Project planning-based product recommendation method and device, equipment, medium and product

Abstract

The invention discloses a product recommendation method and device based on project planning, equipment, a medium and a product, and is applied to the technical field of financial information, and the method comprises the steps: monitoring whether user behavior data of a user to which a project belongs meets a preset push condition or not; if yes, performing multiple rounds of dialogues with the user to which the project belongs based on the dialogue model, and performing project planning demand intention identification; under the condition that the project planning demand intention is identified, generating a questionnaire guide dialogue based on a dialogue model, so that a user to which the project belongs fills a project questionnaire based on the questionnaire guide dialogue; the project questionnaire is analyzed based on the project planning model, and a project planning strategy corresponding to the project questionnaire is generated; calculating the feature similarity between the project planning strategy and the candidate recommended products based on the product recommendation model, and screening out a target product based on the feature similarity; and based on the dialogue model, performing a dialogue of recommending the target product with the user to which the project belongs. The user experience and the product recommendation accuracy can be improved.

Description

Technical Field

[0001] This application belongs to the field of financial information technology, and in particular relates to a product recommendation method, apparatus, equipment, medium and product based on project planning. Background Technology

[0002] In fields requiring refined and personalized allocation of project resources, and the recommendation of corresponding products or services, user demand for project planning services is increasing daily. Project planning services help users plan project resources in advance to ensure a stable supply of resources after project execution, thus fulfilling users' expectations for successful project completion.

[0003] Currently, when recommending products, project planners first need to communicate with users multiple times to collect user and project information. Next, based on personal experience or internal review rules, project planners analyze the collected information, assessing project priority, risk tolerance, and resource gaps. Based on this, they manually select and combine options from the product catalog to create one or more draft project resource allocation plans. Then, these drafts are repeatedly confirmed and revised with users via email or meetings before the final project resource allocation plan is approved. Finally, resource allocation or product ordering is manually executed in the backend management system. This entire process is lengthy, inefficient, and results in a poor user experience. Furthermore, the project resource allocation plans manually selected and combined by project planners may not necessarily be the best match for the user's own resource situation and personalized project resource allocation needs.

[0004] In summary, the current approach to recommending products with personalized resource configurations suffers from poor user experience and low recommendation accuracy. Summary of the Invention

[0005] This application provides a product recommendation method, apparatus, device, medium, and product based on project planning, which can improve user experience and product recommendation accuracy.

[0006] In a first aspect, embodiments of this application provide a product recommendation method based on project planning, the method comprising: Monitor whether the user behavior data of users belonging to the project meets the preset push conditions; Under the condition of meeting the preset push conditions, the system conducts multiple rounds of dialogue with the user to which the project belongs based on the dialogue model, and identifies the intention of the project planning requirement based on the multiple rounds of dialogue. Once the intent of the project planning requirements is identified, a questionnaire is generated based on the dialogue model to guide the dialogue, so that the users of the project can fill in the project survey questionnaire based on the questionnaire-guided dialogue. The project survey questionnaire includes at least project requirement information and user information of the users of the project. The project survey questionnaire is analyzed based on the project planning model to generate a corresponding project planning strategy. The project planning strategy includes at least: overall project goal, phased project goals, project resource storage method, product recommendation requirements, and project disaster recovery strategy. The similarity between the project planning strategy and the candidate recommended products is calculated based on the product recommendation model, and the target product is selected based on the similarity of the features. Based on the dialogue model, a dialogue is conducted with the user of the project to recommend the target product.

[0007] Secondly, embodiments of this application provide a product recommendation device based on project planning, the device comprising: The monitoring module is used to monitor whether the user behavior data of users belonging to the project meets the preset push conditions; The identification module is used to conduct multiple rounds of dialogue with the user to which the project belongs based on a dialogue model when the preset push conditions are met, and to identify the project planning needs intent based on the multiple rounds of dialogue. The first generation module is used to generate a questionnaire-guided dialogue based on the dialogue model when the intention of project planning needs is identified, so that the users of the project can fill in the project survey questionnaire based on the questionnaire-guided dialogue. The project survey questionnaire includes at least project requirement information and user information of the users of the project. The second generation module is used to analyze the project questionnaire based on the project planning model and generate a project planning strategy corresponding to the project questionnaire. The project planning strategy includes at least: the overall project goal, the phased project goals, the project resource storage method, the product recommendation requirements, and the project disaster recovery strategy. The calculation module is used to calculate the feature similarity between the project planning strategy and the candidate recommended products based on the product recommendation model, and to filter out the target products based on the feature similarity. The recommendation module is used to conduct a dialogue with the user of the project to recommend the target product based on the dialogue model.

[0008] Thirdly, embodiments of this application provide an electronic device, the device comprising: A processor and a memory storing computer program instructions; a product recommendation method based on project planning that implements any of the above when the processor executes the computer program instructions.

[0009] Fourthly, embodiments of this application provide a computer storage medium storing computer program instructions, which, when executed by a processor, implement the product recommendation method based on project planning described above.

[0010] Fifthly, embodiments of this application provide a computer program product in which instructions, when executed by a processor of an electronic device, enable the electronic device to perform any of the above-mentioned product recommendation methods based on project planning.

[0011] The product recommendation method, apparatus, device, medium, and product based on project planning in this application embodiment monitor whether the user behavior data of the user to which the project belongs meets preset push conditions. If the preset push conditions are met, a multi-turn dialogue is conducted with the user to which the project belongs based on a dialogue model. This proactively initiates a service process when potential user demand signals are identified. By identifying the user's project planning demand intent based on the multi-turn dialogue, the user's project planning demand intent can be identified efficiently and accurately. Furthermore, upon identifying the project planning demand intent, a questionnaire-guided dialogue is generated based on the dialogue model, allowing the user to fill in a project survey questionnaire. This guides the user to accurately describe project demand information and user information, reducing user operation difficulty and improving user experience. By analyzing the project survey questionnaire based on the project planning model, a project planning strategy corresponding to the project survey questionnaire is generated. The feature similarity between the project planning strategy and candidate recommended products is calculated based on a product recommendation model. Target products are selected based on the feature similarity, improving the accuracy of product recommendations. Finally, by conducting a dialogue with the user to which the project belongs based on the dialogue model to recommend the target product, the user experience is improved. In summary, this application can improve user experience and product recommendation accuracy. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a network architecture diagram of an application scenario of the product recommendation method based on project planning provided in the embodiments of this application; Figure 2 This is a flowchart illustrating a product recommendation method based on project planning provided in one embodiment of this application. Figure 3 This is a schematic diagram of the structure of a product recommendation device based on project planning provided in another embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in another embodiment of this application. Detailed Implementation

[0014] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0015] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0016] It should be noted that the acquisition, storage, use, and processing of data in this application embodiment all comply with the relevant provisions of national laws and regulations.

[0017] It should be noted that in the embodiments of this application, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, it does not mean that the applicant has used or necessarily used the solution.

[0018] The background technology involved in this application is described below.

[0019] In fields requiring refined and personalized configuration of project resources, and the recommendation of corresponding products or services, such as IT solution design, enterprise service customization, complex product configuration, and retirement planning product configuration, the demand for project planning services is increasing daily due to the significant differences in existing resources, risk appetite, and project goals among different users. Project planning services can help users find products or services that match their existing resources, risk appetite, and project goals, enabling them to plan project resources rationally in advance and ensure a stable supply of resources after project execution.

[0020] Currently, before recommending products, there is usually a standard operating procedure (SOP). First, project planners communicate with users multiple times to collect information such as the user's existing resources, risk tolerance, and project goals. This information is then analyzed to assess the project's priority, risk tolerance, and resource gaps. Based on this, one or more draft project resource allocation plans are manually selected and combined from the product catalog. These drafts are then repeatedly confirmed and revised with the user via email or meetings before the final project resource allocation plan is approved. Finally, resource allocation or product ordering is manually executed in the backend management system. This entire process is lengthy, inefficient, and results in a poor user experience. Furthermore, the project resource allocation plans manually selected and combined by project planners may not necessarily be the best match for the user's own resources and personalized project resource allocation needs.

