Large language model aided optimization strategic decision-making system and method

Abstract

The invention discloses a large language model auxiliary optimization strategic decision-making system and a large language model auxiliary optimization strategic decision-making method. The system comprises six core modules. The dynamic knowledge fusion module constructs a three-layer distributed knowledge network, constructs an entity association weight matrix through a bidirectional Transform model based on an attention mechanism, and realizes knowledge dynamic association in combination with a time attenuation factor and a hybrid coding technology. The large language model module performs field fine tuning by adopting incremental pre-training and low-rank adaptation technologies, and introduces an exclusive word segmentation list to improve professional analysis precision. The full-process intelligent writing module covers submodules for report generation, revision and the like, and supports full-life-cycle management of reports. The strategic decision intelligent deduction module integrates scene impact factors, and realizes multi-scene deduction through reinforcement learning and Monte Carlo tree search. The interaction display module provides a visual interface, and the multi-mode interaction module realizes full task chain management. According to the invention, real-time knowledge support and intelligent deduction capability are provided for strategic decision making, and decision making efficiency and accuracy are improved.

Description

Technical Field

[0001] This document relates to the field of strategic decision-making technology, and in particular to a large language model-assisted optimization system and method for strategic decision-making. Background Technology

[0002] In today's complex and ever-changing market and policy environment, strategic decision-making across industries faces unprecedented challenges. From a macro perspective, the profound adjustments in the global economic landscape, the rapid iteration of emerging technologies, and frequent changes in policies and regulations have made it increasingly complex for enterprises and organizations to consider factors when planning their future strategic direction. Traditional strategic decision-making models, relying on expert experience and limited data analysis, have significant limitations. On the one hand, expert knowledge reserves struggle to keep pace with rapidly changing information, and individual cognitive biases can easily lead to one-sided decisions. For example, in the energy industry, relying solely on expert experience in the traditional energy market makes it difficult to accurately plan corporate energy strategies in the face of the rise of new energy sources and significant adjustments in energy policies. On the other hand, data collection and integration are fraught with difficulties. Different types of knowledge bases, such as policy databases, energy databases, general knowledge bases, and energy knowledge bases, operate independently, lacking efficient linkage mechanisms, resulting in information fragmentation.

[0003] While significant progress has been made in the fields of natural language processing and artificial intelligence, with models like ChatGPT demonstrating excellent performance in text generation and question-answering, their application in specialized strategic decision-making scenarios still faces numerous limitations. Existing large language models are mostly general-purpose models, lacking a deep understanding of the professional context and potential needs of strategic decision-making, and are unable to perform efficient knowledge learning and semantic parsing for specific industry-specific strategic decision-making scenarios. In the strategic decision-making simulation stage, traditional methods, mainly based on simple models and historical data, are insufficient to simulate complex and ever-changing real-world scenarios and cannot effectively cope with uncertainties such as sudden policy changes and market shifts. For example, in supply chain management, traditional simulation methods struggle to accurately predict the chain reactions caused by sudden changes in trade policies or disruptions in raw material supply, making it difficult for companies to develop flexible and effective response strategies in advance. Furthermore, the entire strategic research project process, from report processing to strategic decision-making, lacks an integrated collaborative management mechanism. Different stages are handled by different tools or personnel, resulting in a severe disconnect in the task chain, an inability to track task status in real time, and consequently, low project efficiency and high collaboration costs. Summary of the Invention

[0004] According to embodiments of the present invention, a large language model-assisted optimization strategic decision-making system and method are provided, aiming to solve the above-mentioned problems.

[0005] A large language model-assisted optimization strategy decision-making system according to an embodiment of the present invention includes: The dynamic knowledge fusion module consists of a distributed knowledge network comprising a policy base, an energy database, a general knowledge base, an energy knowledge base, and a strategic management model base. Each knowledge base achieves automatic cross-database knowledge association and dynamic updates through a semantic association engine, providing real-time updated integrated knowledge support for strategic decision-making and report processing. The large language model module, based on a locally deployed model architecture, forms the semantic parsing capability for strategic decision-making scenarios through continuous learning of multi-domain knowledge in the dynamic knowledge fusion module. The end-to-end intelligent writing and strategic collaboration module is used to achieve full lifecycle management of strategic reports; The interaction and display module provides a user interface that allows users to input commands and upload files, while also visually presenting system processing results and strategic decision-making information.

