A method for intelligently constructing a tobacco business enterprise strategy map

Abstract

The present application relates to the technical field of data processing, in particular to a kind of tobacco commercial enterprise strategy map intelligent construction method, obtains multidimensional business data;Through first model, preliminary strategic elements are obtained by semantic disassembly;Semantic entity is mapped to structured constraint parameter and is solved by second model optimization to obtain planning scheme;Based on domain rule base, adaptability check is carried out, and in response to the check fails, generate correction feedback signal injection first model and re-disassemble until pass, output target strategy map data;In response to data variation meets preset trigger condition, trigger model re-computation to update data;The present application realizes semantic understanding and operational optimization cooperation and check feedback closed loop, improves the precision and dynamic adaptability of strategy map construction.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method for intelligently constructing strategic maps for tobacco commercial enterprises. Background Technology

[0002] In the field of corporate strategic planning and management, strategic maps, as a core tool connecting macro-strategic goals with micro-execution paths, directly determine the effectiveness of strategy implementation through their construction quality.

[0003] With the deepening of digital transformation, enterprises are accumulating explosive growth in business data, and the types of data are becoming increasingly complex. This includes both unstructured text data such as policies, regulations, and market trends, as well as structured numerical data such as financial indicators and operational status. Traditional strategy map construction mainly relies on manual research and expert discussions. When faced with massive amounts of multi-source heterogeneous data, this approach is not only time-consuming and lengthy, but also makes it difficult to guarantee the comprehensiveness and objectivity of data analysis, causing the strategy formulation cycle to lag behind changes in the market environment.

[0004] In recent years, although artificial intelligence technology has been gradually introduced into enterprise management decision support, existing intelligent solutions still have significant technical shortcomings when dealing with complex strategic planning tasks.

[0005] On the one hand, generative AI, represented by large language models, is good at semantic understanding and content generation of unstructured text. However, when faced with planning problems that require strict mathematical logic and resource constraints, it is prone to the "illusion" phenomenon that is out of touch with the actual resource carrying capacity, and the generated strategic paths often lack feasibility. On the other hand, although traditional operations research optimization algorithms can solve the optimal resource allocation scheme under given constraints, they are highly dependent on structured input parameters and cannot directly understand the macro-strategic intentions described in natural language, resulting in a serious data gap between the semantic layer and the computational layer.

[0006] Existing technologies typically use these two types of models in simple chain or independently, lacking an effective mechanism for mapping and converting semantics into constraint parameters. This makes it difficult to accurately translate macro-strategic intentions into the boundary conditions and objective functions of mathematical models.

[0007] Furthermore, existing intelligent planning systems generally lack automated verification and feedback correction mechanisms based on domain-specific rules. When the planning scheme output by the model does not match industry-specific compliance requirements, business-related rules, or quantitative indicator systems, the system cannot automatically identify the deviation and convert the deviation information into a correction signal that the model can understand. Manual adjustments are only possible after manual review. This open-loop processing method is not only inefficient but also fails to form a data-driven iterative optimization loop, making it difficult for the generated strategic map to continuously adapt to the dynamically changing internal and external environment.

[0008] Meanwhile, when dealing with changes in multi-source heterogeneous data such as policy adjustments and market fluctuations, existing technologies mostly adopt threshold triggering methods based on single indicators, lacking the ability to weightedly integrate and calculate changes in multi-dimensional data. This can easily lead to the accidental triggering of global recalculation due to normal fluctuations in local data, or the omission of systemic risks due to the failure of a single indicator to exceed the limit, resulting in a lack of robustness and accuracy in the dynamic update mechanism of strategic maps.

[0009] Therefore, there is an urgent need to develop a strategic map intelligent construction technology solution that can deeply integrate semantic understanding and operational optimization, possess a closed loop of domain rule verification and feedback, and have the ability to dynamically adapt to multi-source data. Summary of the Invention

[0010] The technical problem to be solved by this invention is to provide a method for intelligent construction of strategic maps for tobacco commercial enterprises, which can realize accurate semantic decomposition of strategic intentions, structured constraint mapping, adaptive verification and correction, and dynamic updating.

