Method and system for executing financial transaction using center agent and a plurality of special artificial intelligence agents

Abstract

A method for establishing an investment strategy is disclosed in a system comprising multiple specialized AI agents, wherein the system comprises the steps of receiving a user’s investment goal, a data fusion agent collecting and analyzing heterogeneous financial data and fusing them into a single comprehensive signal, the data fusion agent transmitting the single comprehensive signal to a strategy formulation agent, the strategy formulation agent generating a graph of thoughts (GoT) containing multiple financial thoughts as nodes based on the investment goal and the single comprehensive signal, the strategy formulation agent determining an investment strategy by transforming the nodes on the graph of thoughts, the strategy formulation agent transmitting the investment strategy to a transaction execution agent, and the transaction execution agent executing a transaction according to the investment strategy. The multiple specialized AI agents include a data fusion agent, a strategy formulation agent, a transaction execution agent, and a risk agent.

Description

Technology Field

[0001] The present invention relates to a method and system for establishing an investment strategy using a plurality of specialized AI agents, and more specifically, to a technology for establishing an investment strategy by analyzing heterogeneous financial data using the Dempster-Shafer Theory (DST), fusing them into a single fused signal. Background Technology

[0002] Recently, there have been active attempts in the financial investment sector to analyze the market using AI and machine learning. Robo-advisors utilize technologies such as recommending predefined portfolio models based on a user's risk tolerance or identifying investment sentiment by analyzing the sentiment of news articles or social media through natural language processing (NLP).

[0003] However, conventional technologies tend to rely primarily on structured data (e.g., stock prices and financial statements) or a single type of unstructured data (e.g., news sentiment). Currently, there is a lack of a robust mechanism to rationally fuse conflicting information from data sources to make consistent decisions when information conflicts—for example, technical analysis indicators might send a 'buy' signal while sentiment analysis sends a 'sell' signal due to negative news articles. Consequently, there is a need for technology that reliably fuses conflicting and uncertain multi-source financial data and generates optimal investment strategies through adaptive reasoning in a complex and dynamic financial environment. Prior art literature

[0004] Korean Registered Patent No. 10-2455325 (October 12, 2022) The problem to be solved

[0005] The present invention aims to provide a method and system for establishing an investment strategy using a plurality of professional AI agents. means of solving the problem

[0006] As a technical means for achieving the above-mentioned task, the first aspect of the present invention provides a method for establishing an investment strategy by a system comprising a plurality of specialized AI agents, wherein the system comprises the steps of receiving a user’s investment goal, a data fusion agent collecting and analyzing heterogeneous financial data and fusing them into a single comprehensive signal, the data fusion agent transmitting the single comprehensive signal to a strategy formulation agent, the strategy formulation agent generating a graph of thoughts (GoT) including a plurality of financial thoughts as nodes based on the investment goal and the single comprehensive signal, the strategy formulation agent determining an investment strategy by modifying the nodes on the graph of thoughts, the strategy formulation agent transmitting the investment strategy to a transaction execution agent, and the transaction execution agent executing a transaction according to the investment strategy. The plurality of specialized AI agents include a data fusion agent, a strategy formulation agent, a transaction execution agent, and a risk agent.

[0007] In addition, a data fusion agent can collect conflicting financial evidence from multiple heterogeneous data sources, assign a confidence level to each of the financial evidence, and use the Dempster-Schaffer Theory (DST) to fuse the confidence levels of each of the financial evidence and generate a single composite signal. The financial evidence may include technical analysis indicators, news sentiment scores, and economic indicators.

[0008] Additionally, the method may further include the step of a strategy formulation agent transmitting an investment strategy to a risk agent; the step of the risk agent verifying whether the investment strategy complies with predefined risk management rules; the step of, if the investment strategy complies with the risk management rules, causing a transaction execution agent to call a financial API to execute a financial transaction; and the step of, if the investment strategy violates the risk management rules, the step of the risk agent transmitting a message to the transaction execution agent to block the transaction.

[0009] In addition, the strategy formulation agent can merge nodes representing multiple inference paths into a single node.

