Business operation strategy determination method and device for business system, and electronic equipment

Abstract

The application discloses a kind of business operation strategy determination method, device and electronic equipment of business system.Therein, the method includes: obtaining the business operation data corresponding to business operation demand in business system;From business operation data, extract multiple business variables;Initial causal diagram is constructed;Based on business operation data, determine the retention index corresponding to multiple causal edges in initial causal diagram respectively;Based on the retention index corresponding to multiple causal edges respectively, initial causal diagram is adjusted, and target causal diagram is obtained;Based on business operation demand and target causal diagram, determine the target operation strategy of business system.The application solves the technical problem that business operation strategy is inaccurate when determining the business operation strategy of business system in related technology.

Description

Technical Field

[0001] This invention relates to the field of data processing, and more specifically, to a method, apparatus, and electronic device for determining business operation strategies for a business system. Background Technology

[0002] In related technologies, corresponding operational strategies need to be formulated based on business operation requirements to ensure the efficient, stable, and sustainable operation of business systems. However, in determining the business operation strategies of business systems, there are technical problems such as inaccurate business operation strategies, leading to imbalances in resource scheduling, decreased operational efficiency, and even business anomalies.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This invention provides a method, apparatus, and electronic device for determining the business operation strategy of a business system, in order to at least solve the technical problem in the related art that the business operation strategy is inaccurate when determining the business operation strategy of a business system.

[0005] According to one aspect of the present invention, a method for determining a business operation strategy for a business system is provided, comprising: acquiring business operation data corresponding to business operation requirements in the business system; extracting multiple business variables from the business operation data, wherein the multiple business variables are variables whose representation index of the business operation requirements is greater than a representation threshold; constructing an initial causal graph, wherein the initial causal graph includes nodes corresponding to the multiple business variables and causal edges formed by causal relationships between the multiple business variables; determining retention indices corresponding to the multiple causal edges in the initial causal graph based on the business operation data; adjusting the initial causal graph based on the retention indices corresponding to the multiple causal edges to obtain a target causal graph; and determining a target operation strategy for the business system based on the business operation requirements and the target causal graph.

[0006] Optionally, determining the retention indices corresponding to the multiple causal edges in the initial causal graph based on the business operation data includes: when the business operation data includes business description text, runtime sequence data, and execution constraint data, determining the text confidence level corresponding to each of the multiple causal edges based on the business description text, wherein the text confidence level is used to characterize the textual support strength of the business description text for the causal relationship corresponding to the corresponding causal edge; determining the temporal confidence level corresponding to each of the multiple causal edges based on the runtime sequence data, wherein the temporal confidence level is used to characterize the temporal support strength of the runtime sequence data for the causal relationship corresponding to the corresponding causal edge; determining the constraint confidence level corresponding to each of the multiple causal edges based on the execution constraint data, wherein the constraint confidence level is used to characterize the constraint support strength of the execution constraint data for the causal relationship corresponding to the corresponding causal edge; and determining the retention indices corresponding to the multiple causal edges in the initial causal graph based on the text confidence level, temporal confidence level, and constraint confidence level corresponding to each of the multiple causal edges.

[0007] Optionally, for any one of the plurality of causal edges, the retention index corresponding to the causal edge is determined in the following manner: a confidence variance is determined based on the text confidence, temporal confidence, and constraint confidence corresponding to the causal edge; a causal conflict index is determined between the text confidence and the constraint confidence, wherein the causal conflict index is used to characterize the degree of conflict between the causal direction represented by the text confidence and the causal direction represented by the constraint confidence; a confidence penalty index is determined based on the confidence variance and the causal conflict index; and a retention index is determined based on the text confidence, temporal confidence, constraint confidence, and confidence penalty index corresponding to the causal edge.

[0008] Optionally, determining the time-series confidence level corresponding to each of the plurality of causal edges based on the runtime sequence data includes: determining the time-lag response strength and historical response strength corresponding to each of the plurality of causal edges based on the runtime sequence data; and determining the time-series confidence level corresponding to each of the plurality of causal edges based on the time-lag response strength and historical response strength corresponding to each of the plurality of causal edges.

[0009] Optionally, determining the target operation strategy of the business system based on the business operation requirements and the target causal graph includes: determining multiple initial operation strategies based on the business operation requirements and the target causal graph; determining the business output increment, execution cost index, operation risk index, and execution difficulty coefficient corresponding to the multiple initial operation strategies; determining the fusion selection index corresponding to the multiple initial operation strategies based on the business output increment, execution cost index, operation risk index, and execution difficulty coefficient corresponding to the multiple initial operation strategies; and determining the target operation strategy from the multiple initial operation strategies based on the fusion selection index corresponding to the multiple initial operation strategies.

[0010] Optionally, for any one of the plurality of initial operating strategies, the business output increment corresponding to the initial operating strategy is determined in the following manner: From the plurality of business variables, a dependent variable corresponding to the initial operating strategy and an effect variable corresponding to the dependent variable are determined; based on the dependent variable and the effect variable, an interference variable is determined from a plurality of residual variables, wherein the plurality of residual variables are other business variables besides the dependent variable and the effect variable, and the interference variable is a business variable that has a causal relationship with both the dependent variable and the effect variable; the difference between the business output under the initial operating strategy and the business output under a reference operating strategy is determined when the variable parameter corresponding to the interference variable is a predetermined parameter, and the difference in output is taken as the business output increment, wherein the reference operating strategy is the operating strategy when the business operation corresponding to the dependent variable is not executed.

[0011] Optionally, after determining the target operation strategy of the business system based on the business operation requirements and the target cause-effect graph, the process includes: determining a strategy output template, wherein the strategy output template includes a requirement description area, an operation constraint area, and an execution content area, wherein the operation constraints in the operation constraint area are used to constrain the execution content in the execution content area; and outputting the target operation strategy according to the strategy output template.

[0012] According to one aspect of the present invention, a business operation strategy determination apparatus for a business system is provided, comprising: a first determining module, configured to acquire business operation data corresponding to business operation requirements in the business system; a second determining module, configured to extract multiple business variables from the business operation data, wherein the multiple business variables are variables whose representation index of the business operation requirements is greater than a representation threshold; a third determining module, configured to construct an initial causal graph, wherein the initial causal graph includes nodes corresponding to the multiple business variables and causal edges formed by causal relationships between the multiple business variables; a fourth determining module, configured to determine retention indices corresponding to the multiple causal edges in the initial causal graph based on the business operation data; a fifth determining module, configured to adjust the initial causal graph based on the retention indices corresponding to the multiple causal edges to obtain a target causal graph; and a sixth determining module, configured to determine a target operation strategy for the business system based on the business operation requirements and the target causal graph.

[0013] According to one aspect of the present invention, an electronic device is provided, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to execute the instructions to implement the business operation strategy determination method of the business system described in any of the preceding claims.

[0014] According to one aspect of the present invention, a computer-readable storage medium is provided, wherein when the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform the business operation strategy determination method of the business system described in any of the preceding claims.

[0015] In this embodiment of the invention, by first acquiring business operation data corresponding to business operation requirements in the business system, and extracting multiple business variables whose representation index of business operation requirements is greater than a representation threshold from the business operation data, it is possible to focus on the key variables most relevant to the current requirements and avoid interference from irrelevant variables on subsequent causal modeling. Then, based on multiple business variables, an initial causal graph containing nodes and causal edges is constructed, which can explicitly structure the causal relationships between variables, providing a basis for the reliability assessment of subsequent causal edges. Finally, based on the business operation data, the retention index corresponding to multiple causal edges in the initial causal graph is determined, which can quantify the credibility of each causal edge and avoid introducing unreliable causal relationships based solely on experience or subjective judgment. The initial causal graph is then adjusted based on the retention indices corresponding to multiple causal edges to obtain the target causal graph. This process removes unreliable causal edges and retains highly reliable ones, ensuring that the target causal graph accurately reflects the causal structure between variables in the business system. Finally, the target operational strategy is determined based on the business operation requirements and the target causal graph. Since the target causal graph eliminates unreliable causal relationships, the strategy derived from it has a causal basis, avoiding decision-making biases caused by causal misjudgments or variable omissions during strategy formulation. This achieves accurate determination of the business operation strategy and solves the technical problem of inaccurate business operation strategies in related technologies. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0017] Figure 1 This is a flowchart of a method for determining the business operation strategy of a business system according to an embodiment of the present invention;

[0018] Figure 2 This is a schematic diagram of the business operation strategy determination process of the business system in an optional embodiment of the present invention;

[0019] Figure 3 This is a schematic diagram of the business operation strategy determination process framework of the business system in an optional embodiment of the present invention;

[0020] Figure 4 This is a structural block diagram of a business operation strategy determination device for a business system according to an embodiment of the present invention. Detailed Implementation

[0021] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0022] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0023] Example 1

[0024] According to an embodiment of the present invention, an embodiment of a method for determining the business operation strategy of a business system is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0025] Figure 1 This is a flowchart of a business operation strategy determination method for a business system according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0026] S102, Obtain business operation data from the business system that corresponds to the business operation requirements;

[0027] This involves business systems, which are comprehensive operating systems that carry out and execute corresponding business activities. For example, these business systems include various business activities such as resource scheduling, task allocation, and production execution.

[0028] This involves business operation requirements, which are descriptions of the tasks, objectives, or problems faced by the business system in the current operating environment. For example, these business operation requirements include improving output efficiency, reducing operating costs, and optimizing resource allocation.

[0029] This includes business operation data, which consists of various types of data generated by the business system during its actual operation, including structured and unstructured data. For example, business operation data includes order data, inventory data, production data, supply chain data, financial data, user operation data, equipment status data, operation reports, and maintenance records.

[0030] S104. Extract multiple business variables from business operation data. Among them, multiple business variables are variables whose representation index of business operation requirements is greater than the representation threshold.