[0011] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A method for intelligently constructing a strategic map for tobacco businesses, comprising the following steps: Acquire multi-dimensional business data from tobacco commercial enterprises, including policy guidance data, market operation data, and internal management data. The first model is used to semantically decompose the multidimensional business data, extract strategic intent information and business constraint information from the multidimensional business data, and obtain preliminary strategic elements. The semantic entities in the preliminary strategic elements are mapped to structured constraint parameters, and the second model is used to optimize and solve the solution based on the structured constraint parameters to obtain a planning scheme that includes resource allocation results and path planning results. The structured constraint parameters include boundary constraint conditions and objective functions; The adaptability of the planning scheme is verified based on the domain rule base, and the verification result is obtained. In response to the verification result indicating that the verification failed, a correction feedback signal is generated based on the verification deviation item and the corresponding correction suggestion, and the correction feedback signal is injected into the first model as a context input to trigger the first model to re-decompose semantics based on the correction feedback signal until the verification passes and the target strategic map data is output. In response to the detection that the data changes of the multidimensional business data meet the preset trigger conditions, the first model and the second model are triggered to re-execute the calculation to update the target strategic map data.

[0012] To solve the above-mentioned technical problems, another technical solution adopted by the present invention is as follows: A smart terminal for constructing a strategic map for a tobacco business enterprise includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the smart construction method for constructing a strategic map for a tobacco business enterprise described above.

[0013] The beneficial effects of this invention are as follows: This invention provides an intelligent method for constructing strategic maps for tobacco commercial enterprises. By deeply integrating large language models and operations research algorithms, it constructs an end-to-end intelligent mapping link from macro-strategic intent to executable planning schemes. It utilizes the semantic understanding capabilities of large language models to accurately decompose multi-dimensional business data such as policy guidance, market operations, and internal enterprise management. Furthermore, by transforming semantic entities into structured constraint parameters containing boundary constraints and objective functions, it enables operations research algorithms to find optimal solutions under hard constraints such as total resources and time nodes. This fundamentally eliminates the technical defect of large language models easily generating the illusion of resource disconnect. Simultaneously, it overcomes the data gaps that traditional operations research models cannot understand natural language strategic intent, ensuring the generated planning schemes are optimized. It combines strategic rationality and resource feasibility. Based on this, a domain rule base is introduced to automatically verify the adaptability of the planning scheme. When the verification fails, the verification deviation items and correction suggestions are constructed as correction feedback signals and injected into the large language model as context, driving the model to re-decompose semantics and form a closed-loop iterative correction mechanism. This solves the technical pain point that open-loop systems cannot automatically identify and correct industry rule deviations. In addition, by performing weighted fusion calculations on multi-source heterogeneous monitoring data, it comprehensively determines whether data changes meet the triggering conditions. This overcomes the defects of single indicator threshold triggering methods, which are prone to misjudgment or omission of systemic risks. This makes the dynamic update of the strategic map more robust and accurate, thus realizing the technical leap from static generation to dynamic adaptation of the strategic map. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating an intelligent construction method for a strategic map of a tobacco business enterprise, according to an embodiment of the present invention. Figure 2 This is a schematic diagram of a smart construction terminal for a strategic map of a tobacco business enterprise, according to an embodiment of the present invention. Label Explanation: 1. A smart terminal for constructing strategic maps for tobacco commercial enterprises; 2. Processor; 3. Memory. Detailed Implementation

[0015] To explain in detail the technical content, objectives, and effects of the present invention, the following description is provided in conjunction with the embodiments and accompanying drawings.

[0016] Please refer to Figure 1 One embodiment of the present invention is as follows: A method for intelligently constructing a strategic map for tobacco businesses, comprising the following steps: Step S100: Obtain multi-dimensional business data of tobacco commercial enterprises, including policy guidance data, market operation data, and internal management data of enterprises.