[0010] In addition, the strategy formulation agent can transform the nodes to ultimately determine a single merged node as the investment strategy.

[0011] In addition, multiple specialized AI agents can interact by exchanging messages containing performatives defined in the FIPA-ACL (Foundation for Intelligent Physical Agents - Agent Communication Language) standard.

[0012] A second aspect of the present invention may provide a computer-readable storage medium having a program recorded thereon for executing the method of the first aspect on a computer. Effects of the invention

[0013] The present invention is a computer-implemented system composed of a collection of multiple AI agents, which has the advantage of reliably fusing conflicting and uncertain multi-source financial data, generating optimal investment strategies by adaptively reasoning in complex and dynamic financial environments, and making robust and scalable autonomous investment decisions through the interaction of multiple autonomous agents constituting the system in a structured manner.

[0014] In addition, fusing conflicting heterogeneous financial data through the Dempster-Schaffer theory (DST) has the advantageous effect of quantifying the level of conflict between data and explicitly modeling uncertainty to generate more reliable comprehensive investment signals. Brief explanation of the drawing

[0015] FIG. 1 is a drawing illustrating an example of a system for establishing an investment strategy according to one embodiment. FIG. 2 is a flowchart illustrating a method for establishing an investment strategy according to one embodiment. FIG. 3 is a diagram showing an example of the structure of a graph of accidents (GoT) according to one embodiment. FIG. 4 is a diagram showing an example of determining an investment strategy using a graph of accidents (GoT) according to one embodiment. FIG. 5 is a block diagram illustrating the configuration of a system according to one embodiment. Specific details for implementing the invention

[0016] The invention will be described in detail below by way of exemplary embodiments with reference to the attached drawings. The following embodiments are intended only to embody the invention and do not limit or restrict the scope of the invention. Anything that can be easily inferred by a person skilled in the art to which the invention pertains from the detailed description and embodiments shall be interpreted as falling within the scope of the invention.

[0017] Terms such as 'composed' or 'comprising' as used in this specification should not be interpreted as necessarily including all of the various components or steps, and should be interpreted as some of the components or steps may not be included, or additional components or steps may be included.

[0018] The terms used in this specification are described in their current, general form considering the functions mentioned herein; however, they may refer to various other terms depending on the intent of those skilled in the art, case law, the emergence of new technologies, etc. Therefore, the terms used in this specification should not be interpreted solely by their names, but should be interpreted based on the meaning of the terms and the overall content of this specification. Furthermore, singular expressions include a plural meaning unless the context clearly indicates a singular meaning.

[0019] The terms “above” and similar designations used in this specification (particularly in the claims) may indicate both singular and plural forms. Furthermore, unless there is a description explicitly specifying the order of the steps describing the method according to this specification, the described steps may be performed in a suitable order. The present invention is not limited by the order in which the described steps are described.

[0020] Phrases such as "in one embodiment" appearing in various places in this specification do not necessarily refer to the same embodiment.

[0021] Some embodiments of this specification may be represented by functional block configurations and various processing steps. Some or all of these functional blocks may be implemented by various numbers of hardware and / or software configurations that execute specific functions.

[0022] Furthermore, the connecting lines or connecting members between the components depicted in the drawings are merely illustrative of functional connections and / or physical or circuit connections. In the actual device, connections between components may be represented by various alternative or added functional connections, physical connections, or circuit connections.

[0023] The embodiments relate to a method and system for establishing an investment strategy using a plurality of professional AI agents, and detailed descriptions of matters widely known to those skilled in the art to which the following embodiments belong are omitted. The present invention will be described in detail below with reference to the attached drawings.

[0024] Preferred embodiments of the present invention will be described in detail below with reference to the drawings.

[0025] FIG. 1 is a diagram illustrating an example of a system for establishing an investment strategy according to one embodiment. Referring to FIG. 1, the system (100) may be connected to a personalized Agentic AI, an LLM model, a robo-advisor, and an authentication / KEY / Action.