[0006] The intelligent strategic decision-making simulation module integrates auxiliary strategic decision-making and data visualization units, supporting multi-scenario strategic simulation and interactive parameter adjustment; The multimodal interaction and task management module provides multimodal entry points such as natural language interaction, file upload, and operation interface interaction, enabling full-chain status tracking and collaborative management from report processing to strategic decision-making. The dynamic knowledge fusion module's semantic association engine uses a bidirectional Transformer model based on an attention mechanism to construct a dynamically updated entity association weight matrix. The large language model module uses incremental pre-training and low-rank adaptation techniques for domain fine-tuning. The strategic decision-making intelligent inference module introduces scenario impact factors to simulate uncertainty inference scenarios.

[0007] The large language model-assisted optimization strategy decision-making method according to embodiments of the present invention includes: Construct a distributed knowledge network and achieve dynamic knowledge fusion. Build a knowledge network that includes a policy base, an energy database, a general knowledge base, an energy knowledge base, and a strategic management model base. Construct a semantic association engine through a bidirectional Transformer model based on an attention mechanism to achieve automatic cross-database knowledge association and dynamic updates, providing real-time updated integrated knowledge support for strategic decision-making and report processing. The localized deployment base model is subjected to domain adaptation learning, and continuous learning is carried out based on multi-domain knowledge obtained by dynamic knowledge fusion. Incremental pre-training and low-rank adaptation techniques are used for domain fine-tuning to form the semantic parsing capability for strategic decision-making scenarios. Implement full lifecycle management of strategic reports to achieve intelligent writing and strategic collaboration throughout the entire process of strategic report generation, revision, review and evaluation; It provides a user interaction and results display interface, supports user input commands and file uploads, and presents system processing results and strategic decision-making related information in a visual manner; Conduct intelligent simulations of strategic decision-making across multiple scenarios, integrate units for assisting strategic decision-making and data visualization, introduce scenario impact factors to simulate uncertainty simulation scenarios, and support multi-scenario strategic simulations and interactive parameter adjustments; It enables multimodal interaction and full task chain management, providing multimodal entry points such as natural language interaction, file upload, and operation interface interaction, and performs status tracking and collaborative management of the entire task chain from report processing to strategic decision-making.

[0008] In this embodiment of the invention, the dynamic knowledge fusion module enables automatic cross-database knowledge association and updating, providing real-time integrated knowledge support for decision-making. The large language model module accurately parses the semantics of strategic decision-making scenarios, improving the understanding of potential user commands. The end-to-end intelligent writing and collaboration module improves report generation efficiency and quality, strengthening the connection with the decision-making model. The interaction and display module presents information visually, allowing users to focus on key content and improving acquisition efficiency. The strategic decision-making intelligent deduction module supports multi-scenario deduction and parameter adjustment, enhancing the scientific nature of decision-making. The multimodal interaction and task management module enables collaborative management of the entire task chain, improving overall process efficiency. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in one or more embodiments of this specification or in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a schematic diagram of a large language model-assisted optimization strategic decision-making system according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating a large language model-assisted optimization strategy decision-making method according to an embodiment of the present invention. Detailed Implementation

[0011] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this document.

[0012] System Implementation Examples According to embodiments of the present invention, a large language model-assisted optimization strategic decision-making system is provided. Figure 1 This is a schematic diagram of a large language model-assisted optimization strategic decision-making system according to an embodiment of the present invention. Figure 1 As shown, the large language model-assisted optimization strategic decision-making system of this invention specifically includes: The dynamic knowledge fusion module consists of a distributed knowledge network comprising a policy base, an energy database, a general knowledge base, an energy knowledge base, and a strategic management model base. Each knowledge base achieves automatic cross-database knowledge association and dynamic updates through a semantic association engine, providing real-time updated integrated knowledge support for strategic decision-making and report processing. The dynamic knowledge fusion module is specifically used for: The association weights of historical knowledge are dynamically adjusted by a time decay factor to ensure the timeliness of knowledge association. When a new knowledge item is added, its semantic similarity with existing knowledge is automatically calculated and the association mapping is completed. At the same time, the historical association weights are dynamically adjusted in combination with the time decay factor. The decay coefficient of the time decay factor is dynamically and adaptively adjusted based on the update frequency of the knowledge domain. A higher decay rate is set for high-frequency update domains and a lower decay rate is set for low-frequency stable domains. A three-layer distributed knowledge network architecture is constructed, including a bottom data layer, an intermediate association layer, and a top service layer. The bottom data layer adopts a distributed storage architecture to support petabyte-level data expansion, the intermediate association layer implements deep entity encoding through a semantic association engine, and the top service layer provides standardized API interfaces for other modules to call.

