Method and device for constructing intelligent agent, electronic equipment and storage medium

Abstract

The present disclosure discloses a method and device for constructing an intelligent agent, electronic equipment and storage medium, and relates to the technical field of computers. The specific implementation scheme is: the intelligent agent is respectively subjected to context learning of business index information of a target field, concept information of the target field, and abnormal change analysis methodology, so that the business index interpretation module in the intelligent agent is used for interpreting business indexes, the concept network module is used for understanding the concept information of the target field, and the data analysis methodology module is used for learning abnormal change analysis methods; the intelligent agent is subjected to context learning of a thinking chain template for abnormal change analysis of the business index, so that the index analysis module of the intelligent agent is used for abnormal change analysis of the business index based on the thinking chain template.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, specifically to the field of artificial intelligence technology such as machine learning and intelligent agent construction, and in particular to a method, apparatus, electronic device and storage medium for constructing an intelligent agent. Background Technology

[0002] Currently, we are in a data-driven digital economy era. Breakthroughs in artificial intelligence (AI) technology, especially the emergence of large language models, are profoundly changing the data processing and analysis paradigms across various industries. In this wave, intelligent agents, as artificial intelligence entities capable of perceiving the environment, understanding objectives, monitoring real-time data, attributing business anomalies, gaining insights into user behavior, and supporting strategic decision-making, are seen as realizing refined operational planning steps and executing tasks to achieve specific goals. They have rapidly moved from theoretical research to commercial applications.

[0003] In data-intensive, complex, and highly competitive business environments, intelligent agents have demonstrated enormous application potential. However, despite the high market enthusiasm, intelligent agents that can truly perform complex analytical tasks stably, reliably, and efficiently in real-world business scenarios and produce results that directly drive business decision-making are still extremely rare. Summary of the Invention

[0004] This disclosure provides a method, apparatus, electronic device, and storage medium for constructing an intelligent agent.

[0005] According to one aspect of this disclosure, a method for constructing an intelligent agent is provided, comprising:

[0006] The agent is subjected to contextual learning of business indicator information, conceptual information of the target domain, and anomaly analysis methodology in the target domain, so that the business indicator interpretation module in the agent is used to interpret business indicators, the concept network module is used to understand the conceptual information of the target domain, and the data analysis methodology module is used to learn the anomaly analysis method.

[0007] The agent learns the context of a thought chain template for analyzing the anomalies of the business metrics, so that the agent's metric analysis module can perform anomaly analysis on the business metrics based on the thought chain template.

[0008] According to another aspect of this disclosure, an apparatus for constructing an intelligent agent is provided, comprising:

[0009] The first learning module is used to perform contextual learning on the agent for business indicator information, conceptual information of the target domain, and anomaly analysis methodology in the target domain, so that the business indicator interpretation module in the agent can be used to interpret business indicators, the concept network module can be used to understand the conceptual information of the target domain, and the data analysis methodology module can be used to learn the anomaly analysis method.

[0010] The second learning module is used to learn the context of the thought chain template for the intelligent agent regarding the anomaly analysis of the business indicators, so that the indicator analysis module of the intelligent agent can perform anomaly analysis of the business indicators based on the thought chain template.

[0011] According to another aspect of this disclosure, an electronic device is provided, comprising:

[0012] At least one processor; and

[0013] A memory communicatively connected to the at least one processor; wherein,

[0014] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the methods described above and any possible implementations.

[0015] According to another aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions for causing the computer to perform the methods described above and any possible implementation thereof.

[0016] According to another aspect of this disclosure, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the aspects and any possible implementations described above.

[0017] According to the technology disclosed herein, the accuracy of the constructed intelligent agent can be effectively improved.

[0018] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0019] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0020] Figure 1 This is a schematic diagram based on the first embodiment of the present disclosure;

[0021] Figure 2This is a schematic diagram according to the second embodiment of the present disclosure;

[0022] Figure 3 This is a schematic diagram according to the third embodiment of the present disclosure;

[0023] Figure 4 This is a schematic diagram according to the fourth embodiment of the present disclosure;

[0024] Figure 5 This is a block diagram of an electronic device used to implement the methods of the embodiments of this disclosure. Detailed Implementation

[0025] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0026] Obviously, the described embodiments are only some, not all, of the embodiments disclosed herein. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without inventive effort are within the scope of protection of this disclosure.

[0027] It should be noted that the terminal devices involved in the embodiments of this disclosure may include, but are not limited to, smart devices such as mobile phones, personal digital assistants (PDAs), wireless handheld devices, and tablet computers; the display devices may include, but are not limited to, personal computers, televisions, and other devices with display functions.

[0028] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0029] The construction of an intelligent agent is the core factor determining its upper limit of capabilities and application effectiveness. Currently, the mainstream construction method relies on the "end-to-end prompting engineering" method, which depends on precisely designed prompt words. This construction method is a "one-time" construction, resulting in insufficient learning of the intelligent agent and low accuracy in corresponding business analysis.