[0026] The system (100) may be a computing device such as a server or a micro server, but is not limited thereto, and may include any type of device capable of transmitting and receiving information to and from other devices or systems to establish investment strategies and provide financial services.

[0027] The system (100) may be configured to include multiple specialized AI agents to handle interactions with other devices, formulate investment strategies, and call external APIs for the purpose of executing transactions. In one embodiment, the system (100) is a computer-implemented system composed of a collection of cooperative AI agents coordinated by a central Supervisor Agent. This architecture has the advantage of overcoming the limitations of a monolithic model and maximizing modularity and efficiency by breaking down complex financial tasks into specialized and manageable sub-tasks and assigning them to individual agents.

[0028] The central agent, functioning as the core control unit or brain of the system, receives the user's investment goals, breaks them down into a multi-stage plan, and can delegate sub-tasks to specialized agents. The central agent manages the entire workflow, status, and communication between agents, and can perform the role of an orchestration layer.

[0029] Multiple specialized AI agents may include data fusion agents, strategy formulation agents, trade execution agents, and risk agents.

[0030] The data fusion agent collects multi-source data, including technical analysis data, fundamental analysis data (e.g., corporate financial statements), macroeconomic indicators, and sentiment scores extracted from Korean and English news and social media. It evaluates the reliability of the collected heterogeneous data and fuses them into a single fused signal using the Demster-Schaffer theory. Since financial market forecasting inherently involves uncertainty, the Demster-Schaffer theory (DST) can be introduced to explicitly address this uncertainty.

[0031] In one embodiment, the data fusion agent considers heterogeneous data sources, such as technical indicators, fundamental analysis results, and news sentiment scores, as independent evidence. By applying the Dempster-Shafer Theory (DST), the degree to which each piece of evidence supports a specific hypothesis (e.g., 'buy', 'sell', 'hold') is quantified as a belief mass assignment, and these are fused using Dempster's rule of combination. Through this process, the level of conflict between data can be quantified, and uncertainty can be explicitly modeled to generate a more reliable comprehensive investment signal.

[0032] The process of generating a single composite signal using the Dempster-Schaffer theory (DST) is as follows.

[0033] 1. Frame of Discernment (Θ): First, define the entire set of mutually exclusive hypotheses. For example, for a specific stock 'Samsung Electronics', Θ can be defined as {Rise, Fall, Change}.

[0034] 2. Basic Probability Assignment (BPA): For each data source (evidence), a confidence level is assigned to each element of the power set of Θ.

[0035] - Evidence 1 (Technical Analysis): Occurrence of a moving average golden cross. This strongly supports the 'bullish' hypothesis. m1({bullish}) = 0.7, m1(Θ) = 0.3 (the remainder 0.3 is uncertainty).

[0036] - Evidence 2 (News Sentiment Analysis): Numerous negative news articles regarding the company occurred. This supports the 'decline' hypothesis. m2({decline}) = 0.6, m2(Θ) = 0.4.

[0037] 3. Dempster Combination Rule: Fuses two pieces of evidence to calculate a new confidence mass. This rule strengthens consensus between evidence and quantifies conflicts. m1 m2 (A) = (1 / (1-K)) * Σ(B ∩ C = A) m1(B) * m2(C) where K is the collision coefficient between the evidence.

[0038] 4. Result: The fused confidence mass reflects confidence in both 'up' and 'down', and a high collision coefficient (K) indicates that market uncertainty is very high. The data fusion agent delivers this final fusion result (e.g., m_final({Up})=0.4, m_final({Down})=0.3, m_final({Side})=0.1, m_final(Θ)=0.2) to the strategy formulation agent.

[0039] Based on the user's goals (e.g., long-term growth, dividend yield) and the comprehensive signal received from the data fusion agent, the strategy formulation agent generates an optimal investment portfolio composition and trading strategy using the graph of thought (GoT) inference method described below with reference to Fig. 3.

[0040] The risk agent verifies the risk tolerance pre-set by the user (e.g., maximum loss rate, maximum holding weight of specific assets) and compliance with regulations of domestic and international financial authorities for all trading plans proposed by the strategy formulation agent.