[0021] In summary, the current approach to recommending products with personalized resource configurations suffers from poor user experience and low recommendation accuracy.

[0022] To address the problems of existing technologies, embodiments of the present invention provide a product recommendation method, apparatus, device, medium, and product based on project planning. The product recommendation method based on project planning provided by embodiments of the present invention is described below.

[0023] Figure 1 This is a network architecture diagram of an application scenario of the product recommendation method based on project planning provided in this application embodiment. For example... Figure 1 As shown, in one application scenario of the product recommendation method based on project planning provided in this application embodiment, there are server 11 and client 12, and server 11 and client 12 are connected in communication.

[0024] Client 12 can collect user behavior data of users belonging to the project, such as by recording user operation events through code embedded in a webpage or application. Client 12 sends the collected user behavior data to server 11 in real time or in batches.

[0025] Server 11 can determine whether the user behavior data of the user to which the project belongs meets the preset push conditions.

[0026] Under preset push conditions, server 11 can send a dialogue command to client 12, causing client 12 to display a dialogue entry on its interface or directly launch a dialogue interface. The user inputs text or voice through client 12; if it is voice, client 12 can first convert it to text. Client 12 then sends the user's input text to server 11.

[0027] Server 11 can invoke the dialogue model to process the current user input combined with the historical dialogue context, generating natural language response text. Server 11 sends the generated response text back to client 12, which displays it to the user, thus completing one round of dialogue. This process is repeated cyclically, forming multi-round dialogues.

[0028] During multi-turn dialogues, server 11 can invoke the intent recognition model to analyze the dialogues and identify the intent behind project planning requirements. The intent recognition model can be part of the dialogue model or an independent model.

[0029] Once the intent of the project planning requirements is identified, the server 11 can call the dialogue model to generate a questionnaire-guided dialogue and send the questionnaire-guided dialogue to the client 12.

[0030] Client 12 presents a questionnaire guidance dialog on the interface. Based on this dialog, users of the project can input or confirm information through client 12 to complete the project survey questionnaire. The content of the project survey document includes at least project requirement information and user information of the users of the project. After the user completes the questionnaire or triggers the submission command, client 12 sends the structured project survey questionnaire data to server 11.

[0031] After receiving the project survey data, server 11 can call the project planning model to analyze the data. Specifically, the project planning model can comprehensively process and calculate the various information in the questionnaire to generate corresponding project planning strategies. The project planning strategy can be a structured data object, and its content can include at least: the overall project goal, the phased project goals, the project resource storage method, product recommendation requirements, and the project disaster recovery strategy.

[0032] Server 11 can invoke the product recommendation model to calculate the feature similarity between the project planning strategy and the candidate recommended products. Specifically, the product recommendation model can encode the structured project planning strategy into a feature vector, and also encode the attribute information of the candidate recommended products into feature vectors of the same dimension. Then, the product recommendation model can calculate the similarity, such as cosine similarity, between the feature vector of the project planning strategy and the feature vectors of each candidate recommended product, and sort and filter them according to the similarity scores from high to low to select the target products.

[0033] After selecting the target product, server 11 can again invoke the dialogue model to generate a recommendation script for the target product and send it to client 12. The dialogue model can combine key elements in the project planning strategy with the characteristics of the target product to generate a natural language explanation, describing the reasons for the recommendation and the matching points. The key elements in the project planning strategy can include phased goals, disaster recovery strategies, etc.

[0034] Client 12 can display recommended phrases for the target product in the dialogue interface, thus completing one round of recommendation dialogue. Afterwards, the user of the project can continue to inquire about the target product through client 12, such as price inquiries, feature details, alternative solutions, etc. Client 12 can send the user's new questions to server 11, which can use a dialogue model combined with the recommendation context to generate a response, which is then displayed through client 12, realizing multi-round interactions around product recommendations.

[0035] Following the recommended dialogue, server 11 can continue to monitor real-time events related to the project, such as resource market price fluctuations and the release of new policy constraints. Server 11 can invoke the event distribution engine to evaluate and classify real-time events related to the project. For events that reach a specific threshold, server 11 can trigger a strategy recalculation process, which involves invoking the project planning model and product recommendation model. Based on the new event information and analysis, it generates an updated project planning strategy and target product list, and initiates a new round of notifications or recommended dialogues through the dialogue model, thereby achieving dynamic iterative optimization of the solution and recommendations.

[0036] In another application scenario of the product recommendation method based on project planning provided in this application embodiment, an electronic device is included. The electronic device may deploy models such as dialogue models, project planning models, and product recommendation models, or their application programming interfaces (APIs), so that the electronic device can call these models. The electronic device can directly interact with the user through a user interface provided by its locally running application or browser.

[0037] Figure 2 This is a flowchart illustrating a product recommendation method based on project planning provided in one embodiment of this application. See also... Figure 2 As shown, the product recommendation method based on project planning provided in this application embodiment can be executed by an electronic device, and the method can include steps 101 to 106.

[0038] Step 101: Monitor whether the user behavior data of the users belonging to the project meets the preset push conditions.

[0039] The user to which the project belongs refers to the user associated with the preset project, such as the person in charge of the preset project, the operator, etc.

[0040] User behavior data refers to data collected through technical means that reflects users' digital activity patterns. User behavior data is typically structured or semi-structured. For example, it may include: the frequency and time periods of a user's login to a business system, the duration of time a user spends on specific functional pages, the number of times a user clicks on certain types of news articles, and the number of times a user has historically inquired about or searched for certain keywords. Specific functional pages may include product introduction pages, solution case study pages, etc.

[0041] Preset push conditions refer to rules or thresholds that are pre-defined to judge user behavior data. Preset push conditions are used to filter out users who are more likely to need project planning, avoiding the waste of resources and user disturbance caused by indiscriminate pushes.

[0042] In some embodiments, electronic devices can collect user behavior data, such as login, browsing, clicking, and searching events, through code embedding on web pages or application pages. The electronic device can use a set of rule-based filters to determine whether the user behavior data meets preset push conditions. Alternatively, the electronic device can employ a trained lightweight classification model to analyze the user behavior data, predict the probability that the user needs the project planning service, and determine that the user behavior data meets the preset push conditions when the predicted probability of needing the project planning service is greater than a preset probability.

[0043] For example, when an electronic device detects that a user frequently visits the project planning case library page and the status of the associated project is "not started", the electronic device can determine that the user behavior data matches a preset high potential demand rule, thereby confirming that the preset push conditions are met.

[0044] Step 102: Under the condition of meeting the preset push conditions, conduct multi-round dialogue with the user to which the project belongs based on the dialogue model, and identify the project planning needs intent based on the multi-round dialogue.

[0045] A dialogue model is a computational model capable of understanding and generating sequences of natural language, typically trained on a deep learning architecture. Multi-turn dialogue refers to multiple consecutive, context-dependent question-and-answer interactions.

[0046] When user behavior data meets preset push conditions, the electronic device can call the dialogue model interface to start a dialogue session on the user interface. The text or speech-to-text content entered by the user, along with the dialogue history, is then input into the dialogue model.

[0047] In some embodiments, the dialogue model can semantically encode and understand the current input through its internal mechanisms, and generate coherent and guided response text in combination with the context, thereby enabling multi-turn interactions.

[0048] At the same time, electronic devices can use natural language processing technology to analyze the user's dialogue content, classify it into a predefined set of intent categories, and identify the intent of project planning requirements.

[0049] For example, an electronic device can proactively initiate a dialogue within the user interface, such as: "Hello, I noticed you've recently been looking into retirement planning product configuration management. Would you like me to provide you with a personalized retirement planning product configuration suggestion?" The user might reply: "What is a personalized retirement planning product configuration?" The electronic device can explain what a personalized retirement planning product configuration is and answer the user's question during the dialogue. If the user replies with responses such as "I need a personalized retirement planning product configuration suggestion" or "Please provide me with a personalized retirement planning product configuration suggestion," then the electronic device, by analyzing the dialogue content, can successfully identify the user's clear intention to request a retirement planning product configuration.