[0030] Figure 1 This is a schematic diagram based on the first embodiment of the present disclosure; as shown Figure 1As shown, this embodiment provides a method for constructing an intelligent agent, which may specifically include the following steps:

[0031] S101. Perform context learning on the intelligent agent for business indicator information, conceptual information of the target domain, and anomaly analysis methodology, respectively, so that the business indicator interpretation module in the intelligent agent can be used to interpret business indicators, the concept network module can be used to understand the conceptual information of the target domain, and the data analysis methodology module can be used to learn the anomaly analysis method.

[0032] In this embodiment, the context learning of business indicator information, conceptual information, and anomaly analysis methodology for the intelligent agent can be performed in parallel or sequentially according to a certain order.

[0033] In this embodiment, in-context learning (ICL) of an agent refers to the ability to understand and perform new tasks by means of only a few examples, instructions or background information provided in the input prompt, without updating the model parameters within the agent.

[0034] Specifically, the agent performs contextual learning of business indicator information for the target domain. In practice, this can be achieved by injecting business indicator information of the target domain into the agent, so that the business indicator interpretation module in the agent can interpret the business indicators. In subsequent work, the business indicator interpretation module can interpret the business indicators.

[0035] Contextual learning of conceptual information in the target domain can be performed on the intelligent agent. Specifically, this can be achieved by injecting conceptual information from the target domain into the intelligent agent so that the corresponding conceptual network module in the intelligent agent can understand the conceptual information of the target domain.

[0036] The agent can learn the context of the anomaly analysis methodology. In practice, this can be achieved by injecting the anomaly analysis methodology into the agent so that the agent's data analysis methodology module can be used to learn the anomaly analysis method.

[0037] In other words, in this embodiment, the constructed intelligent agent can be used to perform anomaly analysis on business metrics in the target domain. For ease of management and analysis, this intelligent agent can include many modules, such as a business metric interpretation module, a core concept network module, a data analysis methodology module, and a metric analysis module, etc.

[0038] In this embodiment, during the construction of the intelligent agent, business indicator information of the target domain can be injected into the intelligent agent first. In this way, the business indicator interpretation module in the intelligent agent can understand the business indicators of the target domain based on the injected business indicator information, and then analyze the anomalies of the business indicators.

[0039] In this embodiment, business metrics may include some core business metrics for business analysis in the target domain. Since the nature of business differs across domains, the business metrics for different domains may also differ.

[0040] In this embodiment, the conceptual information of the target domain can be some unique concepts in the target domain, which can help the intelligent agent understand some proprietary concepts in the target domain, and thus enable more accurate anomaly analysis.

[0041] In this embodiment, the methodology for anomaly analysis can be the methods or approaches used when assisting an intelligent agent in performing anomaly analysis.

[0042] S102. Context learning of the thought chain template for the intelligent agent to analyze changes in business metrics, so that the intelligent agent's metric analysis module can be used to analyze changes in business metrics based on the thought chain template.

[0043] In practice, a thought chain template for analyzing changes in business metrics can be embedded into the agent, so that the agent's metric analysis module can learn to perform anomaly analysis on business metrics based on the thought chain template.

[0044] The thought chain template in this embodiment can be used to guide the reasoning logic of the agent when performing anomaly analysis, enabling the agent to learn how to perform multi-step reasoning at the thinking level when performing anomaly analysis, so as to obtain the final anomaly analysis results of business indicators.

[0045] The agent construction method in this embodiment, by employing the aforementioned layered construction, enables the agent to learn rich knowledge such as business indicator information, conceptual information, anomaly analysis methodology, and anomaly analysis thought chain in the target domain in a hierarchical and comprehensive manner. This improves the agent's professionalism, enhances its intelligence, and effectively increases the accuracy of the constructed agent. Furthermore, compared to existing manually constructed agents, it significantly improves the efficiency of agent construction; compared to existing end-to-end constructed agents, it allows the agent to learn richer knowledge in a hierarchical manner, enhancing its intelligence and enabling it to perform anomaly analysis more accurately and effectively.

[0046] Figure 2 This is a schematic diagram based on the second embodiment of this disclosure; the method for constructing an intelligent agent in this embodiment, as described above... Figure 1Based on the technical solutions of the illustrated embodiments, the technical solutions of this disclosure will be further and more effectively introduced. For example... Figure 2 As shown, the method for constructing an intelligent agent in this embodiment may specifically include the following steps:

[0047] S201. Define the intelligent agent in the target domain;

[0048] Defining the target domain for the intelligent agent can be considered the starting point for agent construction. The goal is to define its "digital identity" and establish the basic principles of its behavior, similar to defining job responsibilities and code of ethics for new employees. This stage is completed in the enterprise's data management platform or a dedicated AI agent factory, implemented through a structured configuration interface.