[0031] This involves business variables, which are quantifiable or categorizable business elements extracted from business operation data. For example, these business variables include output conversion rate, resource utilization rate, response latency, task completion rate, inventory turnover rate, order fulfillment rate, etc. Multiple business variables are used to characterize the business operation status in different dimensions.

[0032] This involves a representation index, which is used to measure the degree to which business variables reflect business operation requirements. For example, the representation index includes correlation coefficient, mutual information value, feature importance score, semantic similarity, etc., and is used to quantitatively assess the correlation strength between each business variable and business operation requirements.

[0033] This involves a representation threshold, which is a preset critical value used to filter business variables. For example, the representation threshold can be set to a correlation coefficient of not less than 0.3, a mutual information value of greater than 0.5, or a feature importance score ranking in the top 10%, etc., to determine whether a business variable is sufficient to represent the business operation requirements. Only variables with a representation index greater than the representation threshold are extracted as business variables for subsequent use.

[0034] S106, Construct an initial causal graph, wherein the initial causal graph includes nodes corresponding to multiple business variables, and causal edges formed by the causal relationships between multiple business variables;

[0035] This involves an initial causal graph, which is a directed graph model built based on business variables and their causal relationships. For example, the initial causal graph includes nodes and causal edges, which are used to explicitly and structurally express the causal structure between variables in the business system.

[0036] This involves nodes, which are graphical elements used to represent business variables in the initial causal graph. For example, each node corresponds to a business variable, and its main function is to map business variables to vertices in the graph structure, making it easier to analyze the causal relationships between variables using graph theory methods.

[0037] This involves causal relationships, which are "cause-effect" dependencies between two business variables. For example, such causal relationships include resource utilization affecting output conversion rate, response delay affecting task completion rate, and inventory turnover affecting order fulfillment rate. These causal relationships are used to describe the direction of action and the path of influence between variables in the business system and are the semantic basis for constituting causal edges.

[0038] This involves causal edges, which are directed lines that connect two nodes in the initial causal graph and represent their causal relationship. For example, the causal edge points from the cause variable node to the result variable node, expressing the causal relationship between business variables in the form of directed edges in the graph structure.

[0039] S108, Based on business operation data, determine the retention index corresponding to multiple causal edges in the initial causal graph;

[0040] This involves a retention index, which measures the degree to which each causal edge in the initial causal graph should be retained. For example, the higher the value of the retention index, the higher the credibility of the causal edge and the more it should be retained in the causal graph. Its main function is to provide a quantitative basis for adjusting the initial causal graph, so that causal edges with high credibility are retained and causal edges with low credibility are removed, thereby obtaining a target causal graph that more accurately reflects the causal structure of the business system.

[0041] S110, Based on the retention indices corresponding to multiple causal edges, the initial causal graph is adjusted to obtain the target causal graph;

[0042] This involves a target causal graph, which is a directed graph obtained by adjusting the initial causal graph. For example, the target causal graph retains the causal edges with higher retention indices from the initial causal graph and removes the causal edges with lower retention indices. Its main function is to provide a causal structure that has been filtered for reliability, ensuring that the causal graph used for subsequent strategy determination can truly reflect the causal relationships between variables in the business system and avoid policy deviations caused by unreliable causal edges.

[0043] S112, based on business operation requirements and the target cause-effect graph, determine the target operation strategy of the business system.

[0044] This involves target operation strategies, which are specific strategies determined based on business operation requirements and target cause-effect graphs to guide adjustments to the operation of the business system. For example, target operation strategies include improving resource utilization, optimizing task scheduling strategies, adjusting output conversion processes, and reducing the occupation of non-core tasks. Their main function is to provide executable operation instructions to the business system to specifically solve the problems described by business operation requirements and improve the effectiveness of business operations.

[0045] Through the steps S102-S112 described above, by first acquiring business operation data corresponding to business operation requirements from the business system, and then extracting multiple business variables whose representation indices for business operation requirements are greater than a representation threshold from the business operation data, we can focus on the key variables most relevant to the current requirements and avoid interference from irrelevant variables in subsequent causal modeling. Next, based on these multiple business variables, we construct an initial causal graph containing nodes and causal edges, which explicitly structures the causal relationships between variables, providing a foundation for the reliability assessment of subsequent causal edges. Finally, based on the business operation data, we determine the retention indices corresponding to multiple causal edges in the initial causal graph, which quantifies the credibility of each causal edge and avoids introducing unreliable information based solely on experience or subjective judgment. The initial causal graph is then adjusted based on the retention indices corresponding to multiple causal edges to obtain the target causal graph. This process removes unreliable causal edges and retains highly reliable ones, ensuring that the target causal graph accurately reflects the causal structure between variables in the business system. Finally, the target operational strategy is determined based on business operation requirements and the target causal graph. Since the target causal graph eliminates unreliable causal relationships, the strategy derived from it has a causal basis, avoiding decision-making biases caused by causal misjudgments or variable omissions during strategy formulation. This achieves accurate determination of the business operation strategy and solves the technical problem of inaccurate business operation strategies in related technologies.

[0046] As an optional embodiment, based on business operation data, the retention indices corresponding to multiple causal edges in the initial causal graph are determined, including: when the business operation data includes business description text, runtime sequence data, and execution constraint data, determining the text confidence scores corresponding to multiple causal edges based on the business description text, wherein the text confidence scores characterize the textual support strength of the business description text for the causal relationships corresponding to the corresponding causal edges; determining the temporal confidence scores corresponding to multiple causal edges based on the runtime sequence data, wherein the temporal confidence scores characterize the temporal support strength of the runtime sequence data for the causal relationships corresponding to the corresponding causal edges; determining the constraint confidence scores corresponding to multiple causal edges based on the execution constraint data, wherein the constraint confidence scores characterize the constraint support strength of the execution constraint data for the causal relationships corresponding to the corresponding causal edges; and determining the retention indices corresponding to multiple causal edges in the initial causal graph based on the text confidence scores, temporal confidence scores, and constraint confidence scores corresponding to multiple causal edges.

[0047] This includes business description text, which is descriptive information related to the business recorded in natural language in the business operation data. For example, the business description text includes operation reports, operation and maintenance records, business analysis documents, etc.

[0048] This includes runtime sequence data, which is structured indicator data recorded in chronological order from the business operation data. For example, the runtime sequence data includes hourly order volume, daily inventory level, weekly output conversion rate, monthly resource utilization rate, etc.

[0049] This involves execution constraint data, which are constraints in business operation data used to limit the feasible scope of business operations. For example, execution constraint data includes cost limits, inventory capacity constraints, production capacity limits, equipment configuration limits, task priority rules, etc.

[0050] This involves text confidence, which measures the strength of textual support for the causal relationship corresponding to the causal edge in the business description text. For example, the higher the text confidence value, the more complete and consistent the description of the corresponding causal relationship in the business description text. Its main function is to provide a credibility assessment from the text dimension for the calculation of the retention index.

[0051] This involves time series confidence, which measures the strength of time series support for the causal relationship corresponding to the causal edge in the runtime data. For example, a higher time series confidence value indicates that the causal variable changes first and the outcome variable responds later and the response is significant in the runtime data. Its main function is to provide a credibility assessment from the data dimension for the calculation of the retention index.

[0052] This involves constraint confidence, which measures the strength of the constraint support provided by the execution constraint data for the causal relationship corresponding to the causal edge. For example, the higher the constraint confidence value, the more the causal relationship represented by the causal edge conforms to the business rules and logical constraints. Its main function is to provide a credibility assessment from the rule dimension for the calculation of the retention index.

[0053] The text confidence score, time-series confidence score, and constraint confidence score can be fused using fixed weights or updated using an adaptive weighting method based on historical validation results. The time-series confidence score can be calculated using conditional independence tests, causal testing methods, structural equation modeling, causal discovery algorithms, historical intervention experiments, or quasi-experimental methods.

[0054] In this optional embodiment, the text confidence level corresponding to each causal edge is first determined based on the business description text, which can quantify the strength of the text's support for the causal relationship from unstructured text; then, the temporal confidence level corresponding to each causal edge is determined based on runtime sequence data, which can quantify the strength of the data's temporal support for the causal relationship from temporal changes; next, the constraint confidence level corresponding to each causal edge is determined based on execution constraint data, which can quantify the strength of the constraints' support for the causal relationship from the business rule level; finally, the retention index corresponding to each causal edge is determined by integrating the confidence levels of the above three dimensions. Since the information of the three dimensions of text, time sequence and rules is integrated, the credibility of each causal edge can be comprehensively evaluated from multi-source data, avoiding evaluation bias caused by relying on a single data source, thereby ensuring that the retention index can truly reflect the reliability of the causal edge, and providing a quantitative basis for subsequent accurate screening of credible causal edges.

[0055] As an optional embodiment, for any one of the multiple causal edges, the retention index corresponding to any causal edge is determined in the following manner: The confidence variance is determined based on the text confidence, temporal confidence, and constraint confidence corresponding to any causal edge; a causal conflict index is determined between the text confidence and the constraint confidence, wherein the causal conflict index is used to characterize the degree of conflict between the causal direction represented by the text confidence and the causal direction represented by the constraint confidence; a confidence penalty index is determined based on the confidence variance and the causal conflict index; and a retention index is determined based on the text confidence, temporal confidence, constraint confidence, and confidence penalty index corresponding to any causal edge.

[0056] This involves confidence variance, which is a quantitative value of the degree of dispersion calculated based on the text confidence, time series confidence, and constraint confidence corresponding to the causal edge. For example, the larger the confidence variance, the greater the discrepancy and the worse the consistency among the confidence of the three dimensions. Its main function is to provide an assessment basis for determining the confidence penalty index from the consistency of multi-source confidence.