[0013] In the three-layer distributed knowledge network architecture of the dynamic knowledge fusion module, the entity deep encoding of the intermediate association layer adopts a hybrid encoding method based on knowledge graph embedding and pre-trained language model vector fusion. Specifically, the structural feature vector of the entity in the knowledge graph and the semantic feature vector generated by the pre-trained language model from the entity description text are weighted and fused through a gating mechanism to generate a composite entity vector containing structural and semantic associations. The weight parameters of the gating mechanism are dynamically optimized through domain knowledge supervised sample training to make the fusion ratio of structural and semantic features adapt to the association characteristics of different knowledge types. At the same time, the domain-specific dynamic parameters are set with dynamic feature channels, and their temporal evolution is captured by a temporal convolutional network and integrated into the composite entity vector. The temporal convolutional network uses multi-scale convolutional kernels to capture the evolution features at different temporal granularities in parallel, and dynamically weights and integrates the multi-scale features through an attention mechanism, so that the entity association weight matrix can simultaneously reflect static semantic associations and dynamic parameter associations.

[0014] The large language model module, based on a locally deployed model architecture, forms the semantic parsing capability for strategic decision-making scenarios through continuous learning of multi-domain knowledge from the dynamic knowledge fusion module; it adopts a model parameter sharding storage and distributed inference architecture, which ensures the security of local deployment while supporting parallel model computing to improve inference efficiency.

[0015] The large language model module is specifically used for: On the basis of the localized deployment model architecture, a dynamic fine-tuning dataset is built for the professional fields of energy strategy, policy and regulations. The dynamic fine-tuning dataset adopts a dual-track construction mode of "core domain knowledge + real-time updated cases". The core knowledge layer contains basic domain concepts and classic cases, and the real-time update layer connects to the new knowledge added by the dynamic knowledge fusion module, and automatically expands the dataset on a weekly basis. By freezing most of the model's parameters and updating only a small number of adaptation layer parameters through low-rank adaptation technology, efficient and continuous learning of knowledge from multiple domains can be achieved. The rank parameter of the low-rank adaptation matrix is ​​dynamically adjusted according to the complexity of the domain knowledge. Higher rank values ​​are set for complex domains such as energy strategy to retain more parameter update space, while lower rank values ​​are set for structured domains such as policies and regulations to achieve lightweight adaptation. At the same time, parameter distillation technology is used to merge the adaptation layer parameters that have been fine-tuned multiple times into an incremental update package to reduce model storage redundancy. A dedicated word segmentation table for strategic decision-making scenarios is introduced to improve the parsing accuracy of professional terms and contexts, and enhance the ability to capture potential needs in users' natural language commands. This dedicated word segmentation table adopts a hierarchical structure of "domain terminology library + scenario context lexicon." The domain terminology library includes professional terms and abbreviation mappings such as energy technology and policy provisions. The scenario context lexicon constructs context-related word sets for sub-scenarios such as strategic planning and risk assessment. A dynamic word frequency threshold mechanism enables automatic discovery of new words and incremental updates to the word segmentation table. When the frequency of a professional term in the input text exceeds a preset threshold, the word segmentation table update process is automatically triggered, and the word embedding vectors are retrained. The end-to-end intelligent writing and strategic collaboration module is used to achieve full lifecycle management of strategic reports. Specifically, this module includes: a report generation submodule, an intelligent revision submodule, a structured review submodule, and a multi-dimensional evaluation submodule. The report generation submodule automatically generates professional reports based on user themes and related knowledge from the dynamic knowledge fusion module, and the generated content can be used as decision analysis material by the strategic management model library. The intelligent revision submodule learns the professional expression standards of strategic reports while correcting text errors. The structured review submodule performs structural verification based on the industry standard outline in the dynamic knowledge fusion module. The evaluation results of the multi-dimensional evaluation submodule are directly fed back to the strategic decision-making support module as a decision-making reference.