[0049] Defining an intelligent agent for a target domain, the core configuration elements include:

[0050] The name can be given to the intelligent agent with a clear and business-oriented name, such as "E-commerce Core Indicator Anomaly Analysis Assistant" or "Growth Insight Intelligent Agent".

[0051] The introduction / role description can be a concise summary of the role, such as: "This is a data analysis intelligence agent focused on monitoring and root cause analysis of core business metrics for e-commerce platforms. Its responsibilities are to objectively, rigorously, and efficiently analyze data anomalies and provide actionable business recommendations."

[0052] A unique identifier is used to assign a unique identity (Identity; Id) within the system for subsequent calls, management, and version tracking.

[0053] By adopting the above method, intelligent agents in the target domain can be accurately and clearly defined.

[0054] S202. Perform context learning on the defined agent about a preset question base, which includes multiple query examples from the target domain;

[0055] In practice, a pre-defined question library can be injected into the defined agent to enable the agent to learn the context of the pre-defined question library.

[0056] The question library in this embodiment can also be called a recommended question library. It can include a set of high-frequency and typical analytical query examples in the target field, serving as the starting point for user interaction with the intelligent agent and lowering the barrier to entry. For example: "What changes have occurred in yesterday's transaction DAU?", "Analyze the growth trend and driving factors of this week's payment GMV", "Compared to last month, which stage of the new customer conversion funnel has the most severe loss this month?", etc.

[0057] Specifically, in this embodiment, a question library is injected into the intelligent agent. This question library can be displayed on the interface when the user uses the device, recommending questions for reference. In this way, the user can know which analytical queries they can ask the intelligent agent based on the displayed question library, effectively guiding the user to ask questions and enhancing user convenience.

[0058] Alternatively, on the other hand, the question database in this embodiment can also be used by the intelligent agent to first detect whether the user's question is related to any question in the question database during anomaly analysis. If it is related, it can be determined that the question belongs to the anomaly analysis problem in the target domain, and subsequent anomaly analysis can be performed; otherwise, if the user's question is not related to any question in the question database, it is considered that the question does not belong to the anomaly analysis problem in the target domain, and other intelligent agents, such as general intelligent agents, can be used for fallback analysis. That is, by adopting this step, the intelligent agent can also perform question detection before anomaly analysis, which can effectively improve the accuracy and efficiency of question processing.

[0059] S203, Configure capability and boundary declaration information for the agent;

[0060] By configuring capability and boundary declaration information, capabilities and boundaries can be declared for the agent to clearly inform the user of the agent's capability scope and limitations. For example: "This agent focuses on descriptive and diagnostic analysis based on existing data and does not perform predictive modeling or act as a customer service representative."

[0061] By adopting the above methods, it is possible to accurately and reasonably describe the capabilities and define the boundaries of intelligent agents.

[0062] S204. Configure behavioral norm information for the intelligent agent;

[0063] The behavioral guidelines in this embodiment may include basic behavioral guidelines and output principles. In this embodiment, a set of mandatory "meta-hints" can be designed, which are injected at the underlying level as part of the system prompt each time the agent is invoked, but are not visible to the end user. This set of meta-hints establishes the "constitutional" principles for all agent outputs, and may include at least one of the following:

[0064] 1) Configure the agent with authentic behavioral norms so that the agent can use them for analysis based on the provided data;

[0065] The authenticity behavior guidelines define the principle of authenticity for intelligent agents, prohibiting the fabrication or assumption of data. Specifically, this can be achieved by injecting authenticity behavior guidelines into the intelligent agent. For example, the injected information could be something like, "All your analyses and conclusions must be strictly based on the provided data; you must not fabricate or assume data," etc.

[0066] 2) Configure traceability principle behavioral norms for intelligent agents so that the intelligent agents have traceability of the data being analyzed;

[0067] The traceability principle defines the traceability of an agent. In practice, this can be achieved by injecting traceability principle behavioral guidelines into the agent. For example, the injected information could be something like, "When analyzing specific data points, their source or calculation method should be noted, and the original data file should be provided in the attachment," etc.

[0068] 3) Configure the agent with structured output principles and behavioral guidelines so that the agent can use them to output the structure of the analysis report; the structure of the analysis report includes at least one of the following: report title information, core conclusion summary information, key data discovery information, dimensional breakdown information, problem summary information, and specific business suggestions information.

[0069] This means that the structured output principle that defines the intelligent agent can be configured by injecting behavioral specification information of the structured output principle into the intelligent agent. For example, the injected information could be something like, "Your output must follow a fixed report structure, which typically includes: report title, summary of core conclusions, key data findings, dimensional breakdown, problem summary, and specific business suggestions."