[0017] Specifically, the process is executed in parallel by an internal data interface unit and an external data crawler unit. Policy-oriented data primarily originates from external policy and regulatory websites and normative documents issued by industry authorities. This data is typically in HTML text streams or PDF documents and requires structured extraction through natural language processing. Market operation data includes regional cigarette consumption demand, retail terminal operating conditions, and competitor dynamics, usually synchronized in real-time from industry databases or marketing systems in JSON or XML format. Internal enterprise management data covers data from business subsystems such as sales, logistics, finance, talent, and compliance, obtained through API interfaces in the form of standardized data tables. The data processing unit cleans, deduplicates, and standardizes this multi-source heterogeneous data, forming a unified structured dataset to provide a high-quality data foundation for subsequent model calculations. This multimodal data fusion acquisition mechanism ensures that the strategic map construction can comprehensively perceive changes in the internal and external environment, avoiding information bias caused by a single data source.

[0018] Step S200: Semantically decompose the multidimensional business data using the first model, extract strategic intent information and business constraint information from the multidimensional business data, and obtain preliminary strategic elements.

[0019] In this embodiment, the first model employs a large language model fine-tuned based on tobacco industry corpus. This model receives the structured data output from step S100 and utilizes its deep semantic understanding capabilities to identify the strategic objectives, development paths, and limiting conditions implicit in the text.

[0020] For example, from an annual work report, strategic intentions such as "promoting high-quality development" and "improving logistics efficiency" can be extracted, along with business constraints such as "strictly controlling non-productive expenditures" and "ensuring zero accidents in safe production." Preliminary strategic elements are intermediate data existing in the form of natural language or semi-structured tags. They have completed the abstraction from raw text to business concepts, but have not yet been transformed into mathematical parameters that computers can directly calculate.

[0021] Step S300: The semantic entities in the preliminary strategic elements are mapped to structured constraint parameters, and the second model is used to optimize and solve the solution based on the structured constraint parameters to obtain a planning scheme that includes resource allocation results and path planning results. The structured constraint parameters include boundary constraints and objective functions.

[0022] This is the core element of achieving semantic and operational research collaboration in this embodiment. As one implementation method, the first model is a large language model, and the second model is an operations research algorithm model; the semantic entities in the initial strategic elements are mapped to structured constraint parameters, including: The semantic entities in the initial strategic elements are transformed into the input parameter format of the operations research algorithm model. The boundary constraints are generated based on the resource constraint information and time constraint information in the semantic entities, and the objective function is generated based on the optimization objective information in the semantic entities.

[0023] To illustrate the mapping mechanism more clearly, a specific data transformation example is given below. Assume that the preliminary strategic elements extracted in step S200 contain the text fragment: "The total budget for the city's cigarette marketing special project in 2024 shall not exceed 5 million yuan, and 70% of the funds must be disbursed before the third quarter." The system first transforms this text into a semantic entity object through named entity recognition and relation extraction, for example, represented as {Entity: Budget, Value: 5,000,000, Unit: CNY, TimeConstraint: Q3, Ratio: 0.7}. Subsequently, the parameter mapping engine, according to predefined transformation rules, parses this semantic entity object into a mathematical expression recognizable by the operations research algorithm model. Specifically, for the resource constraint information "not exceeding 5 million yuan," it generates linear programming inequality constraints: ； in Representing the Budget allocation variables for each marketing sub-project; generate time-series constraints based on the time constraint information of "completing 70% by the third quarter". for: ; in, Representing the The amount of resources allocated to each strategic task; It represents three different business sets (e.g., "Five Quality Improvement Actions", "Six Empowerment Projects", "Core Exclusive Sales Supervision"). This indicates the total budget.

[0024] Meanwhile, if the semantic entity contains optimization objective information such as "maximizing market share" or "minimizing logistics costs," it is transformed into an objective function. or The coefficient vector. Through this deterministic mapping mechanism, the unstructured semantics output by the large language model are accurately translated into the input matrix and constraint vector of the operations research algorithm, fundamentally solving the uncertainty problem of generative AI in numerical programming and ensuring the rigor and feasibility of subsequent mathematical solutions.