[0057] This includes a causal conflict index, which is used to characterize the degree of conflict between the causal direction represented by the text confidence level and the causal direction represented by the constraint confidence level. For example, the higher the value of the causal conflict index, the more serious the conflict between the causal direction described in the text and the causal direction defined by the business rules. Its main function is to provide an assessment basis for the consistency of causal direction in determining the confidence penalty index.

[0058] This involves a confidence penalty index, which is a quantitative indicator used to penalize the credibility of causal edges. For example, the higher the value of the confidence penalty index, the greater the discrepancy and the more serious the conflict between the text confidence and the constraint confidence, and the greater the deduction should be made to the credibility of the causal edge. This index is used to determine the retention index and provide input for the penalty term.

[0059] Determine the causal conflict index between text confidence and constraint confidence, example:

[0060] Extract the causal direction corresponding to the text confidence and the causal direction corresponding to the constraint confidence respectively; compare whether the causal direction corresponding to the text confidence and the causal direction corresponding to the constraint confidence are consistent. If they are consistent, the causal conflict index is 0; if they are inconsistent, the causal conflict index is 0.

[0061] Among them, the causal direction of text confidence is extracted from the business description text, that is, the direction in which the cause variable described in the text points to the result variable. The "cause → effect" relationship direction in the text is identified through dependency parsing or semantic role labeling in natural language processing.

[0062] Among them, the causal direction of constraint confidence is the causal direction derived from the execution constraint data, that is, the allowed direction of interaction between variables as defined by business rules or logical constraints. Through rule parsing or logical reasoning, the order or dependency direction of variables implicit in the constraints is determined.

[0063] In this optional embodiment, the confidence variance is first determined based on text confidence, time-series confidence, and constraint confidence. This quantifies the dispersion among the three dimensions of confidence. Greater dispersion indicates greater differences in the support strength of different data sources for the same causal edge and poorer consistency among multi-source information. Next, a causal conflict index is determined between text confidence and constraint confidence, quantifying the degree of conflict between the causal direction represented by the text and the causal direction represented by the constraints. Finally, a confidence penalty index is determined based on the confidence variance and the causal conflict index. Since a larger confidence variance indicates greater consistency among multi-source confidence... The greater the divergence and the higher the causal conflict index, the more severe the conflict between the text direction and the rule direction. This allows the confidence penalty index to accurately reflect the degree of inconsistency of causal edges in multi-source information. Finally, the text confidence, temporal confidence, constraint confidence, and confidence penalty index are integrated to determine the retention index. Since the penalty term can deduct the credibility of causal edges with inconsistencies in multi-source information, it ensures that causal edges with divergence or conflict between multi-source information receive a lower retention index. This avoids unreliable causal edges being mistakenly retained due to inflated confidence in a single dimension, thus providing a reliable quantitative basis for the accurate adjustment of the causal graph in the future.

[0064] As an optional embodiment, determining the time-series confidence level corresponding to each of the multiple causal edges based on runtime sequence data includes: determining the time-lag response strength and historical response strength corresponding to each of the multiple causal edges based on runtime sequence data; and determining the time-series confidence level corresponding to each of the multiple causal edges based on the time-lag response strength and historical response strength corresponding to each of the multiple causal edges.

[0065] This includes the time lag response strength, which measures the degree of response of the result variable under time lag conditions after a change in the causal variable. For example, the higher the value of the time lag response strength, the more significant the change in the result variable is after a certain time lag following the change in the causal variable.

[0066] This includes historical response strength, which measures the degree of response when the causal variable changes and the result variable changes accordingly in historical operational data. For example, a higher value of historical response strength indicates a closer response relationship between changes in the causal variable and changes in the result variable during historical business operations.

[0067] In this optional embodiment, the time-lag response strength corresponding to each causal edge is first determined based on runtime sequence data. This quantifies the response degree of the result variable after a certain time lag following a change in the causal variable. Then, the historical response strength corresponding to each causal edge is determined based on runtime sequence data. This quantifies the closeness of the response between the causal variable and the result variable during historical operation. Since the time-lag response strength reflects the time-lag characteristics of the causal effect and the historical response strength reflects the stability of the causal effect in historical data, integrating the two can comprehensively evaluate the support of runtime sequence data for causal relationships from two dimensions: time lag and historical response. This determines the time-series confidence level corresponding to each causal edge, ensuring that the time-series confidence level can truly reflect the credibility of the causal edge in the time-series dimension. This avoids evaluation bias caused by relying on only a single time-series indicator and provides a reliable basis from the data dimension for the subsequent calculation of the retention index.

[0068] As an optional implementation, based on business operation requirements and the target cause-effect graph, the target operation strategy of the business system is determined, including: determining multiple initial operation strategies based on business operation requirements and the target cause-effect graph; determining the business output increment, execution cost index, operation risk index, and execution difficulty coefficient corresponding to each of the multiple initial operation strategies; determining the fusion selection index corresponding to each of the multiple initial operation strategies based on the business output increment, execution cost index, operation risk index, and execution difficulty coefficient corresponding to each of the multiple initial operation strategies; and determining the target operation strategy from the multiple initial operation strategies based on the fusion selection index corresponding to each of the multiple initial operation strategies.

[0069] This involves an initial operational strategy, which is a preliminary operational strategy generated based on business operational needs and the target cause-effect graph. For example, the initial operational strategy includes improving resource utilization, optimizing task scheduling strategies, adjusting output conversion processes, and reducing the occupation of non-core tasks.

[0070] This involves incremental business output, which is the increase in business output after implementing the initial operating strategy compared to not implementing the strategy. For example, this incremental business output can be expressed as the increase in output conversion rate, the increase in product output, the increase in order fulfillment rate, etc.

[0071] This includes the execution cost index, which is a quantitative indicator used to measure the resource costs required to execute the initial operation strategy. For example, the higher the value of the execution cost index, the greater the resource, time, and other resource input required to execute the strategy.

[0072] This includes an operational risk index, which is a quantitative indicator used to measure the operational risks to the business system that may be caused by the execution of the initial operational strategy. For example, the higher the value of the operational risk index, the greater the probability or severity of risk events such as decreased system stability, increased response latency, and excessive resource consumption that may occur when the strategy is executed. Its main function is to provide an assessment basis from the risk dimension for determining the integration selection index.

[0073] This includes an execution difficulty coefficient, which is a quantitative indicator used to measure how easy it is to actually implement an initial running strategy in a business system. For example, a higher execution difficulty coefficient indicates that the strategy involves more complex operations, a longer execution cycle, and greater modifications to existing processes.

[0074] This includes the integration selection index, a quantitative indicator used to measure the overall priority of each initial operating strategy. For example, a higher value of the integration selection index indicates that the strategy performs better in terms of the four dimensions of business output increment, execution cost index, operating risk index, and execution difficulty coefficient. This provides a comprehensive quantitative basis for selecting the target operating strategy from multiple initial operating strategies, ensuring that the final selected strategy achieves the optimal balance between business output, cost, risk, and executability.

[0075] In this embodiment, multiple initial operational strategies are first determined based on business operation requirements and the target causal graph, ensuring that the generated initial operational strategies are both demand-driven and supported by a causal structure. Next, the business output increment, execution cost index, operational risk index, and execution difficulty coefficient corresponding to each initial operational strategy are determined, quantifying the comprehensive performance of each strategy from four dimensions: output, cost, risk, and executability. Then, based on the above four indicators, a fusion selection index corresponding to each initial operational strategy is determined. Since the fusion selection index integrates information from the four dimensions of output, cost, risk, and executability, it can comprehensively reflect the overall priority of each strategy. Finally, based on the fusion selection index, the target operational strategy is determined from the multiple initial operational strategies, ensuring that the selected strategy achieves an optimal balance between output, cost, risk, and executability, avoiding strategy selection bias caused by relying on only a single dimension for evaluation, thereby providing the business system with the operational strategy with the best overall performance.

[0076] As an optional embodiment, for any one of multiple initial operating strategies, the business output increment corresponding to any one initial operating strategy is determined in the following way: From multiple business variables, the dependent variable corresponding to any one initial operating strategy and the effect variable corresponding to the dependent variable are determined; based on the dependent variable and the effect variable, interference variables are determined from multiple residual variables, wherein the multiple residual variables are other business variables besides the dependent variable and the effect variable, and the interference variables are business variables that have a causal relationship with both the dependent variable and the effect variable; the output difference between the business output under the initial operating strategy and the business output under the reference operating strategy is determined when the variable parameter corresponding to the interference variable is a predetermined parameter, and the output difference is used as the business output increment, wherein the reference operating strategy is the operating strategy when the business operation corresponding to the dependent variable is not executed.

[0077] This involves a dependent variable, which is the causal variable of the initial operating strategy. That is, the business variable that is actively adjusted or intervened in during the execution of the strategy. For example, if the initial operating strategy is to improve resource utilization, the dependent variable is resource utilization; if the strategy is to adjust the output conversion process, the dependent variable is process parameters. Its main function is to map the strategy operation to specific nodes in the causal graph, serving as the intervention object for causal effect estimation.

[0078] This involves result variables, which are business variables that change due to the causal influence of changes in dependent variables. In other words, they are the result variables that are expected to produce effects after the strategy is executed. For example, the result variable corresponding to the dependent variable resource utilization rate can be the output conversion rate, and the result variable corresponding to the dependent variable task scheduling strategy can be the response delay. Its main function is to map the strategy effect to specific nodes in the causal graph, serving as the target object for calculating the incremental business output.

[0079] This involves residual variables, which are the business variables other than the dependent and result variables among multiple business variables. For example, if the business variables include resource utilization rate, output conversion rate, response delay, task completion rate, and inventory turnover rate, where the dependent variable is resource utilization rate and the result variable is output conversion rate, then the residual variables are response delay, task completion rate, and inventory turnover rate. Their main function is to provide a candidate set for identifying confounding variables.