[0070] 4) Configure the agent with first behavioral information that indicates the principle of objectivity and neutrality, so that the agent's analysis matches the first behavioral information;

[0071] The first line of information is used to indicate the principle of objectivity and neutrality, which can ensure that the agent maintains the objectivity and neutrality of the analytical perspective when performing analytical tasks.

[0072] This principle of objectivity and neutrality applies to intelligent agents. For example, in practical implementation, the corresponding configuration can be achieved by injecting the following information into the intelligent agent: "Maintain objectivity in the analysis perspective and avoid using emotional or subjective language, etc."

[0073] 5) Configure the agent with second behavioral information that indicates safety and compliance principles so that the agent's analysis results match the second behavioral information.

[0074] This principle of security and compliance for intelligent agents can instruct them not to output at least one type of information, including user privacy, unpublished data, and other sensitive information. For example, in practical implementation, the corresponding configuration can be achieved by injecting the following information into the intelligent agent: "Do not output any content involving user privacy, unpublished company financial data, or other sensitive information."

[0075] In this embodiment, when defining the behavior of the agent, the configured behavior specification information can be persistently stored in the agent configuration backend. When any application programming interface (API) calls the agent, the system automatically concatenates the "meta-hint" with the user's query question and then sends it to the final hint prefix of the large language model integrated within the agent, thereby constraining the model's behavior at the underlying level.

[0076] By adopting the above methods, it is possible to comprehensively, accurately, and reasonably regulate the behavior of intelligent agents.

[0077] Step S205: Perform contextual learning on the intelligent agent for business indicator information in the target domain, so that the business indicator interpretation module in the intelligent agent can be used to interpret business indicators.

[0078] For example, the business indicator information in this embodiment includes at least one of the following: the definition of the business indicator, its business significance, and the influencing factors during business indicator analysis. This enables the intelligent agent to fully and accurately understand the business indicators, and thus to more accurately analyze the anomalies of the business indicators.

[0079] Optionally, if there are a large number of business metrics in the target domain, including some general and core metrics, and the general business metrics can already be mastered by the intelligent agent in the general domain, then in this step, only the core business metrics of the target domain can be injected into the intelligent agent.

[0080] For example, in the e-commerce sector, core business metrics may include Daily Active Users in Transactions (DAU) and Gross Merchandise Volume (GMV).

[0081] Among them, the business metric information of transaction DAU can be clearly defined as the number of unique users who complete at least one transaction on a given day, explaining its business significance and analyzing its common influencing factors.

[0082] The payment GMV business metrics information can clearly define the total amount of all successfully paid orders on a given day as payment GMV, explaining its business significance. It can also be broken down into its driving formula: GMV = number of transacting users × average order value; further, it can be broken down into traffic × conversion rate × average order value, etc.

[0083] Step S206: Perform contextual learning on the agent regarding field concept information, so that the corresponding field concept system understanding module in the agent can understand at least one of the definition, scenario, and characteristics of the field concept;

[0084] For example, taking the e-commerce domain as the target domain, the field concept information in this embodiment can include the platform traffic types defined in the field concept system of the e-commerce domain. Typical scenarios include public domain, private domain, commercial domain, and others. Their characteristics include: public domain traffic is mainly affected by recommendation strategies, private domain traffic is mainly affected by user's active behavior, commercial domain traffic is mainly affected by the merchant's GMV, and other traffic is small and influenced by complex factors, so it can be disregarded in the analysis. The field concept information can include the above information, enabling the field concept system understanding module in the intelligent agent to learn this field concept information.

[0085] Step S207: Perform contextual learning on the agent regarding user type information, so that the corresponding user type recognition module in the agent can understand at least one of the definitions of different user types, the core behavioral characteristics of different types of users, and the key indicators of different types of users.

[0086] For example, taking the e-commerce sector as the target domain, the definition of user type information in this embodiment can be based on the user's transaction behavior on the platform. Core behavioral characteristics can include: new e-commerce customers are users who complete a transaction on the platform for the first time, and returning e-commerce customers are users who have not completed a transaction for the first time. Key indicators can include customer acquisition cost (CAC) and first-order conversion rate for new e-commerce customers, and repeat purchase rate, repeat purchase category, repeat purchase GMV, etc. for returning e-commerce customers. The first-order conversion rate for new customers can refer to the number of users who complete a transaction on the platform for the first time / the number of users who have never completed a transaction on the platform but visited the platform on the same day.

[0087] Step S208: Perform context learning on the agent regarding conversion funnel analysis information, so that the corresponding conversion funnel model analysis module in the agent can be used to understand at least one of the conversion funnel hierarchy and the meaning of conversion funnel analysis.