[0041] The transaction execution agent converts the finally approved transaction plan into a structured JSON format API call required by the transaction APIs of domestic and foreign financial institutions, and executes the order using an authenticated session through a secure credential management module. In one embodiment, transactions may be conducted across multiple asset classes (e.g., stocks, bonds, cryptocurrencies, derivatives, etc.) through multiple exchanges (e.g., securities firms, banks, insurance companies, cryptocurrency exchanges, etc.).

[0042] In addition, all interactions between agents take place through the FIPA-ACL protocol. Agents interact using a standardized agent communication language called FIPA-ACL, which enables the exchange of messages with clear semantics, such as inform, request, propose, agree, and refuse. For example, if a strategy formulation agent proposes a specific trading plan, the risk management agent evaluates it and responds with agree or refuse; the entire process can be transparently recorded and managed according to a standardized protocol.

[0043] FIG. 2 is a flowchart illustrating a method for establishing an investment strategy according to one embodiment.

[0044] In step S210, the user's investment goal can be entered. For example, the user's investment goal may be medium-risk long-term growth or dividend yield maximization.

[0045] In step S220, a data fusion agent can collect and analyze heterogeneous financial data and fuse them into a single comprehensive signal. In one embodiment, the data fusion agent collects conflicting financial evidence from a plurality of heterogeneous data sources, assigns a confidence level to each of the financial evidence, and uses the Dempster-Shafer Theory (DST) to fuse the confidence levels for each of the financial evidence and generate a single comprehensive signal. Additionally, the financial evidence may include technical analysis indicators, news sentiment scores, and economic indicators.

[0046] In step S230, the data fusion agent may transmit a single comprehensive signal to the strategy formulation agent. In one embodiment, multiple specialized AI agents may interact by exchanging messages containing performatives defined in the FIPA-ACL (Foundation for Intelligent Physical Agents - Agent Communication Language) standard. Multiple specialized AI agents may interact using a standardized agent communication language called FIPA-ACL, which enables the exchange of messages with clear semantics such as inform, request, propose, agree, and refuse. For example, the transmission of a single comprehensive signal by the data fusion agent to the strategy formulation agent may correspond to inform.

[0047] In step S240, based on an investment goal and a single comprehensive signal, a strategy formulation agent can generate a graph of accidents (GoT) containing multiple financial accidents as nodes. The structure of the graph of accidents (GoT) will be described later with reference to FIG. 3.

[0048] In step S250, the strategy formulation agent can determine an investment strategy by modifying nodes on the graph of thought. In one embodiment, the strategy formulation agent can merge nodes representing multiple inference paths into a single node. Additionally, the strategy formulation agent can determine the final merged single node as the investment strategy by modifying the nodes. An example of determining an investment strategy using the graph of thought (GoT) will be described later with reference to FIG. 4.

[0049] In step S260, the strategy formulation agent may transmit an investment strategy to the trade execution agent. For example, the transmission of an investment strategy by the strategy formulation agent to the trade execution agent may correspond to a request to execute a trade according to the investment strategy.

[0050] In step S270, the transaction execution agent can execute a transaction according to the investment strategy.

[0051] The trade execution agent can interact with third-party brokerage APIs. It can construct a valid JSON request body based on inferred trade decisions. For example, the decision "Buy 10 shares of AAPL at market price" is converted into the following JSON object: JSON {"symbol": "AAPL", "qty": "10", "side": "buy", "type": "market", "time_in_force": "day"}

[0052] Prior to step S270, a strategy formulation agent transmits an investment strategy to a risk agent, and the risk agent verifies whether the investment strategy complies with predefined risk management rules. If the investment strategy complies with the risk management rules, in step S270, the transaction execution agent may call a financial API to execute a financial transaction. However, if the investment strategy violates the risk management rules, the risk agent may send a message to the transaction execution agent to block the transaction. For example, if the strategy formulation agent proposes a specific transaction plan {action: 'BUY', symbol: '005930', quantity: 100}, the risk management agent evaluates this plan {proposal_id: 'xyz', status: 'approved'} and responds with agree or refuse, and the entire process can be transparently recorded and managed according to a standardized protocol.