[0080] This involves confounding variables, which are business variables that have a causal relationship with both the dependent and effect variables. For example, if a residual variable affects both the dependent and effect variables, then that variable is a confounding variable. For instance, task concurrency affects both resource utilization and output conversion rate. Its main function is to act as a confounding factor to be controlled in causal effect estimation, and to block the interference of non-causal paths by fixing its variable parameters to predetermined parameters.

[0081] This involves predefined parameters, which are fixed values ​​of the interference variables in the causal effect estimation. For example, the predefined parameters can be the mean, median, typical value under a specific business scenario, or fixed value specified by the user. The main function is to fix the variable parameters of the interference variables to predefined parameters so that the causal effect of the change of the dependent variable on the dependent variable is not affected by the fluctuation of the interference variables.

[0082] This involves a reference operating strategy, which is the operating strategy when the business operation corresponding to the dependent variable is not executed. That is, the comparison strategy when the dependent variable does not change or remains in the baseline state. For example, if the initial operating strategy is to improve resource utilization, the reference operating strategy is the original operating mode when the resource utilization is not adjusted. Its main function is to serve as a benchmark for calculating the incremental business output. The incremental output brought by the strategy is determined by comparing the business output under the initial operating strategy and the reference operating strategy.

[0083] This involves business output, which is the quantifiable output result generated by the business system under a given operating strategy. For example, business output can be expressed as a specific value of output conversion rate, actual quantity of product output, percentage of order fulfillment rate, etc. Its main function is to determine the incremental business output by calculating the difference between the business output under the initial operating strategy and the reference operating strategy.

[0084] In this embodiment, the dependent variable and its corresponding effect variable for any initial operating strategy are first determined from multiple business variables, and the strategy operation is mapped to specific intervention nodes and result nodes in the causal graph. Then, based on the dependent and effect variables, the interfering variables are determined from the remaining variables. Since there is a causal relationship between the interfering variables and both the dependent and effect variables, if they are not controlled, they will interfere with the estimation of the causal effect of the dependent variable on the effect variable through confounding paths (causal effect estimation can be achieved using dual machine learning, propensity score matching, inverse probability weighting, causal forest, instrumental variable method, regression discontinuity, difference method, or structural causal model). Then, the variable parameters of the interfering variables are fixed to predetermined parameters to ensure that the impact of changes in the dependent variable on the effect variable is not confused by the fluctuations of the interfering variables. Finally, the difference in business output between the initial operating strategy and the reference operating strategy under the condition that the interfering variables are controlled is determined as the incremental business output. Since the confounding effect of the interfering variables is eliminated, the incremental business output can truly reflect the output change effect brought about by the strategy, avoiding the overestimation or underestimation of the strategy output effect due to ignoring the interfering variables, thus providing an accurate basis for the subsequent comprehensive evaluation of the strategy.

[0085] As an optional embodiment, after determining the target operation strategy of the business system based on business operation requirements and the target cause-effect graph, the process includes: determining a strategy output template, wherein the strategy output template includes a requirement description area, an operation constraint area, and an execution content area, and the operation constraints in the operation constraint area are used to constrain the execution content in the execution content area; and outputting the target operation strategy according to the strategy output template.

[0086] This involves a strategy output template, which is a preset framework used to standardize the output format and content structure of the target operation strategy. For example, the strategy output template includes a requirement description area, an operation constraint area, and an execution content area. Its main function is to ensure that the output target operation strategy has a unified structure and complete content, making it easy for the business system to understand and execute.

[0087] This includes a requirement description area, which is the content area in the strategy output template used to describe the business operation requirements. For example, the requirement description area includes a specific description of the business operation requirements, the current problems or goals, etc., to provide a target description for the execution content area.

[0088] This includes an operational constraint area, which is a content area in the strategy output template used to describe the constraints that must be followed when executing the target operational strategy. For example, the operational constraint area includes budget limits, inventory capacity constraints, production capacity limits, resource allocation (such as equipment configuration) limits, task priority rules, etc., which are used to provide constraint boundaries for the execution content in the execution content area to ensure that the strategy is executed within a feasible range.

[0089] This involves operational constraints, which are specific rules listed in the operational constraints area to limit the feasible scope of business operations. For example, operational constraints include cost not exceeding the budget limit, resource consumption not exceeding the available capacity, and execution cycle not exceeding the specified time limit. These constraints are used to limit the boundaries of the execution content in the execution content area to ensure that the output target operation strategy conforms to the actual execution capability of the business system.

[0090] This includes the execution content area, which is an information area in the strategy output template used to describe the specific operational content of the target running strategy. For example, the execution content area includes suggested actions, execution objects, execution intensity, execution time window, expected output effect, etc., and is used to carry the specific executable instructions of the strategy. It is limited by the running constraints in the running constraint area to ensure that the strategy content is output within the constraint boundaries.

[0091] In this embodiment, a strategy output template is first determined, which includes a requirement description area, a runtime constraint area, and an execution content area. The runtime constraints in the runtime constraint area are used to limit the boundaries of the execution content in the execution content area. Then, the target runtime strategy is output according to the strategy output template. Since the specific operation content in the execution content area is limited by the runtime constraints in the runtime constraint area, it is ensured that the execution content in the output target runtime strategy does not exceed the actual execution capability boundary of the business system. At the same time, the requirement description area clarifies the business background and orientation that the strategy is to solve, so that the output strategy has both clear business relevance and executability, avoiding the strategy output from being infeasible or disconnected from business requirements due to inconsistent structure or missing constraints.

[0092] Based on the above embodiments and optional embodiments, an optional implementation method is provided, which is described in detail below.

[0093] In related technologies, it is necessary to formulate corresponding operational strategies based on business operation requirements to ensure the efficient, stable, and sustainable operation of business systems. However, in these technologies, there are technical problems when determining the business operation strategies of business systems, such as inaccurate business operation strategies, leading to imbalances in resource scheduling, decreased operational efficiency, or even business anomalies.

[0094] There is currently no effective solution to the above problems.

[0095] In view of this, an optional embodiment of the present invention provides a method for determining the business operation strategy of a business system, which can effectively solve the above-mentioned technical problems.

[0096] Figure 2 This is a schematic diagram of the business operation strategy determination process of the business system in an optional embodiment of the present invention. Figure 3This is a schematic diagram of the business operation strategy determination process framework of the business system in an optional embodiment of the present invention, such as... Figure 2 and Figure 3 As shown, a detailed description follows.

[0097] S1, Obtain business operation data from the business system that corresponds to the business operation requirements;

[0098] Business operation requirements Example: How to adjust the operational strategy to mitigate or avoid a decline in operational output when the business system's output decreases in the target region.

[0099] Specifically, operational data can be multi-source operational data, including structured and unstructured operational data (such as unstructured operational text). Structured operational data includes order data, inventory data, production data, supply chain data, financial data, user operation data, and equipment status data. Unstructured operational text includes operational reports, maintenance records, etc.

[0100] After obtaining structured and unstructured runtime data, the process also includes:

[0101] The structured operational data is standardized to include... Taking the runtime data corresponding to each data source as an example, the first... Class data source in the first The data vector within each time window is represented as follows:

[0102]

[0103] in, Let r be the business indicator data vector of the r-th data source within the t-th time window; This refers to the data corresponding to each business indicator in the r-th type of data source; Let r be the total number of business metrics contained in the r-th type of data source; This represents the total number of data sources.

[0104] for The Middle Business metrics data The data undergoes missing value imputation, outlier filtering, time alignment, and normalization to obtain standardized indicator data.

[0105]

[0106] in, For the first Standardized indicator data for each business metric; For the first The average data of each business indicator; For the first The standard deviation of data for each business indicator; It is a very small positive constant (smoothing factor / zero protection factor).

[0107] For data with different time frequencies, a unified time window is used. Perform aggregation. For the original set of sample points... , No. The aggregated features of the time windows are represented as follows:

[0108]

[0109] in, For the first The aggregated indicator data of individual business metrics; For the first Standardized indicator data for each business metric; Aggregate functions can be specified for mean, maximum, minimum, quantile, summation, or other business rules. This is the index of the original sampling time. It is the set of all original sampling times within the t-th time window.

[0110] Unstructured text is segmented, denoised, source-labeled, and time-labeled to obtain a set of text fragments:

[0111]

[0112] in, A collection of text fragments; To transform unstructured raw text data (such as operation reports and maintenance logs) into a structured collection of text fragments; This represents the total number of text fragments.

[0113] S2, extract multiple business variables from business operation data, among which multiple business variables are variables whose representation index of business operation requirements is greater than the representation threshold;

[0114] For example, multiple business variables are extracted from business operation data using a large language model, as shown in the formula:

[0115]

[0116]

[0117] in, A collection of business variables (i.e., semantic objects); For large language models; This is the parameter set for a large language model; A summary of structured indicators; This is a semantic extraction prompt template; Represents a set of events that are executed; Represents a set of operational indicators; Represents a set of strategy actions; Represents a set of constraints; Represents a set of risk factors; This represents the set of causal edges.

[0118] For example, business operation data, including a collection of text fragments. Summary of Structured Indicators and current business operation requirements The common input large language model (referred to as the large model) is guided by semantic extraction prompt templates to extract runtime events, runtime indicators, strategy actions, constraints, risk factors, and causal edges. The large model (which can be a general large language model, a business system vertical domain large model, a large language model fine-tuned by business system runtime data, or a multi-model formed by combining a text large model and a table understanding model) is then used. The semantic objects output by the large language model are then subjected to format validation, field completion, and confidence calculation to obtain a set of semantic objects.

[0119] Several business variables were extracted from the business operation data, including:

[0120] Multiple initial variables are extracted from the business operation data; the confidence levels of the variables corresponding to each of the multiple initial variables are determined; based on the confidence levels of the variables corresponding to each of the multiple initial variables, the initial variables with confidence levels greater than the variable confidence level threshold are selected as business variables, thus obtaining multiple business variables.