[0088] For example, taking the e-commerce sector as the target field, the conversion funnel analysis information in this embodiment may include: information on the conversion funnel levels such as exposure -> live stream viewing -> effective live stream viewing -> order placement -> payment, etc. The significance of conversion funnel analysis can be to explain what the conversion rate of each level represents. For example, exposure to live stream viewing reflects the attractiveness of the content, and order placement to payment reflects the payment experience and decision support. Funnel analysis can answer the question "What is the conversion bottleneck?"

[0089] In this embodiment, the above-mentioned Figure 1The step S101 of the illustrated embodiment, which involves contextual learning of conceptual information of the target domain for the intelligent agent, includes steps S206-208 of this embodiment. In actual application scenarios, other conceptual information of the target domain can also be injected into the intelligent agent according to actual needs. The principle is the same, and will not be described in detail here.

[0090] By adopting this approach, the intelligent agent can learn and master more extensive professional knowledge in the target domain, providing a solid foundation for subsequent anomaly analysis.

[0091] S209. Perform contextual learning on the anomaly analysis methodology for the agent, so that the agent's data analysis methodology module can be used to learn the anomaly analysis framework, understand the attribution logic, and combine modules that need to be referenced when combining anomaly analysis based on analysis tasks;

[0092] For example, in this embodiment, the anomaly analysis methodology is injected into the agent to enable the agent to learn the context of the anomaly analysis methodology. Specifically, the anomaly analysis framework in the anomaly analysis methodology can be injected into the agent, and classic anomaly analysis ideas, such as "first look at the whole, then break down the dimensions," can be instilled in the agent.

[0093] In this embodiment, the attribution logic may include common attribution methods such as last attribution and attribution principles such as finding the contributor with the greatest change.

[0094] Based on the modules needed for analyzing changes in the combination of analytical tasks, also known as injection and fusion techniques, this guides the agent to not infuse all knowledge modules at once, but to dynamically select and combine relevant knowledge modules according to the type of task the agent is about to execute. For example, when the task involves DAU analysis, the agent will automatically append the prompt text of "DAU part of the core business indicator interpretation module," "user lifecycle module," and "data analysis methodology module" to the user query and meta prompt. This "on-demand loading, context injection" approach ensures that the agent has sufficient domain knowledge depth in specific tasks while avoiding model performance degradation or attention distraction caused by excessively long prompts. The knowledge modules in the agent, including the aforementioned business indicator interpretation module, field concept system understanding module, user type identification module, and conversion funnel model analysis module, are editable assets that are regularly maintained and updated by dedicated personnel.

[0095] In this embodiment, step S201 defines the intelligent agent in the target domain, and steps S202-S204 determine the "behavioral norms" of the intelligent agent. Steps S205-S209 systematically instill the professional knowledge necessary for completing tasks in a specific domain into the intelligent agent, building a solid "business knowledge framework" for it. This is no longer a fragmented mention of knowledge points, but rather in-depth teaching in a structured manner through carefully designed "knowledge prompting modules".

[0096] For example, in the e-commerce field, during the construction and modularization of the knowledge system, the knowledge required for e-commerce analysis can be broken down into multiple interconnected modules, each corresponding to a highly structured prompt text block. These modules together constitute the agent's "long-term memory" or "knowledge base," which is used for anomaly analysis.

[0097] By adopting this approach, the agent can learn and master the methodology of anomaly analysis in the target domain, providing a reliable basis for subsequent anomaly analysis.

[0098] Specifically, in this embodiment, taking the sequential learning of business indicator information, conceptual information, and anomaly analysis methodology of the target domain for the intelligent agent as an example, the intelligent agent can learn the rich professional knowledge of the target domain in a hierarchical and comprehensive manner, thereby effectively improving the accuracy of the intelligent agent in business analysis.

[0099] S210. Context learning of the thought chain template for the agent to perform anomaly analysis on business metrics, so that the agent's metric analysis module can be used to perform anomaly analysis on business metrics based on the thought chain template.

[0100] In practice, a thought chain template for analyzing changes in business metrics can be injected into the agent, enabling the agent to learn the context of the thought chain template for analyzing changes in business metrics.

[0101] Optionally, the thought chain template in this embodiment includes a reasoning chain of multiple steps, each step including input, at least one processing logic, and output. The thought chain template in this embodiment can be pre-configured by professionals based on experience, possessing professionalism and accuracy.

[0102] After endowing the agent with "behavioral norms" and "business knowledge" in the aforementioned steps, the goal of embedding the thought chain template for anomaly analysis of business indicators into the agent is to teach it "how to work," that is, to transform complex analysis tasks into "programs" or "standard operating procedures" that can be automatically executed and stably reproduced within its own system. By designing a highly structured "thought chain template," the analytical reasoning process can be transformed from free play into a step-by-step standardized operation.

[0103] By learning the context of the thought chain template for analyzing changes in business metrics, the agent can learn and master the reasoning logic of multiple steps in the thought chain of change analysis, thereby improving the professionalism and accuracy of the constructed agent.