[0025] Furthermore, the operations research algorithm model includes linear programming algorithms or genetic algorithms; the boundary constraints include total resource constraints, time node constraints, and compliance requirement constraints. The planning scheme is obtained by solving the second model based on structured constraint parameters, including: The objective function is optimized under boundary constraints using an operations research algorithm model to obtain resource allocation results and path planning results, which are then used as planning schemes.

[0026] In practical applications, for resource allocation problems, linear programming or mixed-integer programming algorithms are preferred to find the global optimum. For complex scenarios involving multi-stage, nonlinear path planning, genetic algorithms can be used for heuristic search. Compliance requirements are usually embedded in the model as hard constraints, such as the scope of business licenses stipulated by laws and regulations, and safety production standards, to ensure that the generated planning scheme is feasible within the framework of laws and industry norms.

[0027] Step S400: Perform adaptability verification on the planning scheme based on the domain rule base and obtain the verification result; in response to the verification result indicating that the verification fails, generate a correction feedback signal based on the verification deviation item and the corresponding correction suggestion, and inject the correction feedback signal as context input into the first model to trigger the first model to re-decompose semantics based on the correction feedback signal until the verification passes, and output the target strategic map data.

[0028] It should be noted that this step constitutes the closed-loop feedback control mechanism of the system. Its specific verification dimensions, negative feedback prompt word construction logic, and iteration termination conditions will be elaborated in detail in subsequent embodiments. In this embodiment, this step ensures that the final output target strategic map data not only satisfies mathematical optimality but also conforms to the business rules and strategic orientation of the specific domain.

[0029] Step S500: After outputting the target strategic map data, the target strategic map data is structured and organized according to multiple preset value dimensions, including financial value dimension, customer value dimension, core operating process dimension and organizational capability support dimension. Under each value dimension, the corresponding strategic objectives, execution paths, responsible parties, time nodes, and quantitative indicators are associated to generate a visual strategic map; In response to receiving instructions to expand a level or penetrate an indicator, the corresponding sub-targets or underlying indicator data are displayed in the visualized strategic map.

[0030] In this embodiment, the Balanced Scorecard approach is used to visualize abstract strategic data. For example, under the "Core Operational Process Dimension," the system automatically associates strategic objectives related to the "Modern Logistics Quality Improvement Action," optimized delivery routes, responsible entities for logistics centers, quarterly assessment milestones, and quantitative indicators such as per-box logistics costs. When a user clicks on a high-level objective on the interface, the system responds with a hierarchy expansion command, drilling down level by level to display sub-objectives and detailed underlying data, achieving seamless penetration from macro-strategy to micro-execution. This structured organization not only improves the readability of strategic information but also provides standardized data anchors for subsequent dynamic monitoring.

[0031] In addition, to ensure data security, this embodiment also includes: The transmission process of multi-dimensional business data and target strategic map data is encrypted using the SSL encryption protocol. The AES-256 encryption algorithm is used to encrypt and store the multi-dimensional business data and target strategic map data. Based on the role-based access control mechanism, strategic map viewing, editing and approval permissions are assigned to roles with different permissions. Record all data access and operation behaviors to generate security audit logs.

[0032] Specifically, the system categorizes roles based on job responsibilities, such as strategic planning, business execution, and audit supervision. Different roles can only access strategic map views and data within their authorized scope. For example, personnel at the municipal level can only view relevant indicators for their city, while provincial-level managers can view aggregated data and detailed breakdowns for the entire province. All data creation, deletion, modification, and query operations are recorded in an immutable security audit log, supporting post-event traceability and compliance review. Through a four-tiered protection mechanism of encrypted transmission, encrypted storage, access control, and audit trail, the system effectively ensures the security and compliance of tobacco commercial enterprises' core strategic data throughout its entire lifecycle.