[0121] The formula for determining the confidence level of each variable corresponding to multiple initial variables is as follows:

[0122]

[0123] in, For the first Initial variables Confidence of the variables; For initial variables With business operation requirements The degree of correlation; Represents the initial variable The credibility of the source; Represents the initial variable Field format integrity Represents the initial variable Consistency with existing knowledge. , , , These are the corresponding weighting coefficients.

[0124] If a semantic object is missing a necessary field, the object is marked as pending completion, and the large model is called again to complete it. If the format requirements are still not met or the confidence level is below the threshold after completion, the semantic object will not proceed to the next step.

[0125] S3, Construct an initial causal graph, which includes nodes corresponding to multiple business variables and causal edges formed by the causal relationships between multiple business variables;

[0126] Example: Retrieve causal edges extracted from a large model. Each causal edge is represented as:

[0127]

[0128] in, Indicates the cause variable. Represents the outcome variable. Describing the cause variable For outcome variables Conditions for its operation Describing the cause variable After the change, the result variable The time lag between responses This indicates the source of the supporting data.

[0129] S4, based on business operation data, determine the retention index corresponding to each of the multiple causal edges in the initial causal graph;

[0130] S4 includes:

[0131] S41, when the business operation data includes business description text, runtime sequence data and execution constraint data, based on the business description text, determine the text confidence corresponding to multiple causal edges respectively, wherein the text confidence is used to characterize the text support strength of the business description text for the causal relationship corresponding to the causal edge.

[0132] For example, based on the business description text, determine the text confidence corresponding to each of the multiple causal edges using the following formula:

[0133]

[0134] in, Causal edge Text confidence; For activation functions, such as the Sigmoid function (used to compress any real number input to a probability value between 0 and 1). , , , These are the corresponding weighting coefficients; Causal edge The frequency with which the corresponding causal relationship is mentioned in the business description text; Causal edge The credibility of the text source of the supporting text (including business description text) corresponding to the causal relationship; Causal edge The degree of consistency between the corresponding causal direction and the corresponding causal direction described in the supporting text; Causal edge The corresponding causal relationships and their supporting texts and business operation requirements The degree of correlation.

[0135] For example, text confidence can be used to measure the degree of support for causal edges in operation reports and maintenance records.

[0136] in, It was determined in the following way:

[0137]

[0138] in, This indicates that the business description text supports... point to The number of supporting texts, Indicates the number of supporting texts in the opposite direction; It is a very small positive constant (smoothing factor / zero protection factor).

[0139] S42, Based on runtime sequence data, determine the temporal confidence level corresponding to multiple causal edges respectively, wherein the temporal confidence level is used to characterize the temporal support strength of runtime sequence data for the causal relationship corresponding to the causal edge;

[0140] Specifically, based on runtime sequence data, the time-series confidence levels corresponding to multiple causal edges are determined, including: based on runtime sequence data, determining the time-lag response strength and historical response strength corresponding to multiple causal edges; and based on the time-lag response strength and historical response strength corresponding to multiple causal edges, determining the time-series confidence levels corresponding to multiple causal edges.

[0141] For example, based on runtime sequence data, the time series confidence scores corresponding to multiple causal edges are determined using the following formula:

[0142]

[0143] in, Causal edge The time series confidence level; For example, the sigmoid function; , , , These are the corresponding weight coefficients; The time lag response strength represents the causal edge. Corresponding cause variable After the change, the outcome variable under the corresponding lag time condition The response intensity; The conditional independence test score represents the score of the causal variable after controlling for other variables. and outcome variables Whether a statistically significant association still exists between them reflects the degree of independent support for the causal relationship; The historical intervention response intensity represents the causal variable in historical operational data. When changes occur due to external intervention or strategy adjustments, the outcome variable The degree of response that changes accordingly; For the stability of variable relationships, denote causal edges. The stability of the corresponding causal relationship over time.

[0144] in, The following methods can be used to determine this:

[0145]

[0146] in, The maximum value operator represents the maximum value within a given set of candidate lag times. In the meantime, find the lag time that maximizes the absolute value of the correlation. ; The correlation coefficient function; Causal variable exist The value at any given moment is used as the historical value of the causal variable; For outcome variables The value at time t.

[0147] Among them, the target causal support strength can be used. Replace the above time lag response strength Or conditional independence test score , The following methods shall be used to determine:

[0148]

[0149] in, The baseline predicted loss indicates that only the set of control variables is used. Used as input information to predict outcome variables Prediction error at that time; To fully predict the loss, it means using the set of control variables simultaneously. and historical values ​​of causal variables Used as input information to predict outcome variables Prediction error at that time; It is a very small positive constant (smoothing factor / zero protection factor); This represents the prediction loss function, such as mean squared error (MSE) or mean absolute error (MAE).

[0150] S43, Based on the execution constraint data, determine the constraint confidence corresponding to each of the multiple causal edges, wherein the constraint confidence is used to characterize the constraint support strength of the execution constraint data for the causal relationship corresponding to the causal edge;

[0151] For example, based on execution constraint data, determine the constraint confidence corresponding to multiple causal edges using the following formula:

[0152]

[0153] in, Causal edge The constraint confidence level; For activation functions; , ,and These are the corresponding weight coefficients; Causal edge Business process consistency is used to represent causal edges. Does it conform to the order and logic of the business system's pre-defined process? Causal edge Consistency in time direction, used to represent causal edges Does the causal direction align with the known chronological order in the business system? Causal edge Consistency of execution constraints, used to represent causal edges Does it comply with the execution constraints of the business system?

[0154] S44. For any one of the multiple causal edges, the retention index corresponding to any causal edge is determined in the following way: the confidence variance is determined based on the text confidence, time series confidence and constraint confidence corresponding to any causal edge.

[0155] S45, determine the causal conflict index between the text confidence and the constraint confidence, wherein the causal conflict index is used to characterize the degree of conflict between the causal direction represented by the text confidence and the causal direction represented by the constraint confidence.

[0156] For example, the causal direction under text confidence and the causal direction under constraint confidence are determined to obtain the causal conflict index. For example, the value is 0 when the two directions are completely consistent, about 0.3 when the directions are ambiguous, and 0.9~1.0 when the directions are explicitly opposite or the causal direction under text confidence is explicitly prohibited by the causal direction under constraint confidence.

[0157] S46. Based on the confidence variance and the causal conflict index, determine the confidence penalty index corresponding to any causal edge;

[0158] For example, based on the confidence variance and the causal conflict index, the confidence penalty index corresponding to any causal edge is determined by the following formula:

[0159]

[0160] in, Causal edge The confidence penalty index; It is the variance function; Causal edge The confidence level variance; Causal edge The causal conflict index between the corresponding text confidence and constraint confidence.

[0161] S47. Based on the text confidence, time confidence, constraint confidence and confidence penalty index corresponding to any causal edge, determine the retention index corresponding to any causal edge.

[0162] For example, based on the text confidence, time-series confidence, constraint confidence, and confidence penalty index corresponding to any causal edge, the retention index corresponding to any causal edge is determined using the following formula:

[0163]

[0164] in, Causal edge The retention index; For example, the sigmoid function; , , These are the corresponding fusion weights; Conflict penalty weighting;

[0165] S5. Based on the retention indices corresponding to multiple causal edges, the initial causal graph is adjusted to obtain the target causal graph;

[0166] For example, based on the retention indices corresponding to multiple causal edges, the initial causal graph is adjusted to obtain the target causal graph. This includes: marking the causal edges in the initial causal graph to obtain the marking states corresponding to multiple causal edges; and adjusting the initial causal graph according to the marking states corresponding to multiple causal edges to obtain the target causal graph.

[0167] In this process, the causal edges in the initial causal graph are marked to obtain the marked states corresponding to multiple causal edges, as shown in the formula:

[0168]

[0169] in, Causal edge The marked state; It is in a trusted state; The status is pending verification. In deleted status; A high-confidence threshold for edge retention; The low-confidence threshold for edge retention.

[0170] The initial causal graph is adjusted according to the labels corresponding to multiple causal edges to obtain the target causal graph. This includes: if a causal edge is in the HighConf state, it is included in the target causal graph; if it is in the Pending state, its usage weight is reduced in the subsequent causal verifier or it is included in the manual review queue; if it is in the Removed state, it is removed from the causal edge set.

[0171] A goal-cause-effect graph is represented as:

[0172]

[0173] in, For the purpose of a cause-and-effect graph; This is the set of business variables in the target cause-effect graph; Let be the set of causal edges in the target causal graph; The weight matrix of causal edges in the target causal graph (e.g., matrix elements are retention exponents). This is the set of labeled states in the target cause-effect graph.

[0174] S6 determines the target operation strategy of the business system based on business operation requirements and the target cause-effect graph.

[0175] S6 includes: determining multiple initial operating strategies based on business operation requirements and the target cause-effect graph; determining the business output increment, execution cost index, operation risk index, and execution difficulty coefficient corresponding to each of the multiple initial operating strategies; determining the fusion selection index corresponding to each of the multiple initial operating strategies based on the business output increment, execution cost index, operation risk index, and execution difficulty coefficient corresponding to each of the multiple initial operating strategies; and determining the target operating strategy from the multiple initial operating strategies based on the fusion selection index corresponding to each of the multiple initial operating strategies.

[0176] For example, based on business operation requirements Retrieve relevant context from the knowledge base and the operational cause-effect graph, including relevant operational metrics, historical strategies, similar cases, business rules, constraints, and cause-effect graph substructures, as shown below:

[0177]

[0178] in, To retrieve the context set; This indicates the first [item] in the business system knowledge base. A text fragment or structured knowledge unit; Indicates an embedded function; Represents the similarity function; This indicates selecting the option with the highest similarity. One result.

[0179] Business operation requirements , Retrieve context set and causal graph substructure Input a large model and generate an initial set of running strategies. :

[0180]

[0181] in, This indicates the initial runtime strategy and the generated prompt template. This indicates the initial operating strategy.