[0104] For example, in the e-commerce field, "Transaction DAU Anomaly Analysis" and "Payment GMV Anomaly Analysis" can be used as templates to break down each task into a fixed, multi-step reasoning chain. Each step has clear inputs, processing logic, and outputs.

[0105] The thought process template for embedding business metrics into the intelligent agent for anomaly analysis in this embodiment is itself a configurable and executable "program." The aforementioned process text serves as a highly detailed "prompt template," containing logical judgments and variable input. During interaction with the model, the system automatically acquires data and fills the results into the template, forming the final prompt for the model, guiding it to "think" and output step by step.

[0106] Optionally, in this embodiment, the intelligent agent constructed using this embodiment can also provide suggestions affecting business metrics when performing anomaly analysis. This allows for closed-loop business applications and continuous iteration based on feedback after the intelligent agent is built. For example, after deploying the completed intelligent agent to the production environment, it can be deeply integrated into the business workflow, and a feedback loop can be established to continuously evolve its capabilities, forming a growth flywheel of "analysis-decision-verification-optimization".

[0107] During use, business personnel can also wake up the intelligent agent at any time in the data platform or collaboration tools using natural language to perform specific analyses;

[0108] Specific verification can employ A / B testing. After the business side takes action based on the agent's suggestions, the effectiveness of the strategy is measured using a standard A / B testing framework. Based on the test results, the effectiveness of the agent's suggestions can be verified; furthermore, the agent can be optimized based on the verification results.

[0109] The system can also perform periodic retraining of the agent. For example, the system can set up scheduled tasks, such as weekly or bi-weekly, to automatically rerun the above-described construction in this embodiment using new data from the most recent period, so that the agent can learn the latest knowledge.

[0110] The intelligent agent constructed in this embodiment is no longer a static tool, but a "digital business partner" with solid basic skills, stable methodology, strong execution and continuous learning capabilities, which can truly play an irreplaceable role in core business scenarios.

[0111] The intelligent agent construction method in this embodiment, by adopting the above-mentioned multi-step hierarchical construction, enables the intelligent agent to learn rich professional knowledge of the target domain in a hierarchical and comprehensive manner, thereby improving the professionalism and intelligence of the intelligent agent and enabling the intelligent agent to learn more accurate thought chains. It also enables the intelligent agent to have stronger reasoning logic in business analysis, effectively improving the professionalism, intelligence and accuracy of intelligent agent construction.

[0112] The intelligent agent construction method in this embodiment can solidify the thought chain obtained from expert experience into the built-in process of the intelligent agent, enabling junior operators to quickly produce analysis reports with professional depth and uniform format, releasing the productivity of the data team, and significantly improving the business analysis efficiency and consistency of the intelligent agent.

[0113] The intelligent agent constructed using the intelligent agent construction method of this embodiment can provide stable, timely, and easy-to-understand anomaly analysis results, helping business teams to quickly grasp the core of the problem and reduce decision-making delays and subjective assumptions.

[0114] The intelligent agent constructed in this embodiment achieves standardization and intelligence of the analysis process at a lower cost, based on the capabilities of general large models, and has significant advantages in terms of flexibility, maintainability and efficiency improvement.

[0115] Figure 3 This is a schematic diagram according to the third embodiment of the present disclosure; this embodiment provides an agent construction apparatus 300, including:

[0116] The first learning module 301 is used to perform contextual learning on the intelligent agent for business indicator information, conceptual information of the target domain, and anomaly analysis methodology, respectively, so that the business indicator interpretation module in the intelligent agent is used to interpret business indicators, the concept network module is used to understand the conceptual information of the target domain, and the data analysis methodology module is used to learn the anomaly analysis method.

[0117] The second learning module 302 is used to perform context learning on the agent for a thought chain template for analyzing the anomalies of the business indicators, so that the agent's indicator analysis module can perform anomaly analysis on the business indicators based on the thought chain template.

[0118] The intelligent agent construction device 300 in this embodiment achieves the same implementation principle and technical effect as the above-mentioned related method embodiments by using the above-mentioned modules. For details, please refer to the description of the above-mentioned related method embodiments, which will not be repeated here.

[0119] Figure 4 This is a schematic diagram according to the fourth embodiment of the present disclosure; the intelligent agent construction apparatus 400 of this embodiment, in the above... Figure 1 Based on the technical solutions of the illustrated embodiments, the technical solutions of this disclosure will be described in further detail. For example... Figure 4 As shown, the intelligent agent construction apparatus 400 of this embodiment includes the above-described... Figure 3 The modules with the same name and function shown are: the first learning module 401 and the second learning module 302.

[0120] For example, in this embodiment, the business indicator information includes at least one of the following: the definition of the business indicator, its business significance, and the influencing factors during the analysis of the business indicator.

[0121] like Figure 4 As shown, the intelligent agent construction device 400 in this embodiment further includes:

[0122] The first configuration module 403 is used to configure capability and boundary declaration information for the intelligent agent.