[0050] Step 103: After identifying the project planning needs and intentions, generate a questionnaire based on the dialogue model to guide the dialogue, so that the users of the project can fill in the project survey questionnaire based on the questionnaire-guided dialogue. The project survey questionnaire includes at least project requirement information and user information of the users of the project.

[0051] Guided dialogue is a communication process led by a dialogue model, using natural language interaction to gradually guide users to provide the information required for a project survey questionnaire. Compared to directly presenting a static questionnaire form, guided dialogue provides a better user experience.

[0052] In some embodiments, electronic devices can collect the information required for a project survey questionnaire through a questionnaire-guided dialogue, and populate the corresponding fields of the project survey questionnaire with the discrete information provided by the user through the questionnaire-guided dialogue.

[0053] Electronic devices can break down the information required for a project survey questionnaire into multiple fields to be filled. Through interactive methods such as asking questions, confirming and clarifying, and dynamically adjusting subsequent questions based on the information obtained, all fields can be filled in step by step. In this way, rigid form filling can be transformed into a flexible interview-style dialogue.

[0054] For example, the dialogue model asks: "To develop a plan, we first need to understand the expected timeframe of the project. How many months do you plan to take to complete the main objectives?" The user answers: "About six to eight months." Based on this answer and several fields to be filled, the dialogue model determines whether the next step is to inquire about the budget range or team size, and generates the corresponding next question.

[0055] Step 104: Analyze the project survey questionnaire based on the project planning model and generate the project planning strategy corresponding to the project survey questionnaire. The project planning strategy shall include at least: the overall project goal, the phased project goals, the project resource storage method, the product recommendation requirements, and the project disaster recovery strategy.

[0056] A project planning model is an algorithmic model capable of in-depth analysis and comprehensive decision-making on project survey questionnaires, outputting project planning strategies. Project planning strategies should at least include elements such as overall objectives, phased objectives, resource storage methods, product recommendation requirements, and disaster recovery strategies.

[0057] In some embodiments, the project planning model may employ classification algorithms to determine suitable resource storage and disaster recovery paradigms, and may use regression or sequence prediction models to quantify the overall goal and break down phase goals. The project planning model can generate an internally consistent, machine-readable project planning strategy.

[0058] Step 105: Calculate the feature similarity between the project planning strategy and the candidate recommended products based on the product recommendation model, and select the target product based on the feature similarity.

[0059] The product recommendation model is used to calculate the feature similarity between project planning strategies and candidate recommended products, and to select target products based on feature similarity.

[0060] In some embodiments, feature similarity can be a quantified matching score. Specifically, the product recommendation model can employ a representation learning model, such as a dual-tower neural network architecture, to calculate feature similarity. In the dual-tower neural network architecture, one tower encodes the structured project planning strategy into a high-dimensional vector, i.e., the strategy feature vector; the other tower encodes the attributes of candidate products into product feature vectors of the same dimension. Subsequently, the product recommendation model calculates the cosine similarity or inner product between these two vectors as the quantified feature similarity. Then, the product recommendation model can traverse or retrieve the candidate product pool, calculate the feature similarity between each candidate recommended product in the pool and the project planning strategy, and sort the candidate recommended products in the pool according to the feature similarity from largest to smallest. The candidate recommended product ranked first can be selected as the most matching target product.

[0061] Step 106: Based on the dialogue model, conduct a dialogue with the users of the project to recommend target products.

[0062] In some embodiments, after identifying the target product, the electronic device can invoke the dialogue model again to recommend the target product to the user through dialogue. Specifically, the electronic device can inform the user of the target product's information, key attributes, and its relevance to the project planning strategy through dialogue, explaining why the target product is recommended and how it meets the specific requirements of the planning strategy. For example, the model might say: "Based on your requirements for high availability and phased deployment, we recommend using a certain database service because it provides cross-availability zone deployment capabilities and supports elastic scaling, which perfectly matches the goals of your first-phase pilot." The model can then continue to answer the user's questions about the product through dialogue.

[0063] The product recommendation method based on project planning provided in this application monitors whether the user behavior data of the user belonging to the project meets preset push conditions. If the preset push conditions are met, a multi-turn dialogue is conducted with the user belonging to the project based on a dialogue model. This proactively initiates a service process when potential user demand signals are detected. By identifying the user's project planning demand intent through multi-turn dialogue, the method can efficiently and accurately identify the user's project planning demand intent. Furthermore, upon identifying the project planning demand intent, a questionnaire is generated based on the dialogue model to guide the dialogue. The user belonging to the project fills out a project survey questionnaire based on this guided dialogue, guiding the user to accurately describe project demand information and user information, reducing user operation difficulty and improving user experience. By analyzing the project survey questionnaire based on the project planning model, a project planning strategy corresponding to the survey questionnaire is generated. The feature similarity between the project planning strategy and candidate recommended products is calculated based on the product recommendation model. Target products are then selected based on feature similarity, improving the accuracy of product recommendations. Finally, by conducting dialogue with the user belonging to the project based on the dialogue model to recommend target products, the user experience is further improved. In summary, this application can improve both user experience and product recommendation accuracy.

[0064] In some embodiments, in order to accurately monitor whether the user behavior data of the users to which the project belongs meets the preset push conditions, step 101 is further refined to include: Step 201: Collect user behavior data of the users belonging to the project; Step 202: Extract user behavior features and basic project attribute features from user behavior data based on the user behavior detection model; Step 203: Determine whether the user behavior data meets the preset push conditions based on user behavior characteristics and project basic attribute characteristics. The user behavior characteristics include at least one of user consultation behavior characteristics, user browsing behavior characteristics, and user login behavior characteristics. The project basic attribute characteristics include at least one of project type and project lifecycle characteristics.

[0065] User behavior data can include user inquiry behavior, user browsing behavior, and user login behavior. User inquiry behavior refers to interactive events where a user actively initiates two-way communication to obtain information or assistance. User browsing behavior refers to events where a user passively receives or unilaterally views information content. User login behavior refers to events where a user completes identity authentication and establishes a valid system session.

[0066] Project fundamental attributes refer to descriptive characteristics that are inherent to the project itself and typically do not change in real time with a single interaction. Project fundamental attributes include at least one of the following: project type and project lifecycle characteristics. The project type can be the industry classification to which the project belongs, and the project lifecycle characteristic can be the stage of the project in a standard lifecycle model.

[0067] In some embodiments, user behavior data may include multiple user behavior logs. Electronic devices can obtain user behavior logs by embedding specific data collection code, i.e., event tracking code, into the front-end interactive interface to listen for and record user interaction events in real time.

[0068] User interaction events include, but are not limited to, user login events, page browsing events, button click events, form field focus / focus events, page scrolling events, and session timeout events. Each time an event is triggered, the event tracking code generates a structured log record. This log record typically includes the user identifier, the timestamp of the event, the event type, the identifier of the object the event affected, and related context parameters.

[0069] For example, when user U1001 visits a page titled "Introduction to Pension Planning Tools" at 10:05 AM on a weekday, the system will generate a log record containing the user identifier U1001, the timestamp of the corresponding date and time, the event type of "page view," and the event object of the page number of the "Introduction to Pension Planning Tools" page.

[0070] User behavior detection models can be obtained by training a pre-defined user behavior detection model using training data. The pre-defined user behavior detection model can be a gradient boosting decision tree model, a recurrent neural network (RNN), a long short-term memory network (LSTM), or a convolutional neural network (CNN). Training data can include training samples and their corresponding labels. Training samples can be a collection of all user behavior logs generated by a user within an observation window. The labels corresponding to the training samples can be manually labeled indicating whether or not the planning service should be initiated for that training sample.

[0071] User behavior characteristics can be statistical quantities or time-series patterns derived from users' historical interaction records. For example, user consultation behavior characteristics can be specified as the number of times a user initiated consultation dialogues through the online customer service portal in the past seven days, and the frequency of keywords related to preset projects in the consultation text. User browsing behavior characteristics can be specified as the average time a user spends on specific types of information articles, and the frequency of repeated visits to product detail pages. User login behavior characteristics can be specified as the distribution of users' typical active time periods, and the average duration of login sessions.

