tendency, and it is discovered that the project approval document is missing from all the basic project approval materials, such as project initiation approval, project site selection, and investment scale description, which is why the AI's voting tendency is to reject, then the verification rigor of the project initiation approval document needs to be increased. On the other hand, if the AI's voting tendency is to approve, while most human voters tend to reject, then the verification rigor of all procedural constraints and threshold constraints needs to be increased. Low traceability effectiveness indicates that human voters at all levels generally cannot understand the AI's voting basis, so the verification rigor needs to be reduced to output more detailed judgment criteria for human review. Judgment criteria refer to the items for approval or rejection of each approval material and quantitative data in the voting strategy. For example, the migration conditions are: all basic project approval materials, such as project establishment approval, project site selection, and investment scale description, are uploaded completely and the files can be parsed normally: Confirmed; the total investment amount of the project is less than 10 million: Confirmed; only when all approval materials and quantitative data are confirmed will the AI ​​investment decision give a confirmed result.

[0063] like Figure 5 The diagram shown is a functional block diagram of an AI-powered intelligent investment decision optimization system according to the present invention.

[0064] The AI-powered intelligent investment decision optimization system 500 described in this invention can be installed in an electronic device. Depending on its functions, the AI-powered intelligent investment decision optimization system includes a constraint construction module 501, a candidate output module 502, a propensity screening module 503, an effectiveness analysis module 504, and an investment decision optimization module 505. The module described in this invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, stored in the memory of the electronic device.

[0065] In this embodiment of the invention, the functions of each module / unit are as follows: The constraint construction module 501 is used to configure AI voting result sets and AI benefit items for AI voting decision-makers, configure human voting result sets and human benefit items for multi-level human voting decision-makers, and construct the procedural constraints and threshold constraints of the AI ​​voting decision-makers. The candidate output module 502 is used to analyze the historical adoption rate of multi-level human voters in the human voting result set through human benefit items, and output the candidate voting tendency of AI voters in the AI ​​voting result set according to procedural constraints and threshold constraints. The tendency screening module 503 is used to screen AI voting tendencies that match candidate voting tendencies based on historical adoption rates, starting from the highest level of human voters and ending at the lowest level of human voters. The effectiveness analysis module 504 is used to push the AI ​​voting tendency to human voters at each level to determine the current voting tendency of human voters at each level. Based on the current voting tendency, the AI ​​adoption rate and traceability effectiveness of the AI ​​voting tendency are analyzed through AI benefit items. The voting optimization module 505 is used to adjust the verification strictness of AI voters regarding procedural constraints and threshold constraints based on the AI ​​adoption rate and traceability effectiveness, so as to optimize AI intelligent voting.

[0066] In detail, the modules in the AI ​​intelligent investment decision optimization system 500 described in this embodiment of the invention employ the same methods as described above during use. Figure 1 The method uses the same technical means as the AI ​​intelligent decision optimization method described above and can produce the same technical effect, so it will not be elaborated here.

[0067] In one embodiment, a computer device is provided, which may be a server or a client, and its internal structure diagram may be as follows: Figure 6 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it implements functions or steps on the server or client side of an AI intelligent decision-making optimization method.

[0068] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: Configure AI voting result sets and AI benefit items for AI voting decision-makers, configure human voting result sets and human benefit items for multi-level human voting decision-makers, and construct the procedural constraints and threshold constraints of the AI ​​voting decision-makers; By analyzing the historical adoption rate of multi-level human voters in the human voting result set through human benefit analysis, and outputting the candidate voting tendency of AI voters in the AI ​​voting result set based on procedural constraints and threshold constraints. Based on historical adoption rates, AI voting tendencies that match candidate voting tendencies are selected from the highest level of human voters to the lowest level of human voters. The AI ​​voting tendency is pushed to human voters at all levels to determine the current voting tendency of human voters at each level. Based on the current voting tendency, the adoption rate and traceability effectiveness of the AI ​​voting tendency are analyzed through AI benefit items. Adjust the stringency of AI voting decisions regarding procedural and threshold constraints based on AI adoption rates and traceability effectiveness to optimize AI-driven intelligent voting.