[0123] Further optional, such as Figure 4 As shown, the intelligent agent construction device 400 in this embodiment further includes:

[0124] The second configuration module 404 is used to configure behavioral norm information for the intelligent agent.

[0125] Further optionally, in one embodiment of this disclosure, the second configuration module 404 is configured to perform at least one of the following steps:

[0126] Configure the agent with authentic behavioral norms so that the agent can perform analysis based on the provided data;

[0127] Configure the agent with traceability principle behavioral norms so that the agent has traceability in the data being analyzed;

[0128] Configure the agent with structured output principle behavior specification information so that the agent can output the structure of the analysis report. The structure of the analysis report includes at least one of the following: report title information, core conclusion summary information, key data discovery information, dimensional decomposition information, problem summary information, and specific business suggestion information.

[0129] Configure the agent with first behavioral information indicating an objective and neutral principle, so that the agent's analysis matches the first behavioral information; and

[0130] Configure the agent with second behavioral information that indicates safety and compliance principles, so that the agent's analysis results match the second behavioral information.

[0131] Further optionally, in one embodiment of this disclosure, the intelligent agent construction apparatus 400 of this embodiment further includes:

[0132] The third learning module 405 is used for:

[0133] The defined agent undergoes context learning with respect to a preset question library, which includes multiple query examples from the target domain.

[0134] Further optionally, in one embodiment of this disclosure, the first learning module 401 may include:

[0135] The business indicator learning unit 4011 is used to perform contextual learning of business indicator information for the target domain on the intelligent agent, so that the business indicator interpretation module in the intelligent agent can interpret the business indicators.

[0136] The concept information learning unit 4012 is used to perform contextual learning of concept information in the target domain on the agent, so that the concept network module in the agent can understand the concept information in the target domain.

[0137] The methodology learning unit 4013 is used to perform contextual learning of the anomaly analysis methodology for the target domain on the agent, so that the data analysis methodology module in the agent can be used to learn the anomaly analysis method.

[0138] Further, optionally, in one embodiment of this disclosure, the concept information learning unit 4012 is configured to perform at least one of the following steps:

[0139] The agent is subjected to contextual learning of field concept information so that the corresponding field concept system understanding module in the agent can understand at least one of the definition, scenario and characteristics of the field concept;

[0140] The agent undergoes contextual learning regarding user type information, enabling the corresponding user type recognition module within the agent to understand at least one of the following: the definition of different user types, the core behavioral characteristics of different user types, and the key indicators of different user types; and

[0141] The agent is subjected to contextual learning regarding conversion funnel analysis information, so that the corresponding conversion funnel model analysis module in the agent can understand at least one of the conversion funnel hierarchy and the meaning of conversion funnel analysis.

[0142] Further, optionally, in one embodiment of this disclosure, the methodology learning unit 4013 is used for:

[0143] The agent is subjected to contextual learning regarding the anomaly analysis methodology, so that the agent's data analysis methodology module can be used to learn the anomaly analysis framework, understand the attribution logic, and combine modules that need to be referenced when performing anomaly analysis based on the analysis task.

[0144] Further optionally, in one embodiment of this disclosure, the thought chain template includes a reasoning chain of multiple steps, each step including an input, at least one processing logic, and an output.

[0145] Further optionally, in one embodiment of this disclosure, when the target field is the e-commerce field, the core business indicators include at least one of daily active users and total transaction volume.

[0146] The intelligent agent construction device 400 in this embodiment achieves the same implementation principle and technical effect as the above-mentioned related method embodiments by using the above-mentioned modules. For details, please refer to the description of the above-mentioned related method embodiments, which will not be repeated here.

[0147] The acquisition, storage, and application of any type of information, such as user personal information, involved in the technical solutions disclosed herein comply with relevant laws and regulations and do not violate public order and good morals.

[0148] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0149] Figure 5 A schematic block diagram of an example electronic device 500 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0150] like Figure 5As shown, device 500 includes a computing unit 501, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 502 or a computer program loaded from storage unit 508 into random access memory (RAM) 503. RAM 503 may also store various programs and data required for the operation of device 500. The computing unit 501, ROM 502, and RAM 503 are interconnected via bus 504. Input / output (I / O) interface 505 is also connected to bus 504.

[0151] Multiple components in device 500 are connected to I / O interface 505, including: input unit 506, such as keyboard, mouse, etc.; output unit 507, such as various types of monitors, speakers, etc.; storage unit 508, such as disk, optical disk, etc.; and communication unit 509, such as network card, modem, wireless transceiver, etc. Communication unit 509 allows device 500 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0152] The computing unit 501 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 501 performs the various methods and processes described above, such as the methods of this disclosure. For example, in some embodiments, the methods of this disclosure may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 508. In some embodiments, part or all of the computer program may be loaded and / or installed on device 500 via ROM 502 and / or communication unit 509. When the computer program is loaded into RAM 503 and executed by the computing unit 501, one or more steps of the methods of this disclosure described above may be performed. Alternatively, in other embodiments, the computing unit 501 may be configured to perform the methods of this disclosure by any other suitable means (e.g., by means of firmware).