[0033] Embodiment 2 of the present invention is as follows: A method for intelligently constructing a strategic map for tobacco commercial enterprises differs from Embodiment 1 in that it further details a closed-loop mechanism for adaptation verification and feedback based on a domain rule base, as well as a dynamic adjustment triggering mechanism driven by multi-source data. These two mechanisms together constitute the core technical support for the system's evolution from "static generation" to "adaptive optimization," effectively resolving the potential conflict between the "illusion" problem of content generated by large language models and the "rigid constraints" of operations research algorithms, and improving the robustness of the strategic map in responding to changes in the internal and external environment.

[0034] In the specific implementation of step S400, the adaptability of the planning scheme is verified based on the domain rule base, and the verification result is obtained, including: The planning scheme is matched and compared with the rule elements in the domain rule base to obtain the verification deviation items; Based on the verification deviation items and corresponding correction suggestions, a correction feedback signal is generated, including: In response to the existence of a verification deviation item, the verification deviation item and the corresponding correction suggestion are constructed as a negative feedback prompt word, and the negative feedback prompt word is used as a correction feedback signal.

[0035] Specifically, the adaptability verification and correction feedback process is not a simple manual review, but a fully automated data structure comparison. The domain rule base stores industry compliance guidelines and business development elements that have been technically abstracted (such as core definitions, quantitative indicator systems, and business-related rules in specific development strategies). These rules are pre-coded into machine-readable verification logic. When the planning scheme output by the second model (including resource allocation results and path planning results) enters the verification module, the system automatically extracts the key parameters and logical links in the scheme and matches them one by one with the standard elements in the rule base. Once a mismatch is found, a verification deviation item is generated. To ensure that the first model can accurately understand the meaning of the deviation and make targeted corrections, the system constructs the verification deviation item and the corresponding correction suggestions as negative feedback prompts according to a preset structured template.

[0036] For example, a typical negative feedback prompt template structure is as follows: "[ERROR] Resource allocation exceeds compliance threshold [RULE_ID: C03-BudgetCap] | Current value: 5.2 million | Limit value: 5 million | Deviation type: Hard constraint violation | Suggested correction direction: Please readjust the marketing budget allocation to ensure the total amount does not exceed 5 million, and prioritize investment in core terminal construction." This negative feedback prompt is injected as context input into the prompt window of the first model, triggering the first model to re-decompose semantics and generate solutions based on this correction feedback signal. This mechanism of converting structured verification results into natural language correction instructions allows the large language model to maintain its semantic understanding advantage while being precisely guided by external deterministic rules, thus effectively overcoming the compliance deficiency problem caused by simply relying on the model's endogenous knowledge.

[0037] Furthermore, to ensure the comprehensiveness and accuracy of the verification, the planning scheme is matched and compared with the rule elements in the domain rule base to obtain the verification deviation items, including: The planning scheme was compared with the rule elements from the dimensions of element completeness, business relevance rationality, and indicator quantification feasibility, and the deviation results of each dimension were obtained. In response to the existence of verification deviation items, the verification deviation items and corresponding correction suggestions are constructed as negative feedback prompts, including: In response to any deviation result exceeding a preset deviation threshold, the deviation result and correction suggestions corresponding to the dimension exceeding the preset deviation threshold are constructed as negative feedback prompt words.

[0038] Among them, the element completeness dimension mainly verifies whether the planning scheme covers all the necessary strategic elements required by the domain rule base (such as whether specific organizational construction or security tasks are omitted), which is achieved by comparing the difference between the output tags of the scheme and the necessary tag set of the rule base through set operations; the business relevance rationality dimension mainly verifies whether the logical causal relationship between each strategic element conforms to industry best practices (such as whether "reduced logistics costs" and "improved sorting efficiency" have established a correct driving relationship), which is automatically verified through knowledge graph path reachability analysis or predefined logical rule matrices; the indicator quantification feasibility dimension verifies whether each quantitative indicator is within a reasonable range formed by historical data and industry benchmarks, to prevent unrealistic, aggressive or conservative goals.

[0039] This embodiment also includes the following steps: When the number of iterations for semantic decomposition reaches the preset maximum number of iterations, or when the deviation results of all dimensions are less than or equal to the preset deviation threshold, the iteration terminates and the target strategic map data is output.