[0016] The report generation submodule is also used for: By embedding dynamic knowledge reference tags, policy provisions and energy data involved in the generated content can be directly traced back to the corresponding knowledge base; By using controllable generation technology to constrain the text generation logic, logical coherence between report chapters can be ensured.

[0017] The interaction and display module provides a user interface that allows users to input commands and upload files, while also visually presenting system processing results and strategic decision-making information.

[0018] The intelligent strategic decision-making simulation module integrates a strategic decision-making support and data visualization unit, supporting multi-scenario strategic simulation and interactive parameter adjustment; specifically, the intelligent strategic decision-making simulation module is used for: By employing a combination of reinforcement learning and Monte Carlo tree search, the model parameters in the strategic management model library are transformed into an inference state space; Using the data analysis results from the dynamic knowledge fusion module as the initial state input, the strategic benefit function is constructed to evaluate each simulation path; The system introduces scenario-based impact factors to simulate uncertainties such as sudden policy changes and market shifts, enabling strategic simulations across multiple scenarios. An interactive simulation sandbox is built using D3.js, which allows users to drag and adjust key parameters to update the simulation path and result prediction in real time.

[0019] The multimodal interaction and task management module provides multimodal entry points such as natural language interaction, file upload, and operation interface interaction, enabling full-chain status tracking and collaborative management from report processing to strategic decision-making. The multimodal interaction and task management module is specifically used for: A task chain engine based on a microservice architecture is adopted to break down report processing and strategic decision-making into reusable service units; Define the dependencies and flow rules between tasks using the BPMN process modeling tool, and use message queues to achieve real-time synchronization of task status. A task progress prediction model is constructed, which uses an LSTM time series prediction network combined with task complexity coefficients to estimate the remaining time based on historical task execution data and push reminders for key nodes.

[0020] The interaction and display module is specifically used for: We employ WebGL and data-driven visualization technology to construct a multi-view interactive interface; we generate instruction intent vectors in real time through a natural language understanding interface and match and recommend them with system functional modules. It uses dynamic force-directed graphs to display knowledge association networks and timeline heatmaps to present strategic data trends, allowing users to focus on key information through box selection and zooming interactive operations.

[0021] The dynamic knowledge fusion module's semantic association engine uses a bidirectional Transformer model based on an attention mechanism to construct a dynamically updated entity association weight matrix. The large language model module uses incremental pre-training and low-rank adaptation techniques for domain fine-tuning. The strategic decision-making intelligent inference module introduces scenario impact factors to simulate uncertainty inference scenarios.

[0022] By employing the embodiments of the present invention, the following beneficial effects are achieved: The dynamic knowledge fusion module enables automatic cross-database knowledge association and updates, providing real-time integrated knowledge support for decision-making. The large language model module accurately parses the semantics of strategic decision-making scenarios, enhancing the understanding of potential user commands and needs. The end-to-end intelligent writing and collaboration module improves report generation efficiency and quality, strengthening the connection with decision-making models. The interaction and display module presents information visually, allowing users to focus on key content and improving acquisition efficiency. The strategic decision-making intelligent deduction module supports multi-scenario deduction and parameter adjustment, enhancing the scientific nature of decision-making. The multimodal interaction and task management module enables collaborative management of the entire task chain, improving overall process efficiency.