[0072] Basic project attributes can be user-related project metadata. Project type attributes can be coded according to the business nature of the project, such as personal pension savings projects or family asset allocation projects. Project lifecycle attributes are marked according to the stage of the project, such as the project initialization stage, the plan development stage, and the project execution monitoring stage.

[0073] In some embodiments, the user behavior detection model first aggregates and divides the input user behavior logs by user identifier and time window. Then, for each user's logs within each time window, it calculates dozens or even hundreds of statistical features, such as the total number of page view events in the past three days, the average number of logins per day in the past week, and the number of days since the last inquiry. Simultaneously, it retrieves the corresponding project type code and project lifecycle stage label from the user profile or project database. Finally, the user behavior detection model outputs a feature vector that integrates dynamic user behavior features and basic project attribute features.

[0074] Preset push conditions can be a set of decision rules or thresholds defined by business logic.

[0075] In some embodiments, the process of determining whether user behavior data meets preset push conditions may involve determining whether user behavior characteristics and project basic attribute characteristics meet decision rules. For example, one rule may stipulate that if the project lifecycle characteristic associated with the user shows that the project has ended, then regardless of the user behavior characteristics, the push conditions are not met. Another rule may stipulate that if the user has triggered a push within the past 24 hours and has not generated further interaction, then the conditions for pushing again are temporarily not met.

[0076] The product recommendation method based on project planning provided in this application collects user behavior data of users belonging to a project, and extracts user behavior features and basic project attribute features from the user behavior data based on a user behavior detection model. It can extract truly predictive user behavior features and basic project attribute features from massive and messy raw logs. By judging whether user behavior data meets preset push conditions based on user behavior features and basic project attribute features, it can accurately identify users with project planning needs, so as to initiate services for users who need project planning needs, while reducing the interference of invalid pushes to users.

[0077] In some embodiments, to further improve the user experience, the method further includes, under the condition that preset push conditions are met: Step 301: Push the project planning visualization component to the users to which the project belongs. The project planning visualization component contains an entry point to trigger multi-turn dialogues. Step 302: In response to the multi-turn dialogue entry being triggered, the dialogue model is invoked to enter a multi-turn dialogue with the user to whom the project belongs.

[0078] A project planning visualization component can be a pre-defined graphical user interface element. For example, it could be an interactive card or overlay module that integrates a brief service description, value proposition, and an entry point to trigger multi-turn dialogues. The entry point to trigger multi-turn dialogues could be a visually recognizable button or clickable area, labeled with text such as "Consult Now," "Start Planning," or something similar.

[0079] In some embodiments, electronic devices can implement push notifications through the component rendering capabilities of a front-end application framework. When the back-end service determines that the preset push conditions are met, it can send a command data packet to the front-end. The command data packet can contain the component's layout information, display content, and behavioral logic. After receiving the command, the front-end application can dynamically render the component in a non-intrusive but sufficiently prominent manner according to the context of the current user interface. For example, when a user browses a personal asset overview page, the component may appear as a sidebar card; after a user completes a transaction, the component may be briefly displayed as a modal overlay.

[0080] When a user interacts with the entry point that triggers a multi-turn dialogue through clicking, touching, or other means, the front-end application can capture this interaction event and perform the following operations through the event handling function: record the triggering behavior for subsequent analysis; hide or remove the project planning visualization component to clean up the interface; and send a request to the back-end service to start a new dialogue session.

[0081] Upon receiving a startup request, the backend service of the electronic device can initialize or connect to a dialogue model dedicated to natural language conversation to enter a multi-turn dialogue with the user to whom the project belongs. The dialogue model can be a pre-trained artificial intelligence model capable of handling coherent multi-turn exchanges. The invocation process may include establishing a dialogue context management structure, assigning a unique session identifier, and binding the session to the current user's identity. Subsequently, the electronic device can open a dialogue window or enter a dialogue page on the user interface. The dialogue model generates and displays the first greeting or introductory message, thus formally entering the multi-turn dialogue phase with the user to whom the project belongs.

[0082] For example, when User A is viewing their investment product list, a card pops up in the lower right corner of the interface, stating, "We've discovered asset allocation optimization opportunities for you. Click to view a personalized plan." User A clicks the "Start Planning" button on the card. The system then opens a dialogue panel in the center of the page. The dialogue panel first sends a message from the dialogue model: "Hello, User A, I will assist you in clarifying your financial goals and customizing a personalized planning plan. We can start by discussing your long-term financial vision. What are your specific expectations for retirement?" Thus, a multi-round dialogue is successfully initiated.

[0083] The product recommendation method based on project planning provided in this application can ensure that the project planning visualization component appears in the scenario and time when it is most likely to be accepted and responded to by the user by pushing the project planning visualization component to the user to which the project belongs. By responding to the entry point of multi-turn dialogue, the dialogue model is called to enter a multi-turn dialogue with the user to which the project belongs, which can avoid the bad experience that may be caused by automatically popping up dialogue windows and ensure user experience.

[0084] In some embodiments, in order to further improve the user experience, step 201 involves conducting multiple rounds of dialogue with the user to whom the project belongs based on the dialogue model, and identifying the project planning requirements intent based on the multiple rounds of dialogue. This can be further refined to include steps 401 to 402.

[0085] Step 401: Based on the dialogue model, generate the script generation sub-model to generate push scripts for sending to users of the project in multi-turn dialogues.

[0086] The dialogue generation sub-model is used in the dialogue model to generate natural language responses.

[0087] The dialogue generation sub-model can be a conditional text generation model, capable of outputting guiding dialogue content for the next round that aligns with the business scenario, based on the current dialogue context, pre-defined task objectives, and user profile characteristics. For example, the dialogue generation model might adjust the complexity of its expression based on the user's cognitive level label in the user profile characteristics to ensure that the customer can easily understand it.

[0088] In some embodiments, the dialogue generation sub-model can be based on a pre-trained language model with a Transformer architecture. When applied to project planning scenarios, it can be fine-tuned in a supervised manner using a large number of professional customer service dialogue records, project planner dialogues, and knowledge texts related to the project planning domain.

[0089] When generating push notification messages, the input to the message generation sub-model can include: encoded current conversation history, user basic information tags obtained from the user center service, and the phased target instructions for this conversation. The message generation sub-model automatically generates one or more candidate messages by comprehensively understanding this input information. The electronic device can then select the message with the highest fluency and target match as the final push notification message sent to the user.

[0090] For example, the current task of the dialogue model is to guide the user into the financial situation assessment phase. The dialogue history shows that the user has expressed an initial desire to conduct retirement planning. The dialogue generation sub-model, combining this information, might generate a dialogue like the following: "To tailor a plan for you, we need to understand your current financial situation. Could you briefly tell me your age, after-tax annual income, and approximate current savings?" Step 402: Obtain the user scripts sent by the user to which the project belongs in multi-round dialogues.

[0091] User dialogue refers to the text content entered by the user in the dialogue interface or the text content converted by speech recognition. Obtaining user dialogue is a prerequisite for intent analysis and interaction progression.

[0092] In some embodiments, the front-end application can receive raw user input through a front-end input box in the dialog interface or an integrated speech recognition module. For voice input, a speech-to-text application programming interface (API) can be called to convert the audio stream into text in real time. The converted text can be encapsulated as structured message data, just like the directly entered text. This data includes a user identifier, session identifier, message sequence number, and the message content itself, and is sent to the back-end service for processing in real time via a secure network connection.

[0093] Step 403: Based on the intent recognition sub-model of the dialogue model, perform project planning-related keyword detection on the user's speech, and identify the project planning requirement intent based on the keyword detection results, the user profile of the user to which the project belongs, historical project planning strategies, and project planning knowledge base.

[0094] The intent recognition sub-model is used in dialogue models to understand the true purpose behind user utterances, and is a classification or sequence labeling component. Project planning requirement intent recognition is a multi-factor decision-making process; the intent recognition sub-model can categorize ambiguous user expressions into specific, actionable service intent categories.

[0095] The intent recognition sub-model can be built by fine-tuning a pre-trained model based on BERT or similar technologies.