[0040] This design of multi-dimensional automatic verification and dual termination conditions avoids invalid iterations caused by misjudgment in a single dimension and prevents the model from falling into an infinite loop, ensuring that the system converges to the optimal solution that satisfies multiple constraints under limited computing resources.

[0041] In the dynamic maintenance phase following step S500, it is monitored whether the data changes of the multidimensional business data meet the preset trigger conditions, including: Real-time acquisition of multi-source heterogeneous monitoring data, including policy change data, market sales data, and internal operation data; A weighted fusion calculation is performed on multi-source heterogeneous monitoring data to obtain a comprehensive deviation value; When the overall deviation value exceeds the preset overall threshold, it is determined that the data change meets the preset trigger condition.

[0042] In this embodiment, the simplistic logic of relying solely on a single indicator (such as sales volume or policy documents) to trigger updates, as in traditional technologies, is abandoned in favor of a multi-source heterogeneous data fusion and perception mechanism. This is because the strategic execution environment of tobacco commercial enterprises is highly complex, and fluctuations in a single indicator are often noise or localized phenomena. Only the coordinated changes in multi-dimensional data can characterize the changes in the systemic environment.

[0043] Specifically, a weighted fusion calculation is performed on multi-source heterogeneous monitoring data to obtain a comprehensive deviation value, including: Calculate the changes in policy change data, market sales data, and internal operating data relative to the baseline data, respectively. Based on a preset weighting strategy, the variation ranges of various data are weighted and summed to obtain a comprehensive deviation value; the preset comprehensive threshold can be customized according to the enterprise's strategic level.

[0044] To more clearly illustrate the weighted fusion calculation logic described above, a specific calculation formula example is given below. Let the overall deviation value be... Its calculation model is as follows: ; in, , , These represent the policy change index, actual market sales value, and internal operational efficiency value, respectively, within the current monitoring period. , , These represent the corresponding benchmark values ​​(such as the value for the same period of the previous year or the pre-set value for strategic planning). , , These are the weight coefficients of the three components, and they satisfy... In this formula, the policy change index The semantic similarity difference between the latest policy text and the benchmark policy text can be quantified using natural language processing technology; market sales and internal operational data are directly taken from the business system. Through this weighted summation method, the system can comprehensively assess the overall strength of environmental changes. For example, when market sales decline slightly but policies undergo significant adjustments, the system's weighting is adjusted accordingly. High, overall deviation value It is still possible to quickly exceed the threshold and trigger an update, thus avoiding the defect of missing policy risks due to a single sales indicator not exceeding the limit; conversely, if it is only a normal fluctuation of a few internal operating indicators, the comprehensive deviation value will be smoothed and filtered out when other dimensions are stable, thus avoiding frequent false triggers of the system.

[0045] Furthermore, the technical advantage of customizing preset comprehensive thresholds based on enterprise strategic levels lies in adapting to the management granularity requirements of different levels. For example, provincial tobacco commercial enterprises, whose strategic focus is on macro trends and long-term planning, have a relatively high tolerance for environmental fluctuations. Therefore, the preset comprehensive threshold can be configured with a wider range (e.g., ±5%) to reduce unnecessary strategic fine-tuning and maintain strategic focus. Conversely, for city-level and lower-level enterprises, whose core responsibility is precise execution and implementation, the sensitivity to environmental changes is higher. Therefore, the preset comprehensive threshold can be configured with a narrower range (e.g., ±3%) to respond more quickly to subtle changes in the market and policies, ensuring agility at the execution level. This hierarchical threshold configuration mechanism allows the same intelligent construction system to flexibly adapt to the differentiated needs of different management levels, significantly improving the practicality and accuracy of strategic maps in real-world applications.