[0069] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: Configure AI voting result sets and AI benefit items for AI voting decision-makers, configure human voting result sets and human benefit items for multi-level human voting decision-makers, and construct the procedural constraints and threshold constraints of the AI ​​voting decision-makers; By analyzing the historical adoption rate of multi-level human voters in the human voting result set through human benefit analysis, and outputting the candidate voting tendency of AI voters in the AI ​​voting result set based on procedural constraints and threshold constraints. Based on historical adoption rates, AI voting tendencies that match candidate voting tendencies are selected from the highest level of human voters to the lowest level of human voters. The AI ​​voting tendency is pushed to human voters at all levels to determine the current voting tendency of human voters at each level. Based on the current voting tendency, the adoption rate and traceability effectiveness of the AI ​​voting tendency are analyzed through AI benefit items. Adjust the stringency of AI voting decisions regarding procedural and threshold constraints based on AI adoption rates and traceability effectiveness to optimize AI-driven intelligent voting.

[0070] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0071] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Furthermore, any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory.

[0072] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above.

[0073] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0074] Finally, it should be noted that in the above embodiments, each embodiment can be combined with each other or independent. Deleting any one of them will not affect the technical implementation of other embodiments. The above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An AI-powered intelligent decision-making optimization method, characterized in that, The method includes: Configure AI voting result sets and AI benefit items for AI voting decision-makers, configure human voting result sets and human benefit items for multi-level human voting decision-makers, and construct the procedural constraints and threshold constraints of the AI ​​voting decision-makers; By analyzing the historical adoption rate of multi-level human voters in the human voting result set through human benefit analysis, and outputting the candidate voting tendency of AI voters in the AI ​​voting result set based on procedural constraints and threshold constraints. Based on historical adoption rates, AI voting tendencies that match candidate voting tendencies are selected from the highest level of human voters to the lowest level of human voters. The AI ​​voting tendency is pushed to human voters at all levels to determine the current voting tendency of human voters at each level. Based on the current voting tendency, the adoption rate and traceability effectiveness of the AI ​​voting tendency are analyzed through AI benefit items. Adjust the stringency of AI voting decisions regarding procedural and threshold constraints based on AI adoption rates and traceability effectiveness to optimize AI-driven intelligent voting.

2. The AI ​​intelligent decision optimization method as described in claim 1, characterized in that, Based on historical adoption rates, AI voting tendencies that match candidate voting tendencies are selected from the highest-level human voter to the lowest-level human voter, including: Based on the matching results between the target strategy and historical input features, the target adoption rate corresponding to the target strategy is selected from the historical adoption rates; Based on the target adoption rate, starting from the highest level of human voter and going down to the lowest level of human voter, the deduced voting tendency of each level of human voter regarding the candidate voting tendency is deduced step by step. When the predicted voting tendencies of all levels of human voters are consistent with the candidate voting tendencies, the candidate voting tendencies regarding the target strategy are taken as the AI ​​voting tendencies of the AI ​​voters.

3. The AI ​​intelligent decision optimization method as described in claim 2, characterized in that, Based on the target adoption rate, starting from the highest-level human voter and going down to the lowest-level human voter, the deduced voting tendency of each level of human voter regarding the candidate's voting tendency is derived step by step, including: The projected voting tendencies of the highest-level human voters are determined based on the voting tendencies of the candidates. Starting from the predicted voting tendency of the highest-level human voter, the human vote result corresponding to the highest adoption rate among the target adoption rates of the current-level human voter is taken as the predicted voting tendency of the current-level human voter. If the predicted voting tendency of the current level of human voters is consistent with the voting tendency of the corresponding candidate, then it continues to be passed down to the next level; If the predicted voting tendency of the current level of human voter is inconsistent with the corresponding candidate voting tendency, the transmission to the next level will stop, so as to realize the process of reversing the predicted voting tendency of each level of human voter for the candidate voting tendency.