[0053] For example, if an investment strategy violates risk management rules, the transaction may be automatically blocked, and the strategy formulation agent may be asked to re-establish the investment strategy or the user may be asked to provide explicit confirmation.

[0054] FIG. 3 is a diagram showing an example of the structure of a graph of accidents (GoT) according to one embodiment.

[0055] To construct a high-level portfolio and formulate an optimal strategy, a strategy formulation agent can generate a graph of thoughts (GoT) that includes multiple financial thoughts as nodes.

[0056] Existing AI reasoning methods have clear limitations. For example, Chain-of-Thought (CoT) has a linear structure that proceeds with reasoning along a single path according to a fixed order. This is akin to following a recipe in sequence, making it unsuitable for solving complex and multifaceted problems. Tree-of-Thought (ToT), an improvement over CoT, introduced a tree structure capable of exploring multiple reasoning paths; however, it remains limited to hierarchical and forward-looking exploration, thus having limitations in modeling interconnected thought processes.

[0057] The strategy formulation agent moves beyond reasoning methods limited to linear or tree structures and adopts a graph framework of thought that models the investment strategy formulation process as a dynamic graph. Each node represents a financial thought, such as "incorporating Korean blue-chip stocks," "increasing the weight of US tech stocks," or "hedging exchange rate risk." The strategy formulation agent connects these thoughts on the graph and synthesizes an optimal portfolio under complex constraints by merging different reasoning paths (e.g., combining growth and stability strategies) or iteratively improving plans through feedback loops for specific thoughts. This is very similar to the non-linear thinking processes of humans solving complex problems.

[0058] In one embodiment, a strategy formulation agent can perform graph of thought (GoT) inference based on a single aggregate signal received and user goals using the Dempster-Schaffer theory.

[0059] First, the strategy formulation agent dynamically constructs a graph with 'thoughts' as nodes and 'dependencies' as edges. Then, it creates thought nodes such as "Inclusion in Korean KOSPI 200 Index Tracking ETF," "Search for high-dividend stocks," "Adjusting weight of US IT sector," and "Assessment of global real estate sector risk." Subsequently, it connects the logical dependencies between each thought as edges.

[0060] Next, you can transform the graph and perform inference based on it.

[0061] - Information Aggregation (Aggregation): Combines multiple sub-thought nodes (e.g., analysis results of individual tech stocks) into a single top-level node (investment opinion on the IT sector).

[0062] - Path Merging: Merges two independently explored inference paths (e.g., the 'High Growth' path and the 'Low Volatility' path) to create a new node called the "Portfolio Combining Growth and Stability."

[0063] - Feedback Loop (Refinement): If the result of a specific node (global real estate sector risk assessment) is negative, this information is fed back to the parent node (asset allocation strategy) to dynamically modify the overall strategy.

[0064] The path with the highest score obtained through the graph exploration and transformation process is adopted as the final investment strategy.

[0065] FIG. 4 is a diagram showing an example of determining an investment strategy using a graph of accidents (GoT) according to one embodiment.

[0066] Referring to Figure 4, the result of fusing various data sources (e.g., technical indicators, news sentiment, etc.) using the Dempster-Schaffer theory (e.g., confidence level for “Recession” calculated as 0.6) and user goals (growth maximization, medium level of risk tolerance) are transmitted to a strategy formulation agent.

[0067] Based on a single comprehensive signal and the user's goals, the strategy formulation agent can construct a graph of thoughts. Here, various financial thoughts are formed as nodes: Node A is inclusion in the KOSPI 200 ETF, Node B is searching for high-dividend stocks, Node C is adjusting the weight of the US IT sector, and Node D is global real estate risk assessment. Logical dependencies between each thought are connected via edges, and the graph can be dynamically configured according to interrelationships (dependence, complementarity).