[0046] Please refer to Figure 2 Embodiment 3 of the present invention is as follows: A smart construction terminal 1 for a strategic map of a tobacco commercial enterprise includes a processor 2, a memory 3, and a computer program stored in the memory 3 and executable on the processor 2. When the processor 2 executes the computer program, it implements the steps in the smart construction method for a strategic map of a tobacco commercial enterprise described in Embodiment 1 or Embodiment 2 above.

[0047] This invention discloses an intelligent method for constructing strategic maps for tobacco commercial enterprises. By building a computational architecture that integrates semantic understanding and operations optimization, a first model performs semantic decomposition on multi-dimensional business data to extract strategic intent. Semantic entities are then mapped into structured constraint parameters containing boundary constraints and objective functions, which are then optimized and solved by a second model. This mechanism fundamentally solves the data gap problem between unstructured strategic text and structured mathematical models, ensuring that macro-level strategic intent can be accurately transformed into computable mathematical constraints. It avoids the resource constraint illusion caused by relying solely on generative models, thus improving the mathematical optimality and practical executability of the planning scheme.

[0048] Simultaneously, a closed loop for adaptability verification and negative feedback correction based on a domain rule base was established. When verification fails, the verification deviation items and correction suggestions are constructed as negative feedback prompts and injected into the first model as context input, triggering a re-semantic decomposition. This mechanism, which transforms structured verification deviations into natural language correction signals that the model can understand, enables automated iterative optimization of the planning process. It effectively overcomes the problem of model output being disconnected from specific domain business rules, significantly improving the adaptability and compliance of the strategic map to industry-specific requirements.

[0049] Furthermore, a weighted fusion of multi-source heterogeneous data is used to calculate a comprehensive deviation value to trigger dynamic adjustments. This comprehensively considers the interconnected impact of data from multiple dimensions, including policy changes, market sales, and internal operations, and allows for customized threshold configuration based on the enterprise's strategic level. Compared to simple threshold judgments based on a single indicator, this mechanism can more accurately characterize systemic changes in the internal and external environment, effectively filter out local noise interference, ensure the strategic map's sensitivity to environmental changes, avoid wasting system resources due to unnecessary frequent recalculations, and improve the robustness and intelligence of dynamic management.

[0050] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention specification and drawings, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for intelligently constructing a strategic map for tobacco commercial enterprises, characterized in that, Including the following steps: Acquire multi-dimensional business data from tobacco commercial enterprises, including policy guidance data, market operation data, and internal management data. The first model is used to semantically decompose the multidimensional business data, extract strategic intent information and business constraint information from the multidimensional business data, and obtain preliminary strategic elements. The semantic entities in the preliminary strategic elements are mapped to structured constraint parameters, and the second model is used to optimize and solve the solution based on the structured constraint parameters to obtain a planning scheme that includes resource allocation results and path planning results. The structured constraint parameters include boundary constraint conditions and objective functions; The adaptability of the planning scheme is verified based on the domain rule base, and the verification result is obtained. In response to the verification result indicating that the verification failed, a correction feedback signal is generated based on the verification deviation item and the corresponding correction suggestion, and the correction feedback signal is injected into the first model as a context input to trigger the first model to re-decompose semantics based on the correction feedback signal until the verification passes and the target strategic map data is output. In response to the detection that the data changes of the multidimensional business data meet the preset trigger conditions, the first model and the second model are triggered to re-execute the calculation to update the target strategic map data.

2. The intelligent construction method for a strategic map of a tobacco commercial enterprise according to claim 1, characterized in that, The first model is a large language model, and the second model is an operations research algorithm model; Mapping semantic entities in the preliminary strategic elements to structured constraint parameters includes: The semantic entities in the preliminary strategic elements are transformed into the input parameter format of the operations research algorithm model, wherein the boundary constraints are generated based on the resource constraint information and time constraint information in the semantic entities, and the objective function is generated based on the optimization objective information in the semantic entities.

3. The intelligent construction method for a strategic map of a tobacco commercial enterprise according to claim 2, characterized in that, The operations research algorithm model includes linear programming algorithms or genetic algorithms; The boundary constraints include total resource constraints, time node constraints, and compliance requirement constraints. The planning scheme is obtained by solving the second model based on the structured constraint parameters, including: The objective function is optimized under the boundary constraints by the operations research algorithm model to obtain resource allocation results and path planning results, and the resource allocation results and path planning results are used as the planning scheme.