4. The AI ​​intelligent decision optimization method as described in claim 1, characterized in that, Configure AI investment decision-makers with AI investment decision result sets and AI benefit items, including: An AI voting result set is established from the confirmation and rejection results; An AI benefit item is established by combining the AI ​​adoption rate item and the traceability validity item. The AI ​​adoption rate item is the percentage of levels where the AI ​​voting results are consistent with the human voting results at each level out of the total number of human voting levels. The traceability validity item is the percentage of levels where the human voting results are not nonexistent in response to the AI ​​voting results out of the total number of human voting levels.

5. The AI ​​intelligent decision optimization method as described in claim 1, characterized in that, Configure a set of human voting results and human benefit items for multi-level human voting participants, including: A set of manual voting results is established based on the confirmed results, rejected results, and no results; A human benefit item is established based on the human adoption rate item, which is the percentage of times that the human voting results at each level are consistent with the AI ​​voting results out of the total number of times voting is participated in at each level.

6. The AI ​​intelligent decision optimization method as described in claim 1, characterized in that, Constructing the procedural constraints and threshold constraints for the AI ​​voter includes: In the voting process of the AI ​​voter, procedural constraints are determined by the migration conditions between the voting node and its adjacent nodes; Threshold constraints are determined by quantifying the threshold range corresponding to the indicators.

7. The AI ​​intelligent decision optimization method as described in claim 1, characterized in that, Analysis of the historical adoption rate of multi-level human voters within the human voting result set through human benefit item analysis includes: For each type of historical input feature, the historical adoption rate of each level of human voter in the human vote result set is calculated through the human benefit item; The historical adoption rate includes the confirmed adoption rate and the rejected adoption rate.

8. The AI ​​intelligent decision optimization method as described in claim 1, characterized in that, Based on procedural constraints and threshold constraints, the AI ​​voter's candidate voting tendency within the AI ​​voting result set is output, including: For different investment decisions on the same project, select target strategies that meet procedural constraints and threshold constraints; Generate candidate voting tendencies that are confirmed at each level regarding the target strategy.

9. The AI ​​intelligent decision optimization method as described in claim 1, characterized in that, Adjust the stringency of AI voter verification regarding procedural and threshold constraints based on AI adoption rates and traceability effectiveness, including: When the AI ​​adoption rate is lower than the first preset threshold, the verification strictness of procedural constraints and threshold constraints is increased. When the traceability validity is lower than the second preset threshold, the verification strictness of procedural constraints and threshold constraints is reduced; When the AI ​​adoption rate is not lower than the first preset threshold and the traceability effectiveness is not lower than the second preset threshold, the strictness of the verification of procedural constraints and threshold constraints is maintained.

10. An AI-powered intelligent investment decision optimization system, characterized in that, The system includes: The constraint construction module is used to configure the AI ​​voting result set and AI benefit item for AI voting decision-makers, configure the human voting result set and human benefit item for multi-level human voting decision-makers, and construct the procedural constraints and threshold constraints of the AI ​​voting decision-makers; The candidate output module is used to analyze the historical adoption rate of multi-level human voters in the human voting result set through human benefit items, and output the candidate voting tendency of AI voters in the AI ​​voting result set based on procedural constraints and threshold constraints. The bias screening module is used to screen AI voting biases that match candidate voting biases, based on historical adoption rates, starting from the highest level of human voters and ending at the lowest level of human voters. The effectiveness analysis module is used to push the AI ​​voting tendency to human voters at each level to determine the current voting tendency of human voters at each level. Based on the current voting tendency, the AI ​​adoption rate and traceability effectiveness of the AI ​​voting tendency are analyzed through AI benefit items. The voting optimization module is used to adjust the strictness of AI voting on procedural and threshold constraints based on the AI ​​adoption rate and traceability effectiveness, so as to optimize AI intelligent voting.

Images