[0068] The nodes of the thought graph can be modified; for example, nodes that are less valid or conflict with the goal, such as pruned nodes, can be removed. Additionally, if the result of Node D (Global Real Estate Risk Assessment) is negative, this information can be fed back to the parent Node A (KOSPI 200 ETF Inclusion) to dynamically modify the overall strategy.

[0069] Additionally, nodes can be integrated to create Node E (focused search for US non-IT high-dividend stocks), which serves as an intermediate strategy node synthesizing the results of the preceding Nodes B and C, merging multiple paths (e.g., growth stocks, dividend stocks) into one to simultaneously consider growth potential and stability.

[0070] Afterwards, the graph is explored and the most promising path is selected from the node integration results to adopt a final strategy of 30% KOSPI, 40% US non-IT high-dividend stocks, … This represents a portfolio strategy derived through graph inference, reflecting the results of DST-based data fusion and user goals.

[0071] FIG. 5 is a block diagram illustrating the configuration of a system according to one embodiment.

[0072] As illustrated in FIG. 5, in one embodiment of the present invention, the system (100) may include a control unit (130), a communication unit (150), and a memory (170). However, not all of the illustrated components are essential components. The system (100) may be implemented with more components than illustrated, or with fewer components. The above components will be examined in turn below.

[0073] The control unit (130) typically controls the overall operation of the system (100). The control unit (130) may be equipped with at least one processor. Depending on its function and role, the control unit (130) may include multiple processors or a single processor in an integrated form. A processor may primarily refer to a central processing unit (CPU), an application processor (AP), a graphics processing unit (GPU), etc. Additionally, the CPU, AP, or GPU may include one or more cores within it, and the CPU, AP, or GPU may operate using an operating voltage and a clock signal. However, while the CPU or AP is composed of a few cores optimized for serial processing, the GPU may be composed of thousands of smaller and more efficient cores designed for parallel processing.

[0074] For example, the control unit (130) can control the output unit (not shown), the communication unit (150), etc., by executing programs stored in the memory (170). The control unit (130) can control the communication unit (150) to transmit and receive information with external devices (e.g., user device (not shown), secure storage (not shown), and financial institution server (not shown)).

[0075] The communication unit (150) may include one or more components that enable the system (100) to communicate with an external device. The external device may be a computing device such as the system (100), but is not limited thereto.

[0076] The memory (170) stores data that supports various functions of the control unit (130). The memory (170) can store a number of application programs (or applications) running on the control unit (130), data for the operation of the control unit (130), and instructions. At least some of these application programs may be downloaded from an external device or another server via wireless communication. Additionally, the application programs may be stored in the memory (170), installed on the control unit (130), and driven by a processor to perform the operation (or function) of the control unit (130).

[0077] The memory (170) may include at least one type of storage medium among flash memory type, hard disk type, SSD type (Solid State Disk type), SSD type (Silicon Disk Drive type), multimedia card micro type, card type memory (e.g., SD or XD memory, etc.), RAM (random access memory; RAM), SRAM (static random access memory), ROM (read-only memory; ROM), EEPROM (electrically erasable programmable read-only memory), PROM (programmable read-only memory), magnetic memory, magnetic disk, and optical disk. Additionally, the memory (170) may include web storage that performs storage functions over the internet.

[0078] A system (100) according to another embodiment of the present invention may include a plurality of specialized AI agents, such as a data fusion agent, a strategy formulation agent, a transaction execution agent, and a risk agent. The system (100) is implemented as an AI agent that performs tasks autonomously. The software module of the aforementioned agent may correspond to the hardware module of FIG. 5, for example, the control unit (130) may include a plurality of specialized AI agents, such as a data fusion agent, a strategy formulation agent, a transaction execution agent, and a risk agent.

[0079] The present invention described above can be modeled as computer-readable code on a medium on which a program is recorded. A computer-readable medium includes all types of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable media include HDD (Hard Disk Drive), SSD (Solid State Disk), SSD (Silicon Disk Drive), ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage device, etc., and also include being modeled in the form of a carrier wave (e.g., transmission over the Internet).