[0182] Each initial execution strategy is structured and coded:

[0183]

[0184] in, Indicates the strategy name. Indicates the object to be executed. Indicates the intensity of execution. Indicates the execution cycle. Indicates the target indicator. Indicates resource demand, Indicates potential risks. Indicates the applicable conditions.

[0185] Set up a state transition mechanism for the initial running strategy:

[0186]

[0187] in, Initial running strategy The strategy state, Indicates a candidate state. Indicates the verification state. Indicates supporting data state, Indicates the output state. It indicates a refusal.

[0188] The initial running strategy and initial state for large model generation are as follows:

[0189]

[0190] A strategy in a candidate state cannot be directly used as the final suggestion output; it must be verified by a causal validator.

[0191] For any one of the multiple initial operating strategies, the business output increment corresponding to any one initial operating strategy is determined in the following way: From multiple business variables, the dependent variable corresponding to any one initial operating strategy and the result variable corresponding to the dependent variable are determined; based on the dependent and result variables, interference variables are determined from multiple residual variables, where the multiple residual variables are other business variables besides the dependent and result variables, and the interference variables are business variables that have a causal relationship with both the dependent and result variables; the difference in output between the business output under the initial operating strategy and the business output under the reference operating strategy is determined when the variable parameters corresponding to the interference variables are predetermined parameters, and this difference in output is taken as the business output increment, where the reference operating strategy is the operating strategy when the business operation corresponding to the dependent variable is not executed.

[0192] For example, a causal validator is used to validate the initial running strategy. The causal validator identifies the target running metric (i.e., the outcome variable) corresponding to each initial running strategy. For the initial running strategy... The target operating indicators are expressed as For example, the target performance indicators for adjusting the replenishment cycle can be inventory turnover rate, stockout rate, or order fulfillment rate.

[0193] Identifying confounding variable sets based on target cause-effect graphs, such as for initial running strategies. and target operating indicators Mixed variable set Represented as:

[0194]

[0195] in, This represents a confounding variable identification function based on the backdoor criterion.

[0196] In actual implementation, options that meet the blocking requirements can be selected. arrive The minimal set of variables for noncausal paths ,as follows:

[0197]

[0198] in, Indicates the initial running strategy The potential consequences; It is a confounding variable; This is the set of candidate variables.

[0199] For the initial running strategy and benchmark strategy The average causal effect is expressed as:

[0200]

[0201] in, This is the expectation operator, used for averaging. This indicates proactive intervention; These are strategy variables (intervention variables).

[0202] Furthermore, dual machine learning can be used to estimate causal effects, assuming... To handle variables, we first estimate two auxiliary functions:

[0203]

[0204]

[0205] in, This is an estimate of the propensity score function; This is an estimate of the outcome regression function.

[0206] Then construct the residual variables:

[0207]

[0208]

[0209] in, To handle the residuals of variables; The residuals of the target performance indicators are used as the residuals of the outcome variables.

[0210] Causal effect parameters are estimated using orthogonal moment conditions:

[0211]

[0212] in, These are estimates of the causal effect parameters; For causal effect parameters; For the first The residuals of each target performance indicator; For the first The residuals of each treatment variable; To handle the total number of variables.

[0213] Inverse probability weighted estimation can also be used to estimate causal effects. Let:

[0214]

[0215] in, To provide a given confounding variable Under the condition, policy variable The conditional probability value that takes the value 1; It is a probability function; To handle variables, This indicates that the initial running strategy has been executed. This indicates that the baseline strategy has been executed.

[0216] Then, the first The estimated treatment effect of each initial running strategy is as follows:

[0217]

[0218] in, For the first The inverse probability-weighted estimator of an initial running strategy; For the first The processing variables of an initial running strategy; For the first The target performance metrics for the initial operating strategy; For the first Confounding variables for an initial running strategy; This is the estimated propensity score function.

[0219] In addition, propensity score matching, causal forests, instrumental variables, regression discontinuities, difference-in-differences, or structural causal models can be used to estimate the causal effects of the initial running strategy.

[0220] For the initial running strategy It can also perform counterfactual simulations, and the target cause-effect graph structure of the business system is represented as follows:

[0221]

[0222] in, For the first One endogenous variable; For the first One exogenous disturbance variable; For a set of variables; Representing endogenous variables The set of parent nodes in the target cause-effect graph; This represents a structure function.

[0223] For the initial running strategy By intervening to replace the original strategy generation mechanism, counterfactual results are obtained:

[0224]

[0225]

[0226]

[0227] in, This represents the simulation results of the target operating indicators after executing the initial operating strategy. This represents the simulation results of the target operating indicators after implementing the baseline strategy. This indicates the increase in counterfactual business output; These are strategy variables (intervention variables).

[0228] Calculate the overall verification score of the initial running strategy. :

[0229]

[0230] in, Indicates execution cost, Indicates business risk. Indicates the confidence level of causal verification. , , , , These are the corresponding weight coefficients; This is the normalization function.

[0231] like If the value is above the verification threshold and the risk is below the risk threshold, then the initial operating strategy passes causal verification.

[0232]

[0233] in, Indicates the verification threshold. This indicates the risk threshold. Initial running strategies in the Rejected state will not proceed to the final suggestion generation stage.

[0234] S7. Determine the strategy output template. According to the strategy output template, output the target operation strategy. The strategy output template includes a requirement description area, an operation constraint area, and an execution content area. The operation constraints in the operation constraint area are used to constrain the execution content in the execution content area.

[0235] For example, construct a counterfactual support packet by encapsulating the initial running strategy, which has passed causal verification, into a counterfactual support packet. For the initial running strategy... Its counterfactual support data packet Represented as:

[0236]

[0237] in, This includes the initial running strategy number, strategy name, execution target, and target running metrics; This includes causal paths and the controlled set of confounding variables; This includes causal effect estimates, counterfactual business outputs, and confidence levels; This includes execution costs and resource consumption; This includes business risks and risk thresholds; Including constraints such as cost and production capacity; Including supporting data sources; Indicates the policy state.

[0238] The counterfactual support data package can be stored and transmitted in the form of structured text, tables, knowledge graph subgraphs, vectorized support data representations, or database records.

[0239] A set of support data packets is formed by combining multiple counterfactual support data packets corresponding to the verified initial running strategies. :

[0240]

[0241] The counterfactual support data package is converted into a structured representation readable by large models, such as structured text, tables, predefined formats, key-value pair formats, knowledge graph subgraphs, or vectorized support data representations. The system explicitly labels the policy availability status, causal basis, business output scope, cost range, risk warnings, and resource constraints within the support data package.

[0242] If no initial running strategy passes causality verification, i.e. If the current data is insufficient, a final recommendation will not be generated. Instead, a message will be output stating "Insufficient supporting data is required, and supplementary data or manual review is needed." The low-reliability initial operation strategy and the reasons for its failure will be returned to the user or administrator.

[0243] The strategy output template includes a requirement description area (e.g., a problem description area), an operational constraint area (which may further include a prohibited use area and a constraint explanation area), and an execution content area (which may further include a supporting data package area and an output format area). The problem description area describes the current business operation requirements; the supporting data package area is used to input counterfactual supporting data packages that have passed causal verification; the prohibited use area lists initial operational strategies that have failed verification, have insufficient supporting data, or violate resource constraints; the output format area specifies the recommended content format; and the constraint explanation area restricts large models from referencing strategies outside of supporting data packages, from exaggerating output increments, from deleting risk warnings, and from ignoring resource constraints.

[0244] Business operation requirements Counterfactual Support Data Package and prompt template Input a large model and generate a templated running strategy. :

[0245]

[0246] The templated execution strategy generation process is represented as a conditional generation process with supporting data constraints:

[0247]

[0248] in, This represents the space of content that can be generated, defined by the counterfactual data packet. If a policy is not included... If a content is in a state that is not outputtable, it does not belong to the space of generateable content. It is the conditional probability distribution function; This is a candidate template-based execution strategy.

[0249] The templated operation strategy includes suggested actions, execution targets, execution intensity, execution time window, causal basis, expected output increment, execution cost, risk warning, preconditions, monitoring indicators, and feedback methods.

[0250] S8, Constraint Consistency Verification and Proposal Regeneration;

[0251] Constraint consistency verification is performed on the templated execution strategy generated from the large model:

[0252]

[0253] in, This indicates a templated execution strategy; This indicates a set of counterfactual support data packets; This represents the set of business system resources and rule constraints; For verification operation; This indicates that the test has been passed. This indicates a list of violations.

[0254] Specifically, constraint consistency verification includes:

[0255] Cost consistency verification:

[0256]

[0257] in, To execute the templated execution strategy Total cost at the time (estimated); This represents the upper limit of the total cost.

[0258] Inventory consistency verification:

[0259]

[0260] in, To execute the templated execution strategy Total inventory at the time; This represents the upper limit of the total inventory.

[0261] Capacity consistency verification:

[0262]

[0263] in, To execute the templated execution strategy Production capacity demand at that time; This represents the upper limit of production capacity.

[0264] Operational consistency verification:

[0265]

[0266] in, To execute the templated execution strategy The complexity of the operation at that time; This represents the upper limit of operational complexity.

[0267] Risk consistency verification:

[0268]

[0269] in, To execute the templated execution strategy System operation risk value at that time; This represents the upper limit of the system's operational risk value.

[0270] Causal support for data consistency verification:

[0271]

[0272] in, To execute the templated execution strategy A collection of causal evidence fragments at the time; A set of data packets to support counterfactual claims.

[0273] The above constraints are uniformly written as a vector of violation terms. :

[0274]

[0275] in, The status indicator indicates a violation of cost constraints; This indicates a violation of inventory constraints. This indicates a violation of capacity constraints. The operational constraint violated the status flag; This serves as a status indicator for violations of risk thresholds. This constitutes a violation of the state identifier for causal evidence; This is a comprehensive violation status identifier. Each component indicates whether a constraint has been violated.