[0096] The intent recognition sub-model can detect project planning-related keywords in user speech. Specifically, the intent recognition sub-model can perform sequence labeling on the input user speech and identify entities or keywords that are highly related to the preset project domain, such as pension, retirement, savings rate, risk tolerance, etc.

[0097] Then, the intent recognition sub-model can make a comprehensive intent judgment based on the detected keywords, the user profile of the user to which the project belongs, historical project planning strategies, and the project planning knowledge base. Specifically, the intent recognition sub-model can vectorize and concatenate the above information, and output the final intent category probability distribution through a classification layer.

[0098] For example, a user says they are currently concerned about how much money they will receive each month. The intent recognition sub-model identifies "each month" and "money" through keyword detection. Combined with the user profile showing a middle-aged customer, no historical strategies, and a strong correlation between monthly payments and pension payment planning in the knowledge base, it is highly probable that their intent category is a need for pension benefit calculation.

[0099] The user profile can be a static tag obtained from the user center service, such as customer lifecycle stage and asset level. Historical project planning strategies can be obtained from the solution management service, showing whether the user has previously developed solutions and their core parameters. The project planning knowledge base is a structured domain knowledge graph containing the relationships between concepts, products, and strategies.

[0100] The product recommendation method based on project planning provided in this application generates push messages for sending to users of a project in multi-turn dialogues through a dialogue model-based message generation sub-model. This method can dynamically generate dialogue content that fits the context, avoiding fixed message replies and improving user experience. By acquiring user messages sent by users of a project in multi-turn dialogues, and using the intent recognition sub-model of the dialogue model to detect project planning-related keywords in the user messages, the method identifies project planning needs based on the keyword detection results, user profiles of users of the project, historical project planning strategies, and a project planning knowledge base. This accurately identifies users' project planning needs and improves user experience.

[0101] In some embodiments, to further improve user experience and increase product recommendation accuracy, the project-based product recommendation method further includes: Step 501: Identify the guidance stage corresponding to the user to which the project belongs. The guidance stage is the project planning strategy corresponding to the user who has not filled out the project questionnaire, has filled out the project questionnaire, or has generated the project questionnaire. Step 502: Based on the preset prompt rules corresponding to the guidance phase, send prompt information representing the guidance phase of the user to which the project belongs to the project planning customer service.

[0102] In some embodiments, electronic devices can identify the bootstrapping phase through process status records.

[0103] Specifically, when it is detected that the project survey record associated with the current user and session does not exist, the onboarding phase can be identified as the uncompleted project survey phase.

[0104] When a project survey record is detected as existing and in a submitted state, but no associated project planning strategy record is found, the guidance phase can be identified as the completed project survey phase.

[0105] When both submitted project survey records and project planning strategy records in a generated or pending-confirmation state are detected, the guidance phase can be identified as the generated strategy phase.

[0106] Preset prompt rules can be a set of pre-configured information triggering and sending logics bound to different guidance stages. For example, when it is detected that a user is in the stage of not filling out a project survey questionnaire, a prompt message is sent to customer service. The message content may include: The customer has not yet started filling out the questionnaire; please guide the customer to start the information collection process. When it is detected that the project survey questionnaire has been filled out, the prompt message may be: The customer has completed the questionnaire; the system is generating a planning strategy; please prepare to explain the plan to the customer. When it is detected that the strategy has been generated, the prompt message may be: The customer's customized planning strategy has been generated; you can proceed to the plan explanation and confirmation stage.

[0107] The prompt message is used to provide contextual hints to the project planning customer service personnel serving the user, so that the project planning customer service personnel can provide timely human assistance to the user.

[0108] The product recommendation method based on project planning provided in this application identifies the guidance stage corresponding to the user of the project and sends prompt information representing the guidance stage of the user to the project planning customer service based on the preset prompt rules corresponding to the guidance stage. This ensures that customer service can receive standardized operation guidance prompts at different stages, so that customer service does not need to manually query or remember the progress of each customer. It enables customer service to quickly understand the current service context, focus on the professional service itself, reduce the omission or disorder of service steps due to differences in the personal experience of customer service, and improve the overall user experience of project planning services.

[0109] In some embodiments, in order to improve the accuracy of product recommendations, step 301 is further refined to include steps 601 to 603.

[0110] Step 601: Based on the questionnaire pre-filling engine, call at least one data service system to extract multi-source user data of the users to which the project belongs.

[0111] Multi-source user data consists of fragments of user-related information stored across different data service systems. At least one data service system may include a core business system, a customer relationship management system, a historical product purchase record database, an online behavioral analysis platform, etc. Electronic devices can use a questionnaire pre-filling engine to call the corresponding application programming interface or data access channel of each data service system to retrieve multi-source user data of the project's users from that system.

[0112] Step 602: After standard formatting and desensitization of the multi-source user data, fill it into the project survey questionnaire template to obtain a pre-filled questionnaire.

[0113] Since the multi-source user data extracted by the questionnaire pre-filling engine from different data service systems is heterogeneous and may include structured data, semi-structured data, and unstructured data, there are differences in the format encoding of the multi-source user data, and it may contain sensitive information. Therefore, it is necessary to perform standard formatting and de-identification processing on the multi-source user data.

[0114] Standard formatting transforms heterogeneous data into a unified, standardized structure. Standard formatting can include data cleaning and data mapping. Data cleaning includes removing duplicate records, handling null values, correcting obvious formatting errors, and standardizing units of measurement. Data mapping uses a predefined field mapping table to match fields from different data sources to corresponding standardized fields in the project survey template. For example, it can calculate and map the customer age field from system A and the user birthday field from system B to the age integer field in the questionnaire template.

[0115] Anonymization refers to the process of transforming or replacing sensitive personal information before data entry to protect user privacy and meet data security regulations. Anonymization rules can vary depending on the data type. For example, direct identifiers, such as ID card numbers and mobile phone numbers, can be partially masked. Indirect sensitive information, such as precise income or asset values, can be converted into a range.

[0116] The questionnaire pre-filling engine automatically fills in the corresponding answer areas of the standardized and de-identified data according to the structure of the questionnaire template, resulting in a pre-filled questionnaire. Some questions in the pre-filled questionnaire already have suggested answers or default values ​​to save users time in filling out the questionnaire.

[0117] Step 603: Based on the dialogue model's dialogue generation sub-model, generate a questionnaire guidance dialogue, and send the questionnaire guidance dialogue and pre-filled questionnaire to a multi-turn dialogue to guide the project's users to continue filling in the pre-filled questionnaire.

[0118] In some embodiments, the electronic device can invoke a dialogue generation sub-model, inputting the current dialogue context and the state of the pre-filled questionnaire into it, and then using the dialogue generation sub-model to generate a questionnaire guidance dialogue. This guidance dialogue informs the user that part of the content in the project survey questionnaire has already been pre-filled based on the user's information. The purpose of pre-filling is to save the user's time, and the user needs to review and supplement the already filled content. Simultaneously, the guiding dialogue can be used to direct the user's next steps.

[0119] Electronic devices can send the questionnaire prompt dialogue and pre-filled questionnaire data package to the front end simultaneously. In the multi-turn dialogue interface on the front end, the user will first see the generated prompt dialogue, followed by the dynamic rendering of the project survey questionnaire form, with the relevant fields automatically filled in with the processed data. The user can directly modify, confirm, or fill in any remaining blank fields on the form.

[0120] For example, the dialogue model might send the following message: "Dear User B, to help you complete the assessment quickly, we have pre-filled some basic information for you based on your business records with our bank, such as your age, occupation, and the fund products you hold. Please check and supplement other information, such as your family expenditures and specific risk preferences, so that we can generate a more accurate plan." Simultaneously, a questionnaire form with some fields already filled out is displayed in the dialogue window.