[0080] Meanwhile, the configuration of the system (100) illustrated in FIG. 5 is an example, and each component of the system (100) may be integrated, added, or omitted according to the specifications of the system (100) being implemented. That is, as needed, two or more components may be combined into a single component, or a single component may be subdivided into two or more components. Furthermore, the functions performed by each component (or module) are intended to explain the examples, and the specific operations or devices thereof do not limit the scope of the present invention.

[0081] Meanwhile, the embodiments of the present invention described above can be written as a program executable on a computer and can be implemented on a general-purpose digital computer that operates the program using a computer-readable recording medium.

[0082] The above computer-readable recording media includes storage media such as magnetic storage media (e.g., ROM, floppy disk, hard disk, etc.) and optical reading media (e.g., CD-ROM, DVD, etc.).

[0083] Although embodiments of the present invention have been described above with reference to the attached drawings, those skilled in the art will understand that the present invention may be implemented in other specific forms without changing its technical concept or essential features. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. Explanation of the symbols

[0085] 100: System 130: Control unit 150: Communications Department 170: Memory

Claims

Claim 1 A method for establishing an investment strategy by a system comprising multiple specialized AI agents coordinated by a supervisor agent, wherein the supervisor agent, which manages the entire workflow, state, and inter-agent communication and acts as an orchestration layer, receives a user's investment goal; wherein the supervisor agent breaks down the investment goal into a multi-stage plan; wherein the supervisor agent delegates sub-tasks to multiple specialized AI agents based on the broken-down multi-stage plan; wherein a data fusion agent collects and analyzes heterogeneous financial data based on the sub-tasks and fuses it into a single comprehensive signal; wherein the data fusion agent transmits the single comprehensive signal to a strategy formulation agent; wherein, based on the sub-tasks and the single comprehensive signal, the strategy formulation agent generates a Graph of Thoughts (GoT) containing multiple financial thoughts as nodes; wherein the strategy formulation agent determines an investment strategy by modifying the nodes on the graph of thoughts; and wherein the strategy formulation agent transmits the investment strategy to a risk agent. A method comprising: a step in which the risk agent verifies whether the investment strategy complies with predefined risk management rules; and a step in which the transaction execution agent executes a transaction according to the investment strategy based on the verification result of the risk agent; wherein a plurality of specialized AI agents include the data fusion agent, the strategy formulation agent, the transaction execution agent, and the risk agent. Claim 2 A method according to claim 1, wherein the step of the data fusion agent collecting and analyzing heterogeneous financial data and fusing them into a single composite signal comprises: the step of the data fusion agent collecting conflicting financial evidence from a plurality of heterogeneous data sources—the financial evidence including technical analysis indicators, news sentiment scores, and economic indicators—; the step of the data fusion agent assigning a reliability to each of the financial evidence; and the step of the data fusion agent fusing the reliability of each of the financial evidence using the Dempster-Shafer Theory (DST) and generating the single composite signal. Claim 3 A method according to claim 1, wherein the step of the transaction execution agent executing a transaction according to the investment strategy based on the verification result of the risk agent comprises: a step of causing the transaction execution agent to call a financial API to execute a financial transaction if the investment strategy complies with the risk management rule; and a step of the risk agent sending a message to the transaction execution agent to block the transaction if the investment strategy violates the risk management rule. Claim 4 A method according to claim 1, wherein the step of the strategy formulation agent determining an investment strategy by modifying the nodes on the graph of the thought includes the step of the strategy formulation agent merging nodes representing a plurality of inference paths into a single node. Claim 5 In claim 4, the step of the strategy formulation agent determining an investment strategy by modifying the nodes on the graph of the thought includes the step of the strategy formulation agent determining a single node that is finally merged by modifying the nodes as the investment strategy. Claim 6 A method according to claim 1, wherein the plurality of specialized AI agents interact by exchanging messages including performatives defined in the FIPA-ACL (Foundation for Intelligent Physical Agents - Agent Communication Language) standard. Claim 7 A computer-readable storage medium having a program recorded thereon for executing the method of any one of paragraphs 1 through 6 on a computer.

Images