4. The intelligent construction method for a strategic map of a tobacco commercial enterprise according to claim 1, characterized in that, The planning scheme is adapted based on the domain rule base, and the verification results are obtained, including: The planning scheme is matched and compared with the rule elements in the domain rule base to obtain the verification deviation item; Based on the verification deviation items and corresponding correction suggestions, a correction feedback signal is generated, including: In response to the existence of the verification deviation item, the verification deviation item and the corresponding correction suggestion are constructed as a negative feedback prompt word, and the negative feedback prompt word is used as the correction feedback signal.

5. The intelligent construction method for a strategic map of a tobacco commercial enterprise according to claim 4, characterized in that, The planning scheme is matched and compared with the rule elements in the domain rule base to obtain the verification deviation items, including: The planning scheme is compared with the rule elements from the dimensions of element completeness, business relevance rationality, and indicator quantification feasibility, respectively, and the deviation results of each dimension are obtained. In response to the existence of the aforementioned verification deviation item, the verification deviation item and its corresponding correction suggestion are constructed as negative feedback prompt words, including: In response to the deviation result in any dimension exceeding a preset deviation threshold, the deviation result corresponding to the dimension exceeding the preset deviation threshold and the correction suggestion are constructed as the negative feedback prompt word; The method further includes: In response to the number of iterations for semantic decomposition reaching a preset maximum number of iterations, or the deviation results of all dimensions being less than or equal to the preset deviation threshold, the iteration is terminated and the target strategic map data is output.

6. The intelligent construction method for a strategic map of a tobacco commercial enterprise according to claim 1, characterized in that, Monitoring whether the data changes of the multidimensional business data meet preset trigger conditions includes: Real-time acquisition of multi-source heterogeneous monitoring data, including policy change data, market sales data, and internal operation data; The multi-source heterogeneous monitoring data are weighted and fused to obtain a comprehensive deviation value; In response to the overall deviation value exceeding a preset overall threshold, it is determined that the data change meets the preset triggering condition.

7. The intelligent construction method for a strategic map of a tobacco commercial enterprise according to claim 6, characterized in that, The multi-source heterogeneous monitoring data are weighted and fused to obtain a comprehensive deviation value, including: Calculate the magnitude of change of the policy change data, the market sales data, and the internal operation data relative to the baseline data, respectively. Based on a preset weighting strategy, the variation ranges of various types of data are weighted and summed to obtain the comprehensive deviation value; The preset comprehensive threshold can be customized according to the enterprise's strategic level.

8. The intelligent construction method for a strategic map of a tobacco commercial enterprise according to claim 1, characterized in that, After outputting the target strategic map data, the method further includes: The target strategic map data is structured and organized according to multiple preset value dimensions, including financial value dimension, customer value dimension, core operating process dimension, and organizational capability support dimension. Under each of the aforementioned value dimensions, the corresponding strategic objectives, execution paths, responsible entities, time nodes, and quantitative indicators are associated to generate a visual strategic map; In response to receiving a hierarchical expansion command or an indicator penetration command, the corresponding level of sub-targets or underlying indicator data is displayed in the visualized strategic map.

9. The intelligent construction method for a strategic map of a tobacco commercial enterprise according to claim 1, characterized in that, It also includes the following steps: The transmission process of the multidimensional business data and the target strategic map data is encrypted using the SSL encryption protocol. The multidimensional business data and the target strategic map data are encrypted and stored using the AES-256 encryption algorithm; Based on the role-based access control mechanism, strategic map viewing, editing and approval permissions are assigned to roles with different permissions. Record all data access and operation behaviors to generate security audit logs.

10. A smart terminal for constructing a strategic map for a tobacco business enterprise, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps in the intelligent construction method for a strategic map of a tobacco commercial enterprise as described in any one of claims 1-9.

Images