[0121] The product recommendation method based on project planning provided in this application involves using a questionnaire pre-filling engine to call at least one data service system to extract multi-source user data of users belonging to the project. This multi-source user data is then formatted and anonymized before being filled into a project survey questionnaire template to obtain a pre-filled questionnaire. This reduces the amount of information users need to input, shortens questionnaire completion time, and improves user experience. Furthermore, the data extracted from at least one data service system is more accurate and reliable than manual completion from memory, improving product recommendation accuracy. A dialogue generation sub-model based on a dialogue model generates a questionnaire guidance dialogue, and the questionnaire guidance dialogue and the pre-filled questionnaire are sent to multiple rounds of dialogue, increasing user participation and ensuring a positive user experience.

[0122] In some embodiments, in order to improve user experience and product recommendation accuracy, steps 701 to 704 are included before step 104.

[0123] Step 701: Add the project planning customer service representative as a member of the conversation to the multi-round conversation, so that the user of the project and the project planning customer service representative can have a conversation in the multi-round conversation.

[0124] Project planning customer service representatives can be qualified professionals or consultants within the relevant industry. For example, in elderly care resource planning services, project planning customer service representatives can be qualified elderly care planning professionals.

[0125] In some embodiments, when the dialogue process reaches a point requiring human intervention, the electronic device can trigger a dialogue member addition command according to preset rules, such as when a complex inquiry is detected, the user explicitly requests to transfer to a human agent, or the process enters the solution confirmation stage. Specifically, if the user's language contains complex questions, ambiguous expressions, or specific keywords such as "human agent," "customer," or "manager," the project planning customer service representative can be added as a dialogue member to the multi-round dialogue. Similarly, if the confidence level of the project planning model's analysis of the questionnaire is found to be below a preset threshold, the project planning customer service representative can be added as a dialogue member to the multi-round dialogue. Furthermore, when the process enters key decision-making stages such as solution confirmation or product purchase, the project planning customer service representative can also be added as a dialogue member to the multi-round dialogue.

[0126] Then, the electronic device can select a project planning customer service account from the pool of certified project customer service accounts based on predefined customer service allocation rules, such as user level, region, problem complexity, and customer service availability. Subsequently, the electronic device can add the unique identifier of this customer service account to the member list data of the multi-turn dialogue session, thereby adding the project planning customer service representative to the dialogue session's member list. This allows the device logged into this customer service account to receive all historical and future messages of the dialogue in real time and has the permission to send messages to the dialogue. Simultaneously, the electronic device can also update the dialogue member panel in the user interface, displaying the project planning customer service representative's avatar or name.

[0127] Subsequently, messages sent by the project planning customer service will be presented to the user in the dialogue flow, just like the messages generated by the dialogue model. At the same time, messages sent by the user will also be synchronized to the operation interface of the project planning customer service, so that the project user and the project planning customer service can have a dialogue in multiple rounds.

[0128] Step 702: Send product infographics for various product types to the multi-turn conversation.

[0129] Product infographics are data representations that present the core attributes and comparative relationships of a product in a graphical and visual way, such as pie charts, bar charts, radar charts, or information cards.

[0130] Product infographics for various product types can intuitively display resource allocation schemes, proportions, and other information for different products, transforming abstract product attributes into intuitive visual graphical representations to help users understand complex project planning and configuration concepts.

[0131] In some embodiments, when the project planning service is an asset planning service, the various product types may include various financial instruments that may be involved in asset project planning, such as protection insurance, stable bond funds, and growth stock funds. The product infographic may be a scatter plot of the risk and return distribution of different types of products, a pie chart of the asset allocation ratio, or an area graph of historical volatility curves.

[0132] Step 703: Send the one-click authorization visualization component to the multi-turn dialogue.

[0133] The One-Click Authorization Visualization Component is an interactive interface element designed to guide users through the legal authorization process necessary for undertaking project planning services with minimal effort.

[0134] The one-click licensing visualization component can include a non-editable summary of the core terms of the licensing agreement, customer confirmation options presented as checkboxes, and a button representing final confirmation.

[0135] Specifically, electronic devices can determine the type of authorization required for the current business process. For example, a data sharing authorization may be used to access other user account information for financial analysis, or a product purchase authorization may be used for potential subsequent product transactions. Based on the authorization type required for the current business process, a corresponding authorization document template is generated. The front-end then uses this template to render an authorization pop-up or a one-click authorization visualization component embedded in the dialog stream.

[0136] Step 704: In response to the user of the project authorizing the visualization component with one click, confirming the user's authorization to generate the project planning strategy, proceed to the step of analyzing the project questionnaire based on the project planning model and generating the project planning strategy corresponding to the project questionnaire.

[0137] A one-click operation refers to a single action taken by the user in the confirmation area of ​​the one-click authorization visual component, such as clicking or tapping. This one-click operation is considered as the user's explicit consent to the authorization content specified in the component.

[0138] In some embodiments, electronic devices can listen for user interaction events with the confirmation button in the one-click authorization visualization component via the front end. When a user interaction event with the confirmation button in the one-click authorization visualization component is detected, the user authorization for the project is confirmed, and a project planning strategy is generated.

[0139] The product recommendation method based on project planning provided in this application adds project planning customer service representatives as dialogue members in a multi-turn conversation, allowing users of the project to engage in dialogue with the project planning customer service representatives. This enables more accurate acquisition of users' project planning needs, thereby improving the accuracy of product recommendations. By sending product information charts of various product types to the multi-turn conversation, the difficulty for users to understand different product types can be reduced. Furthermore, by sending a one-click authorization visualization component to the multi-turn conversation, users can easily authorize the generation of project planning strategies, improving the user experience.

[0140] In some embodiments, to further improve the user experience, step 701 may be followed by step 801 and / or step 802.

[0141] Step 801: Send the query portal of the expert knowledge base to the multi-round dialogue, so that the project planning customer service can query project-related knowledge through the query portal and send the query results of project-related knowledge to the multi-round dialogue.

[0142] The expert knowledge base is a structured digital database that stores domain-specific expertise, policies and regulations, typical cases, solution templates, and other related content.

[0143] The query entry point can be a search box or a menu for browsing by category, providing an interactive interface that allows project planning customer service representatives to input search criteria and access information from the expert knowledge base.

[0144] An electronic device can generate or invoke a search control and embed it within the multi-turn dialogue interface framework, specifically for use by project planning customer service representatives. When a customer service representative enters a query into the control, the electronic device captures the input and converts it into a machine-processable query request. Subsequently, the electronic device performs a retrieval operation in a database storing structured professional knowledge content. This retrieval process may be based on keyword matching, semantic similarity calculation, or a combination thereof. After the retrieval is complete, the electronic device formats the most relevant knowledge content with the highest matching degree to form the query result, which is first fed back to the project planning customer service representative's interactive interface. The customer service representative can review the result and, through a confirmation action, instruct the electronic device to insert the query result as a new message into the multi-turn dialogue history currently shared with the project's user.

[0145] For example, in a multi-turn dialogue interface, a new search box appears in the toolbar area of ​​the project planning customer service representative. The representative enters a description of their question about a specific tax policy. The system searches for relevant content in the background knowledge entries and displays a summary of the found policies and key case points to the representative. If the representative deems this information helpful to the user, they click the send button, and the information then appears in the main dialogue window under the name of the system or the customer service representative for the user to read.

[0146] Step 802: Add the expert user as a dialogue member to the multi-turn dialogue.

[0147] Expert users refer to user accounts that possess specific professional qualifications, distinct from project planning customer service.

[0148] The electronic device can perform an operation that extends session access to an expert user, enabling them to participate in real-time dialogue. Specifically, the electronic device can allow participants in the current dialogue, such as a project planning customer service representative, to initiate an expert invitation. Subsequently, the electronic device selects an online expert user from a predefined set of expert users based on preset expert matching rules, such as matching by professional field tags, and sends a request to the selected expert user's terminal device to join the current dialogue session. Once the expert user accepts the request, the electronic device modifies the member permission list of the multi-turn dialogue session, adding the expert user's account identifier. After this operation is completed, the expert user's client interface will be able to receive the complete dialogue history and real-time message stream, and the content entered by the expert user will also be synchronized to all dialogue members in real time.