[0276] For example:

[0277]

[0278] in, Indicates an indicator function, This indicates that the condition is true, meaning the constraint is violated; , indicating that the condition is false, that is, the constraint is satisfied.

[0279] Constraint consistency verification can be implemented using rule engines, symbolic constraint solvers, classification models, natural language consistency judgment models, or a combination of the above methods.

[0280] If the templated execution strategy passes the validation, the final suggestion, i.e., the target execution strategy, will be output:

[0281]

[0282] If there are violations of the templated execution strategy, a list of violations will be generated:

[0283]

[0284] in, For the first Constraint violation status indicator.

[0285] Violations may include exceeding cost limits, insufficient inventory, insufficient production capacity, citing strategies outside of supporting data packages, lack of causal evidence, exaggeration of business output, or lack of risk warnings.

[0286] Return the list of violations, the original templated execution strategy, and the counterfactual support data package to the large model:

[0287]

[0288] in, This indicates that the prompt template will be regenerated.

[0289] Repeat the above steps until the templated execution strategy passes the consistency check or reaches the maximum number of regenerations. The regeneration process is represented as:

[0290]

[0291] in, For the first The template-based execution strategy; For the first The template-based execution strategy; For the first List of violations.

[0292] If the verification still fails after reaching the maximum number of regenerations, a manual review prompt will be output, along with a list of unresolved violations.

[0293] Among them, the proposed regeneration (i.e., policy regeneration) can be carried out by one-time rewriting, multi-round interactive rewriting, local correction based on violations, or overall regeneration based on the initial running policy after reordering.

[0294] S9 performs a feedback closed-loop update.

[0295] After implementing the final recommendations, collect implementation feedback data, including actual operating indicators, implementation costs, resource consumption, risk events, and strategy implementation status.

[0296] Calculate the deviation between actual results and counterfactual simulation results. :

[0297]

[0298] in, This indicates the actual result after the strategy is executed. This represents the counterfactual simulation prediction results. To prevent constants with a denominator of zero.

[0299] Calculate the strategy feedback score :

[0300]

[0301] in, Indicates actual business output. Indicates the actual execution cost. Indicates the intensity of the risk event. This indicates the number of times the constraint has been violated in the history of this strategy type. This indicates the adoption status by management. , , , , These are the corresponding weighting coefficients.

[0302] Next, the knowledge base is updated by writing execution cases, actual output increments / actual output, risk events, etc. into the business system knowledge base, and updating the corresponding document vectors, business tags, and search indexes.

[0303] Update the edge weights of the target causal graph:

[0304]

[0305] in, Represents the historical causal edge weight matrix. This represents the causal edge weight matrix estimated based on the execution feedback. This represents the historical smoothing coefficient.

[0306] Update the initial running strategy to generate weights:

[0307]

[0308] in, This indicates that the old strategy generates weights. This indicates the weights generated by the updated strategy. This represents the historical smoothing coefficient.

[0309] Write the updated initial running strategy generation weights into the large model prompt template or the initial running strategy sorting module, so that when the subsequent large model generates initial running strategies, it will give priority to strategy types with high business output, low risk, few constraint violations and high adoption rate.

[0310] Execution feedback can come from actual operational data of the business system, policy execution status, risk event records, etc.

[0311] The following description, with specific examples, further illustrates this point.

[0312] Taking the business operation requirement as an example: "The incremental output of the target domain corresponding to the business system has decreased in the past three weeks. How to adjust the operation strategy to slow down or avoid the decline in operational output?"

[0313] First, extract runtime events and policy information from business system reports, operational analysis reports, and maintenance records. For example, the large model extracts:

[0314] Runtime events: Target domain runtime output increment decreases, output conversion rate decreases, resource utilization rate decreases, and runtime response latency increases.

[0315] Operational metrics: incremental output, output conversion rate, resource utilization rate, response latency, and task completion rate.

[0316] Strategic actions: Improve resource utilization, optimize task scheduling strategies, increase operational resource investment, adjust output conversion processes, and reduce the occupation of non-core tasks.

[0317] Constraints: Maximum total amount of running resources, task scheduling cycle limit, budget not exceeding a set threshold, and fixed priority of critical tasks.

[0318] Risk factors: Excessive resource consumption leading to decreased stability, optimization results falling short of expectations, increased task scheduling conflicts, and further deterioration of response latency.

[0319] Causal edges: Resource utilization rate → Operational output increment; Output conversion rate → Operational output increment; Response latency → Output conversion rate; Task scheduling strategy → Resource utilization rate; Resource input → Resource utilization rate.

[0320] Then, the causal edges were verified using three sources: runtime sequence data, runtime logs, and business constraints. For the causal edge "resource utilization → incremental runtime output," it was found to be supported by textual data in both the runtime analysis report and historical optimization reviews. Historical time-series data also showed a lagged response of increased output after improved resource utilization, and it conformed to business constraints; therefore, it was marked as a high-confidence causal edge. For the causal edge "resource input directly increases incremental runtime output," its textual support was weak, and it was not supported by business constraints; therefore, it was removed or marked as a low-confidence causal edge.

[0321] Next, the large model generates several initial operating strategies, including: improving resource utilization, optimizing task scheduling strategies, increasing runtime resource investment, adjusting output transformation processes, and reducing the occupation of non-core tasks. All of these strategies are marked as candidate states, and the final recommendation is not directly output.

[0322] Subsequently, the causal validator validates each initial running strategy. For the "improve resource utilization" strategy, the target metrics are identified as output conversion rate, runtime output increment, and response latency, and confounding variables are identified as task concurrency, historical load level, resource base capacity, version, and region requirements. The expected improvement in conversion rate, reduction in response latency, impact on output increment, and risk level of this strategy are obtained through causal effect estimation and counterfactual simulation. For the "optimize task scheduling strategy," its intervention effect on runtime output increment and resource consumption is similarly estimated, and the scheduling cycle change and risk are calculated.

[0323] If the "improve resource utilization" strategy passes causal verification, it is encapsulated into a counterfactual support data package, recording the initial operating strategy, target indicators, causal path, controlled confounding variables, intervention effect, counterfactual business output, execution cost, business risk, resource constraints, budget constraints, and confidence level. If the "optimize task scheduling strategy" strategy has low business output or excessive resource consumption, it is marked as a low-priority strategy or a not recommended strategy.

[0324] Subsequently, the large model generates a templated execution strategy based on counterfactual evidence supporting data packets. For example, it suggests optimizing resource utilization in the target domain over the next two weeks, keeping utilization within a set threshold range, while pausing the addition of new non-core tasks and monitoring output conversion rate, response latency, runtime output increment, and task completion rate. This templated execution strategy then enters the consistency verification module.

[0325] The consistency verification module checks whether the suggestion exceeds the resource budget, whether it will cause the response latency to fall below the threshold, whether it meets resource capacity and business constraints, and whether it references causal evidence in the counterfactual support data packet. If it finds that the templated execution strategy contains the content of "increasing additional large-scale resource investment," but the strategy fails the causal verification, a violation item "policy reference outside the support data packet" is generated, and the violation item, the original templated execution strategy, and the counterfactual support data packet are returned to the large model for rewriting. The large model deletes the resource investment content without supporting data and regenerates a templated execution strategy that meets the constraints.

[0326] Finally, the system outputs operational decision recommendations that pass consistency verification. After the business system executes the recommendations, it collects actual operational output increments, response latency, resource utilization, operating costs, and feedback, and compares the actual results with the counterfactual simulation results. If the actual results are better than the predicted results, the generation weight of similar resource optimization strategies in similar scenarios is increased; if the actual output increment decreases or risk events increase, the generation weight of such strategies is decreased, and the business system knowledge base and causal graph edge weights are updated.

[0327] Through the above steps, the initial operating strategy generated by the large model can be verified through causal validation, counterfactual support data packet constraints, and consistency checks before forming the final recommendation, thereby improving the credibility, interpretability, and executability of the business system's operational decision recommendations.

[0328] The above optional implementation methods can achieve at least the following beneficial effects:

[0329] (1) Compared with related technologies, this invention first obtains business operation data corresponding to business operation requirements in the business system, and extracts multiple business variables whose representation index of business operation requirements is greater than the representation threshold from the business operation data. This allows the focus on the key variables most relevant to the current requirements and avoids interference from irrelevant variables on subsequent causal modeling. Then, based on multiple business variables, an initial causal graph containing nodes and causal edges is constructed, which can explicitly structure the causal relationships between variables and provide a basis for the reliability assessment of subsequent causal edges. Finally, based on the business operation data, the retention index corresponding to multiple causal edges in the initial causal graph is determined, which can quantify the credibility of each causal edge and avoid introducing unreliable factors based solely on experience or subjective judgment. The initial causal graph is then adjusted based on the retention indices corresponding to multiple causal edges to obtain the target causal graph. This process removes unreliable causal edges and retains highly reliable ones, ensuring that the target causal graph accurately reflects the causal structure between variables in the business system. Finally, the target operational strategy is determined based on the business operation requirements and the target causal graph. Since the target causal graph eliminates unreliable causal relationships, the strategy derived from it has a causal basis, avoiding decision-making biases caused by causal misjudgments or variable omissions during strategy formulation. This achieves accurate determination of the business operation strategy and solves the technical problem of inaccurate business operation strategies in related technologies.