[0023] Method Implementation Examples According to embodiments of the present invention, a method for optimizing strategic decision-making using a large language model is provided. Figure 2 This is a flowchart of a large language model-assisted optimization strategy decision-making method according to an embodiment of the present invention. Figure 2 As shown, the large language model-assisted optimization strategy decision-making method of this invention specifically includes: S1. Construct a distributed knowledge network and achieve dynamic knowledge fusion. Build a knowledge network that includes a policy base, an energy database, a general knowledge base, an energy knowledge base, and a strategic management model base. Construct a semantic association engine through a bidirectional Transformer model based on an attention mechanism to achieve automatic cross-database knowledge association and dynamic updates, providing real-time updated integrated knowledge support for strategic decision-making and report processing. S2. Perform domain adaptation learning on the localized deployment base model, carry out continuous learning based on multi-domain knowledge obtained by dynamic knowledge fusion, and use incremental pre-training and low-rank adaptation techniques to fine-tune the domain and form the semantic parsing capability for strategic decision-making scenarios. S3. Implement full lifecycle management of strategic reports to achieve intelligent writing and strategic collaboration throughout the entire process of strategic report generation, revision, review and evaluation; S4. Provides a user interaction and result display interface, supports user input commands and file uploads, and presents system processing results and strategic decision-making related information in a visual manner; S5. Conduct intelligent simulation of strategic decision-making in multiple scenarios, integrate auxiliary strategic decision-making and data visualization units, introduce scenario impact factors to simulate uncertainty simulation scenarios, and support multi-scenario strategic simulation and interactive parameter adjustment; S6 enables multimodal interaction and full task chain management, providing multimodal entry points such as natural language interaction, file upload, and operation interface interaction, and performing status tracking and collaborative management of the entire task chain from report processing to strategic decision-making.

[0024] This embodiment is a method embodiment that corresponds one-to-one with the above system embodiments. For specific implementation, please refer to the above system embodiments, which will not be repeated here.

[0025] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A large language model-assisted strategic decision-making system, characterized in that... include: The dynamic knowledge fusion module consists of a policy base, an energy database, a general knowledge base, an energy knowledge base, and a strategic management model base, forming a distributed knowledge network. The distributed knowledge network achieves automatic cross-database knowledge association and dynamic updates through a semantic association engine. The large language model module continuously learns the domain knowledge involved in the dynamic knowledge fusion module based on a locally deployed model architecture. The end-to-end intelligent writing and strategic collaboration module is used to achieve full lifecycle management of strategic reports; The interaction and display module provides a user interface that allows users to input commands, upload files, and present system processing results and strategic decision-making information in a visual manner. The intelligent strategic decision-making simulation module integrates auxiliary strategic decision-making and data visualization units, supporting multi-scenario strategic simulation and interactive parameter adjustment; The multimodal interaction and task management module provides multimodal entry points such as natural language interaction, file upload, and operation interface interaction, enabling full-chain status tracking and collaborative management from report processing to strategic decision-making. The dynamic knowledge fusion module's semantic association engine uses a bidirectional Transformer model based on an attention mechanism to construct a dynamically updated entity association weight matrix. The large language model module uses incremental pre-training and low-rank adaptation techniques for domain fine-tuning. The strategic decision-making intelligent inference module introduces scenario impact factors to simulate uncertainty inference scenarios.

2. The system according to claim 1, characterized in that, The dynamic knowledge fusion module is specifically used for: The association weights of historical knowledge are dynamically adjusted by the time decay factor to ensure the timeliness of knowledge association. When a new knowledge item is added, its semantic similarity with existing knowledge is automatically calculated and the association mapping is completed. At the same time, the historical association weights are dynamically adjusted by combining the time decay factor. A three-layer distributed knowledge network architecture is constructed, including a bottom data layer, an intermediate association layer, and a top service layer. The bottom data layer adopts a distributed storage architecture to support petabyte-level data expansion, the intermediate association layer implements deep entity encoding through a semantic association engine, and the top service layer provides standardized API interfaces for other modules to call.

3. The system according to claim 2, characterized in that, In the three-layer distributed knowledge network architecture, the entity deep encoding of the intermediate association layer adopts a hybrid encoding method based on knowledge graph embedding and pre-trained language model vector fusion. Specifically, the structural feature vector of the entity in the knowledge graph and the semantic feature vector generated by the pre-trained language model of the entity description text are weighted and fused through a gating mechanism to generate a composite entity vector containing structural association and semantic association. Dynamic feature channels are set with domain-specific dynamic parameters, and their temporal evolution is captured by a temporal convolutional network and integrated into the composite entity vector.

4. The system according to claim 1, characterized in that, The large language model module is specifically used for: Based on the localized deployment of the basic model architecture, a dynamically evolving fine-tuning dataset is constructed for the professional fields of energy strategy and policy regulations; By applying low-rank adaptation techniques, the original parameters are frozen, and only the newly added low-rank matrix is ​​optimized to adapt to downstream tasks, so as to achieve efficient and continuous learning of knowledge from multiple domains. We introduce a word segmentation table specifically for strategic decision-making scenarios to improve the parsing accuracy of professional terms and domain-specific contexts.