[0153] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0154] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0155] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0156] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0157] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0158] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0159] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0160] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for constructing an intelligent agent, comprising: The agent is subjected to contextual learning of business indicator information, conceptual information of the target domain, and anomaly analysis methodology in the target domain, so that the business indicator interpretation module in the agent is used to interpret business indicators, the concept network module is used to understand the conceptual information of the target domain, and the data analysis methodology module is used to learn the anomaly analysis method. The agent learns the context of a thought chain template for analyzing the anomalies of the business metrics, so that the agent's metric analysis module can perform anomaly analysis on the business metrics based on the thought chain template.

2. The method according to claim 1, wherein, The business indicator information includes at least one of the following: the definition of the business indicator, its business significance, and the influencing factors during the analysis of the business indicator.

3. The method according to claim 1, wherein, Before performing context learning on the intelligent agent regarding business metrics information of the target domain, conceptual information of the target domain, and anomaly analysis methodology, the method further includes: Configure the capabilities and boundary declaration information for the intelligent agent.

4. The method according to claim 1, wherein, Before performing context learning on the intelligent agent regarding business metrics information of the target domain, conceptual information of the target domain, and anomaly analysis methodology, the method further includes: Configure behavioral specification information for the intelligent agent.

5. The method according to claim 4, wherein, Configure behavioral specification information for the intelligent agent, including at least one of the following: Configure the agent with authentic behavioral norms so that the agent can perform analysis based on the provided data; Configure the agent with traceability principle behavioral norms so that the agent has traceability in the data being analyzed; Configure the agent with structured output principle behavior specification information so that the agent can output the structure of the analysis report. The structure of the analysis report includes at least one of the following: report title information, core conclusion summary information, key data discovery information, dimensional decomposition information, problem summary information, and specific business suggestion information. Configure the agent with first behavioral information indicating the principle of objectivity and neutrality, so that the agent's analysis matches the first behavioral information; as well as Configure the agent with second behavioral information that indicates safety and compliance principles, so that the agent's analysis results match the second behavioral information.

6. The method according to claim 1, wherein, Before performing context learning on the intelligent agent regarding business metrics information of the target domain, conceptual information of the target domain, and anomaly analysis methodology, the method further includes: The defined agent undergoes context learning with respect to a preset question library, which includes multiple query examples from the target domain.

7. The method according to claim 1, wherein, Contextual learning of conceptual information about the target domain is performed on the agent so that the conceptual network module in the agent can understand the conceptual information of the target domain, including at least one of the following: The agent is subjected to contextual learning of field concept information so that the corresponding field concept system understanding module in the agent can understand at least one of the definition, scenario and characteristics of the field concept; The agent is subjected to contextual learning about user type information so that the corresponding user type recognition module in the agent can understand at least one of the definitions of different user types, the core behavioral characteristics of different types of users, and the key indicators of different types of users. as well as The agent is subjected to contextual learning regarding conversion funnel analysis information, so that the corresponding conversion funnel model analysis module in the agent can understand at least one of the conversion funnel hierarchy and the meaning of conversion funnel analysis.

8. The method according to claim 7, wherein, Contextual learning of the anomaly analysis methodology is performed on the agent to enable the data analysis methodology module within the agent to learn the anomaly analysis method, including: The agent is subjected to contextual learning regarding the anomaly analysis methodology, so that the agent's data analysis methodology module can be used to learn the anomaly analysis framework, understand the attribution logic, and combine modules that need to be referenced when performing anomaly analysis based on the analysis task.

9. The method according to claim 1, wherein, The thought chain template includes a reasoning chain of multiple steps, each step including an input, at least one processing logic, and an output.

10. The method according to claim 1, wherein, When the target field is e-commerce, the business metrics include at least one of daily active users and total transaction volume.

11. An apparatus for constructing an intelligent agent, comprising: The first learning module is used to perform contextual learning on the agent for business indicator information, conceptual information of the target domain, and anomaly analysis methodology in the target domain, so that the business indicator interpretation module in the agent can be used to interpret business indicators, the concept network module can be used to understand the conceptual information of the target domain, and the data analysis methodology module can be used to learn the anomaly analysis method. The second learning module is used to learn the context of the thought chain template for the intelligent agent regarding the anomaly analysis of the business indicators, so that the indicator analysis module of the intelligent agent can perform anomaly analysis of the business indicators based on the thought chain template.

12. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1-10.

13. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-10.

14. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method according to any one of claims 1-10.

Images