[0330] (2) Compared with related technologies, this invention first determines the confidence variance based on text confidence, time-series confidence, and constraint confidence, which can quantify the degree of dispersion among the three dimensions of confidence. The greater the dispersion, the greater the difference in the support strength of different data sources for the same causal edge, and the worse the consistency among multi-source information. Then, it determines the causal conflict index between text confidence and constraint confidence, which can quantify the degree of conflict between the causal direction represented by the text and the causal direction represented by the constraint. Finally, it determines the confidence penalty index based on the confidence variance and the causal conflict index, since the greater the confidence variance, the greater the consistency among multi-source information. The greater the discrepancy between confidence levels and the higher the causal conflict index, the more severe the conflict between the text direction and the rule direction. This allows the confidence penalty index to accurately reflect the degree of inconsistency of causal edges in multi-source information. Finally, the text confidence, temporal confidence, constraint confidence, and confidence penalty index are integrated to determine the retention index. Since the penalty term can deduct the confidence of causal edges with inconsistencies in multi-source information, it ensures that causal edges with discrepancies or conflicts between multi-source information receive a lower retention index. This avoids unreliable causal edges being mistakenly retained due to inflated confidence in a single dimension, thus providing a reliable quantitative basis for the accurate adjustment of the causal graph in the future.

[0331] (3) Compared with related technologies, this invention first determines multiple initial operation strategies based on business operation requirements and target cause-effect graphs, so that the generated initial operation strategies are both demand-oriented and supported by causal structure; then determines the business output increment, execution cost index, operation risk index and execution difficulty coefficient corresponding to each initial operation strategy, which can quantify the comprehensive performance of each strategy from four dimensions: output, cost, risk and executability; then determines the fusion selection index corresponding to each initial operation strategy based on the above four indicators. Since the fusion selection index integrates information from four dimensions: output, cost, risk and executability, it can comprehensively reflect the comprehensive priority of each strategy; finally, the target operation strategy is determined from multiple initial operation strategies based on the fusion selection index, ensuring that the selected strategy achieves the optimal balance between output, cost, risk and executability, avoiding strategy selection bias caused by relying on only a single dimension evaluation, thereby providing the business system with the best comprehensive performance operation strategy.

[0332] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, because according to the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0333] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0334] Example 2

[0335] According to embodiments of the present invention, an apparatus for implementing the above-described method for determining business operation strategies of a business system is also provided. Figure 4 This is a structural block diagram of a business operation strategy determination device for a business system according to an embodiment of the present invention, such as... Figure 4 As shown, the device includes: a first determining module 402, a second determining module 404, a third determining module 406, a fourth determining module 408, a fifth determining module 410, and a sixth determining module 412. The device will be described in detail below.

[0336] The first determining module 402 is used to obtain business operation data corresponding to business operation requirements in the business system.

[0337] The second determining module 404 is connected to the first determining module 402 and is used to extract multiple business variables from the business operation data, wherein the multiple business variables are variables whose representation index of business operation requirements is greater than the representation threshold.

[0338] The third determining module 406 is connected to the second determining module 404 and is used to construct an initial causal graph, wherein the initial causal graph includes nodes corresponding to multiple business variables and causal edges formed by the causal relationships between multiple business variables.

[0339] The fourth determining module 408, connected to the third determining module 406, is used to determine the retention index corresponding to multiple causal edges in the initial causal graph based on business operation data.

[0340] The fifth determining module 410, connected to the fourth determining module 408, is used to adjust the initial causal graph based on the retention indices corresponding to multiple causal edges to obtain the target causal graph.

[0341] The sixth determining module 412, connected to the fifth determining module 410, is used to determine the target operation strategy of the business system based on business operation requirements and the target cause-effect graph.

[0342] It should be noted here that the first determining module 402, the second determining module 404, the third determining module 406, the fourth determining module 408, the fifth determining module 410, and the sixth determining module 412 mentioned above correspond to steps S102 to S112 in the method for determining the business operation strategy of the implementation business system. The instances and application scenarios implemented by multiple modules and corresponding steps are the same, but are not limited to the content disclosed in the above embodiment 1.

[0343] Example 3

[0344] According to another aspect of the present invention, an electronic device is also provided, comprising: a processor; and a memory for storing processor-executable instructions, wherein the processor is configured to execute instructions to implement the business operation strategy determination method of the business system described above.

[0345] Example 4

[0346] According to another aspect of the present invention, a computer-readable storage medium is also provided, which, when the instructions in the computer-readable storage medium are executed by the processor of an electronic device, enables the electronic device to perform the business operation strategy determination method of the business system described above.

[0347] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0348] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0349] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0350] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0351] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0352] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0353] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for determining the business operation strategy of a business system, characterized in that, include: Obtain business operation data from the business system that corresponds to the business operation requirements; Multiple business variables are extracted from the business operation data, wherein the multiple business variables are variables whose representation index of the business operation requirements is greater than the representation threshold. Construct an initial causal graph, wherein the initial causal graph includes nodes corresponding to the plurality of business variables, and causal edges formed by the causal relationships between the plurality of business variables; Based on the business operation data, the retention index corresponding to each of the multiple causal edges in the initial causal graph is determined. Based on the retention indices corresponding to the multiple causal edges, the initial causal graph is adjusted to obtain the target causal graph; Based on the business operation requirements and the target cause-effect graph, the target operation strategy of the business system is determined.

2. The method according to claim 1, characterized in that, The step of determining the retention index corresponding to each of the multiple causal edges in the initial causal graph based on the business operation data includes: When the business operation data includes business description text, runtime sequence data, and execution constraint data, based on the business description text, the text confidence level corresponding to each of the multiple causal edges is determined, wherein the text confidence level is used to characterize the text support strength of the business description text for the causal relationship corresponding to the causal edge. Based on the runtime sequence data, the temporal confidence level corresponding to each of the plurality of causal edges is determined, wherein the temporal confidence level is used to characterize the temporal support strength of the runtime sequence data for the causal relationship corresponding to the causal edge; Based on the execution constraint data, the constraint confidence level corresponding to each of the plurality of causal edges is determined, wherein the constraint confidence level is used to characterize the constraint support strength of the execution constraint data for the causal relationship corresponding to the causal edge; Based on the text confidence, temporal confidence, and constraint confidence corresponding to the multiple causal edges, the retention index corresponding to the multiple causal edges in the initial causal graph is determined.

3. The method according to claim 2, characterized in that, For any one of the plurality of causal edges, the retention index corresponding to that causal edge is determined in the following manner: Based on the text confidence, time-series confidence, and constraint confidence corresponding to any causal edge, determine the confidence variance; Determine the causal conflict index between the text confidence and the constraint confidence, wherein the causal conflict index is used to characterize the degree of conflict between the causal direction represented by the text confidence and the causal direction represented by the constraint confidence; Based on the confidence variance and the causal conflict index, determine the confidence penalty index corresponding to any causal edge; Based on the text confidence, time confidence, constraint confidence, and confidence penalty index corresponding to any causal edge, determine the retention index corresponding to any causal edge.

4. The method according to claim 2, characterized in that, The step of determining the temporal confidence level corresponding to each of the plurality of causal edges based on the runtime sequence data includes: Based on the runtime sequence data, determine the time lag response strength and historical response strength corresponding to the plurality of causal edges, respectively; Based on the time lag response strength and historical response strength corresponding to the multiple causal edges, the time series confidence level corresponding to the multiple causal edges is determined.

5. The method according to claim 1, characterized in that, The process of determining the target operation strategy of the business system based on the business operation requirements and the target cause-effect graph includes: Based on the aforementioned business operation requirements and the target cause-effect graph, multiple initial operation strategies are determined; Determine the incremental business output, execution cost index, and operational risk index, as well as the execution difficulty coefficient, corresponding to the multiple initial operating strategies; Based on the incremental business output, execution cost index, and operational risk index corresponding to the multiple initial operating strategies, and the execution difficulty coefficient, determine the fusion selection index corresponding to the multiple initial operating strategies. Based on the fusion selection index corresponding to the multiple initial operating strategies, the target operating strategy is determined from the multiple initial operating strategies.

6. The method according to claim 5, characterized in that, For any one of the multiple initial operation strategies, the incremental business output corresponding to any one of the initial operation strategies is determined in the following manner: From the plurality of business variables, determine the dependent variable corresponding to any one of the initial operating strategies, and the result variable corresponding to the dependent variable; Based on the dependent variable and the effect variable, interference variables are determined from multiple residual variables. The multiple residual variables are other business variables besides the dependent variable and the effect variable among the multiple business variables, and the interference variables are business variables that have a causal relationship with both the dependent variable and the effect variable. When the variable parameter corresponding to the interference variable is a predetermined parameter, the difference in output between the business output under the initial operating strategy and the business output under the reference operating strategy is determined, and the difference in output is taken as the business output increment. The reference operating strategy is the operating strategy when the business operation corresponding to the dependent variable is not executed.

7. The method according to any one of claims 1 to 6, characterized in that, After determining the target operation strategy of the business system based on the business operation requirements and the target cause-effect graph, the process includes: Determine the strategy output template, wherein the strategy output template includes a requirement description area, a running constraint area, and an execution content area, wherein the running constraints in the running constraint area are used to constrain the execution content in the execution content area; Output the target operation strategy according to the output template of the strategy.

8. A device for determining the business operation strategy of a business system, characterized in that, include: The first determination module is used to obtain business operation data corresponding to business operation requirements in the business system. The second determining module is used to extract multiple business variables from the business operation data, wherein the multiple business variables are variables whose representation index of the business operation requirements is greater than a representation threshold. The third determining module is used to construct an initial causal graph, wherein the initial causal graph includes nodes corresponding to the plurality of business variables, and causal edges formed by the causal relationships between the plurality of business variables; The fourth determining module is used to determine the retention index corresponding to each of the multiple causal edges in the initial causal graph based on the business operation data. The fifth determining module is used to adjust the initial causal graph based on the retention indices corresponding to the plurality of causal edges to obtain the target causal graph; The sixth determining module is used to determine the target operation strategy of the business system based on the business operation requirements and the target cause-effect graph.

9. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the instructions to implement the business operation strategy determination method of the business system as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, When the instructions in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device is able to perform the business operation strategy determination method of the business system as described in any one of claims 1 to 7.

Images