5. The system according to claim 1, characterized in that, The full-process intelligent writing and strategic collaboration module specifically includes: a report generation submodule, an intelligent revision submodule, a structured review submodule, and a multi-dimensional evaluation submodule; The report generation submodule automatically generates professional reports based on the user's topic and the associated knowledge of the dynamic knowledge fusion module, and the generated content can be called by the strategic management model library as decision analysis material. The intelligent revision submodule is used to simultaneously learn the professional expression standards of strategic reports while correcting text errors; The structured auditing submodule performs structure verification based on the industry standard outline in the dynamic knowledge fusion module; The evaluation results of the multi-dimensional evaluation submodule are directly fed back to the strategic decision-making support module as a reference for decision-making.

6. The system according to claim 1, characterized in that, The strategic decision-making intelligent deduction module is specifically used for: By employing a combination of reinforcement learning and Monte Carlo tree search, the model parameters in the strategic management model library are transformed into an inference state space; Using the data analysis results from the dynamic knowledge fusion module as the initial state input, the strategic benefit function is constructed to evaluate each simulation path; The system introduces scenario-based impact factors to simulate uncertainties such as sudden policy changes and market shifts, enabling strategic simulations across multiple scenarios. An interactive simulation sandbox is built using D3.js, which allows users to drag and adjust key parameters to update the simulation path and result prediction in real time.

7. The system according to claim 1, characterized in that, The interaction and display module is specifically used for: We employ WebGL and data-driven visualization technology to construct a multi-view interactive interface; we generate instruction intent vectors in real time through a natural language understanding interface and match and recommend them with system functional modules. It uses dynamic force-directed graphs to display knowledge association networks and timeline heatmaps to present strategic data trends, allowing users to focus on key information through box selection and zooming interactive operations.

8. The system according to claim 1, characterized in that, The multimodal interaction and task management module is specifically used for: A task chain engine based on a microservice architecture is adopted to break down report processing and strategic decision-making into reusable service units; Define the dependencies and flow rules between tasks using the BPMN process modeling tool, and use message queues to achieve real-time synchronization of task status. Build a task progress prediction model to estimate the remaining time based on historical task execution data and push reminders for key milestones.

9. The system according to claim 5, characterized in that, The report generation submodule is also used for: By embedding dynamic knowledge reference tags, policy provisions and energy data involved in the generated content can be directly traced back to the corresponding knowledge base; By using controllable generation technology to constrain the text generation logic, logical coherence between report chapters can be ensured.

10. A method for large language model-assisted optimization strategic decision-making based on the large language model-assisted optimization strategic decision-making system according to any one of claims 1-9, characterized in that, include: Construct a distributed knowledge network and achieve dynamic knowledge fusion. Build a knowledge network that includes a policy base, an energy database, a general knowledge base, an energy knowledge base, and a strategic management model base. Construct a semantic association engine through a bidirectional Transformer model based on an attention mechanism to achieve automatic cross-database knowledge association and dynamic updates, providing real-time updated integrated knowledge support for strategic decision-making and report processing. The localized deployment base model is subjected to domain adaptation learning, and continuous learning is carried out based on multi-domain knowledge obtained by dynamic knowledge fusion. Incremental pre-training and low-rank adaptation techniques are used for domain fine-tuning to form the semantic parsing capability for strategic decision-making scenarios. Implement full lifecycle management of strategic reports to achieve intelligent writing and strategic collaboration throughout the entire process of strategic report generation, revision, review and evaluation; It provides a user interaction and results display interface, supports user input commands and file uploads, and presents system processing results and strategic decision-making related information in a visual manner; Conduct intelligent simulations of strategic decision-making across multiple scenarios, integrate units for assisting strategic decision-making and data visualization, introduce scenario impact factors to simulate uncertainty simulation scenarios, and support strategic simulations and interactive parameter adjustments across multiple scenarios; It enables multimodal interaction and full task chain management, providing multimodal entry points such as natural language interaction, file upload, and operation interface interaction, and performs status tracking and collaborative management of the entire task chain from report processing to strategic decision-making.

Images