[0149] For example, when a project planning customer service representative clicks the "Request Expert Support" button in the chat toolbar, the electronic device can match an online expert whose field matches the project planning service's domain. The expert's device receives an invitation notification prompting them to join an ongoing customer conversation. Upon accepting, the expert's client immediately loads the entire history of the conversation, and a new input box appears where the expert can directly enter professional advice. This advice is then displayed in real-time to both the project's user and the project planning customer service representative.

[0150] The product recommendation method based on project planning provided in this application provides a more professional response to complex user questions by sending the query entry of the expert knowledge base into a multi-turn dialogue and / or adding expert users as dialogue members into the multi-turn dialogue. This improves the accuracy of product recommendations to users and enhances the user experience.

[0151] In some embodiments, to further improve the user experience, steps 901 to 907 are included after step 106.

[0152] Step 901: Monitor at least one project-related event for the user to whom the project belongs.

[0153] After engaging in a conversation with the user of the project to recommend the target product, the electronic device can continuously monitor at least one project-related event for the user. A project-related event can be any event related to the user of the project, the execution process of the project itself, the resources that the project depends on, or the external environment in which the project exists.

[0154] For example, project-related events may include changes in the project resources held by the user, a decrease in derivative resources generated by project execution, and changes in the user's risk appetite.

[0155] Step 902: Determine the event classification of project-related events. Event classification includes: global impact event level, local impact event level, or operational impact event level.

[0156] In some embodiments, electronic devices can classify monitored project-related events according to preset event level classification rules. Different event levels indicate that project-related events have different impact ranges. Among them, the global impact event level refers to events that may disrupt the core objectives of the project or require a complete overhaul of the overall framework; the local impact event level refers to events that only affect a certain module or parameter of the project and require targeted adjustments; and the operational impact event level refers to events that only provide potential signals for updating the profile but do not immediately trigger adjustments to the solution.

[0157] Step 903: Based on project-related events and their corresponding event classifications, predict the fluctuation curves of the project's preset indicators within a preset future time period.

[0158] Project-defined indicators are key quantitative values ​​used to measure the health or achievement of project goals. For example, project-defined indicators may include total project resource supply, the amount of derivative resources generated during project execution, risk level, and project resource liquidity ratio. An indicator fluctuation curve is a prediction of a continuous or discrete sequence of future indicator values ​​over time.

[0159] In some embodiments, electronic devices may employ time series prediction models, such as Long Short-Term Memory (LSTM) networks or Convolutional Neural Networks (CNNs), to predict the fluctuation curves of preset project indicators over a future preset time period, in conjunction with project-related events and their corresponding event classifications. The time series prediction model can be trained using the fluctuation curves of preset project indicators over a period of time following the occurrence of project-related events of different event classifications in historical data. The preset time series prediction model may be a Recurrent Neural Network (RNN), a Long Short-Term Memory (LSTM) network, or a Convolutional Neural Network (CNN), etc.

[0160] Step 904: Input the indicator fluctuation curve into the future event prediction model to predict future events of the project.

[0161] Future event prediction models can be trained by using historical data to analyze indicator change patterns and subsequent specific events. Preset event prediction models can be, for example, recurrent neural networks (RNNs), long short-term memory networks (LSTMs), or convolutional neural networks (CNNs).

[0162] Future event prediction models can infer potential or accompanying derivative events from trend data. Therefore, after inputting the indicator fluctuation curve into the future event prediction model, the model can output one or more project future events that may occur in the future.

[0163] Step 905: Input project future events, project questionnaires, project-related events, and project planning strategies into the project strategy migration model to generate updated project planning strategies.

[0164] The project strategy transfer model can be obtained by supervising the training of a pre-defined strategy transfer model using a historical strategy adjustment event dataset and the corresponding updated strategies adopted in the past.

[0165] The historical strategy adjustment event dataset can include: historical project planning strategies, historical project-related events, historical project questionnaires, and historical actual follow-up events.

[0166] Here, "historical project planning strategy" can be a structured feature of the project planning strategy adopted by a client at time point T. "Historical project-related events" can be project-related events that actually occurred after time point T. "Historical project questionnaires" can be the original project questionnaires completed by the client before time point T. "Historical actual subsequent events" can be a series of subsequent events that actually occurred after time point T, used to replace predicted future events during training, serving as the basis for the model's learning and adjustment patterns.

[0167] The updated strategies actually adopted in history are recorded after time point T. For the above events, the adjusted project planning strategies actually developed by experts for the client or ultimately adopted by the client are the targets that the model needs to learn and approximate.

[0168] Step 906: Calculate the feature similarity between the updated project planning strategy and the candidate recommended products based on the product recommendation model, and select the updated target products based on the feature similarity.

[0169] The specific implementation method for calculating feature similarity can be found in the above embodiments, and will not be repeated here.

[0170] Step 907: Based on the dialogue model, conduct a dialogue with the user of the project to recommend the target product for update.

[0171] In some embodiments, the electronic device can invoke the dialogue model again to explain to the user the reasons for the environmental changes and strategy adjustments in a dialogue format, and recommend newly selected target products to guide the user to update the target products.

[0172] The product recommendation method based on project planning provided in this application can trigger a replanning process by continuously monitoring events, predict future project timelines, deduce potential risks or opportunities from current events, proactively inform users of environmental changes and strategy adjustment logic, and recommend updated products, thereby improving user experience.

[0173] In some embodiments, the product recommendation method based on project planning can be used for product recommendation in elderly care planning projects, and the specific implementation is as follows.

[0174] We continuously monitor the behavioral data of users involved in retirement planning projects on mobile banking apps or online banking. Preset push notification criteria may include: users are over 40 years old, have browsed the retirement information section more than 3 times in the past 7 days, and have zero holdings of retirement-themed products in their asset accounts.

[0175] Analyze users' active time periods and push a visual component entry for elderly care project planning services to customers during their usual login times, such as 8:00 PM to 10:00 PM, through the message center or the APP homepage.

[0176] On the workbench, project planning customer service representatives can enter keywords such as "elderly care planning" and "elderly care resource allocation" into the natural language search box. Through intelligent association, they can directly invoke the workflow for quantitative allocation of elderly care resources, create a service session for the customer in the current conversation, enter multiple rounds of dialogue, and identify the intent of elderly care planning needs based on the multiple rounds of dialogue.

[0177] Once the intent to plan for retirement is identified, a pre-filled questionnaire engine retrieves user information, existing resources, and owned products from at least one data service system. This information is then pre-filled into a project survey questionnaire template, resulting in a completed questionnaire. The pre-filled questionnaire is then sent to a multi-round dialogue to guide the user to continue filling out the questionnaire. Furthermore, real-time validation can be performed during the completion process, prompting the user to confirm any unreasonable values. For example, if annual resource consumption is found to be significantly lower than a reasonable range, a prompt will appear asking the user to confirm.

[0178] After the user completes the pre-filled questionnaire, the project planning model analyzes the questionnaire and generates a corresponding project planning strategy. This strategy includes at least: the overall goal of the elderly care plan, the phased goals of the elderly care plan, the storage method for elderly care resources, the recommended needs for elderly care products, and the disaster recovery strategy for the elderly care project. This strategy is presented to the user and project planning customer service in a pie chart format during multiple rounds of dialogue. If more detailed user information is needed to optimize the plan, a one-click authorization component will pop up, allowing the user to complete data authorization upon confirmation.

[0179] When project planning managers need assistance, they can access the expert knowledge base through the dialogue entry point, such as interpreting the terms and conditions of a complex elderly care planning product, or invite elderly care project planning experts to join the dialogue for real-time consultation.

[0180] Then, a product recommendation model is used to filter target products with matching characteristics from the entire market product pool based on the project planning strategy, and recommendations can be made through dialogue. Recommendation script templates can be customized for different product types and user profiles, such as age and risk preference.

[0181] After a user purchases a product, events related to their retirement plan can be continuously monitored, such as significant changes in the user's resources, announcements of adjustments to the resource allocation of a recommended product, and new changes in retirement planning policies. When such events are detected, the impact can be assessed using an event grading model. For events requiring strategy adjustments, a project strategy migration model is employed to generate updated retirement planning strategies and product recommendation lists. Furthermore, a dialogue model is used to proactively initiate follow-up communication with the user, explaining the reasons for the adjustments and recommending new product options, thereby achieving long-term dynamic management and support for retirement planning.

[0182] Based on the product recommendation method based on project planning provided in the above embodiments, this application also provides a specific implementation of a product recommendation device based on project planning. Please refer to the following embodiments.

[0183] Figure 3 This is a schematic diagram of the structure of a product recommendation device based on project planning provided in another embodiment of this application. See also... Figure 3 The product recommendation device 30 based on project planning provided in this application embodiment includes: Monitoring module 31 is used to monitor whether the user behavior data of the users to which the project belongs meets the preset push conditions; The identification module 32 is used to conduct multi-round dialogues with the user to which the project belongs based on a dialogue model when the preset push conditions are met, and to identify the project planning needs intent based on the multi-round dialogues. The first generation module 33 is used to generate a questionnaire to guide the dialogue based on the dialogue model when the intention of project planning needs is identified, so that the users of the project can fill in the project survey questionnaire based on the questionnaire-guided dialogue. The project survey questionnaire includes at least project requirement information and user information of the users of the project. The second generation module 34 is used to analyze the project questionnaire based on the project planning model and generate the project planning strategy corresponding to the project questionnaire. The project planning strategy includes at least: the overall project goal, the phased project goals, the project resource storage method, the product recommendation requirements, and the project disaster recovery strategy. Calculation module 35 is used to calculate the feature similarity between the project planning strategy and the candidate recommended products based on the product recommendation model, and to select the target product based on the feature similarity. Recommendation module 36 is used to conduct dialogues with users of the project to recommend target products based on a dialogue model.

[0184] In some possible implementations, the monitoring module can also be used for: Collect user behavior data of users belonging to the project; Based on the user behavior detection model, user behavior features and basic project attribute features are extracted from user behavior data; The system determines whether user behavior data meets the preset push conditions based on user behavior characteristics and project basic attribute characteristics. The user behavior characteristics include at least one of user consultation behavior characteristics, user browsing behavior characteristics, and user login behavior characteristics. The project basic attribute characteristics include at least one of project type and project lifecycle characteristics.

[0185] Among some possible implementations, the product recommendation device based on project planning also includes: The push module is used to push project planning visualization components to users of the project. The project planning visualization components contain entry points that trigger multi-turn dialogues. The response module is used to respond to the triggering of a multi-turn dialogue entry point by invoking the dialogue model to enter a multi-turn dialogue with the user to which the project belongs.

[0186] In some possible implementations, the identification module is specifically used for: A dialogue model-based sub-model generates push messages for sending to users of the project during multi-turn dialogues. Obtain the user scripts sent by users belonging to the project in multiple rounds of dialogue; The intent recognition sub-model based on the dialogue model detects project planning-related keywords in user speech, and identifies project planning requirements intent based on the keyword detection results, user profiles of the users to which the project belongs, historical project planning strategies, and project planning knowledge base.

[0187] Among some possible implementations, the product recommendation device based on project planning also includes: The second identification module is used to identify the guidance stage corresponding to the user of the project. The guidance stage is the project planning strategy corresponding to the user who has not filled out the project questionnaire, has filled out the project questionnaire, or has generated the project questionnaire. The sending module is used to send prompts indicating the user's guidance phase to the project planning customer service based on the preset prompt rules corresponding to the guidance phase.

[0188] In some possible implementations, the first generation module is specifically used for: The questionnaire pre-filling engine calls at least one data service system to extract multi-source user data of the users to which the project belongs; After standard formatting and anonymization of multi-source user data, it is filled into the project survey questionnaire template to obtain a pre-filled questionnaire. Based on the dialogue model, a sub-model for generating dialogue is used to generate a questionnaire to guide the dialogue. The questionnaire to guide the dialogue and the pre-filled questionnaire are then sent to multiple rounds of dialogue to guide users of the project to continue filling in the pre-filled questionnaire.

[0189] Among some possible implementations, the product recommendation device based on project planning also includes: Add a module to add project planning customer service representatives as dialogue members in multi-turn conversations, enabling project users to engage in dialogues with project planning customer service representatives in multiple rounds. The second sending module is used to send product information charts of various types of products to multi-round dialogues; The third sending module is used to send the one-click authorization visualization component to multi-turn dialogues; The third generation module is used to respond to the project user's one-click authorization of the visualization component, confirm the project user's authorization to generate project planning strategies, and enter the steps of analyzing the project questionnaire based on the project planning model to generate the project planning strategy corresponding to the project questionnaire.

[0190] Among some possible implementations, the product recommendation device based on project planning also includes: The fourth sending module is used to send the query entry of the expert knowledge base to the multi-round dialogue, so that the project planning customer service can query project-related knowledge through the query entry and send the query results of project-related knowledge to the multi-round dialogue. And / or, a second add module for adding expert users as dialogue members to multi-turn dialogues.

[0191] Among some possible implementations, the product recommendation device based on project planning also includes: The monitoring module is used to monitor at least one project-related event for the user to whom the project belongs; The second determination module is used to determine the event classification of project-related events. The event classification includes: global impact event level, local impact event level, or operational impact event level. The prediction module is used to predict the fluctuation curve of the project's preset indicators within a preset period of time based on project-related events and their corresponding event classifications. The second preset module is used to input the indicator fluctuation curve into the future event prediction model to predict the future events of the project. The fourth generation module is used to input project future events, project questionnaires, project-related events, and project planning strategies into the project strategy migration model to generate updated project planning strategies. The filtering module is used to calculate the feature similarity between the updated project planning strategy and the candidate recommended products based on the product recommendation model, and to filter out the updated target products based on feature similarity. The update module is used to conduct dialogues with users of the project to recommend updated target products based on a dialogue model.

[0192] Each module of the product recommendation device based on project planning provided in this application embodiment can realize the functions of each step of the product recommendation method based on project planning provided in the above embodiment, and can achieve its corresponding technical effects. For the sake of brevity, it will not be described in detail here.

[0193] Figure 4 This is a schematic diagram of the structure of an electronic device provided in another embodiment of this application. See also... Figure 4 The client anomaly detection method described in the above embodiments can be further described in this embodiment as a client anomaly detection device 40, which includes a processor 41 and a memory 42 storing computer program instructions; the processor 41 executes the computer program instructions to implement any of the client anomaly detection methods described in the above embodiments.

[0194] The product recommendation method based on project planning in the above embodiments can be implemented using a computer storage medium. This computer storage medium stores computer program instructions; when these instructions are executed by a processor, they implement any of the product recommendation methods based on project planning in the above embodiments.

[0195] This application also provides a computer program product, including a computer program, which, when executed, implements any of the product recommendation methods based on project planning described above.

[0196] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0197] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0198] The above are merely specific embodiments of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A product recommendation method based on project planning, characterized by, The method comprises: monitoring whether user behavior data of a user to which the project belongs meets a preset push condition; in the case where the preset push condition is met, carrying out multi-round dialogue with the user to which the project belongs based on a dialogue model, and carrying out project planning demand intention identification based on the multi-round dialogue; in the case where the project planning demand intention is identified, generating a questionnaire guide dialogue based on the dialogue model, so that the user to which the project belongs fills in a project questionnaire based on the questionnaire guide dialogue, the project questionnaire comprising at least project demand information and user information of the user to which the project belongs; analyzing the project questionnaire based on a project planning model, generating a project planning strategy corresponding to the project questionnaire, the project planning strategy comprising at least a project total goal, a project phased goal, a project resource storage mode, a product recommendation demand and a project disaster recovery strategy; calculating a feature similarity between the project planning strategy and a candidate recommended product based on a product recommendation model, and screening a target product based on the feature similarity; carrying out dialogue with the user to which the project belongs for recommending the target product based on the dialogue model.

2. The method of claim 1, wherein, Monitoring whether user behavior data of a user to which the project belongs meets a preset push condition comprises: collecting user behavior data of the user to which the project belongs; extracting user behavior features and project basic attribute features from the user behavior data based on a user behavior detection model; judging whether the user behavior data meets a preset push condition based on the user behavior features and the project basic attribute features, wherein the user behavior features comprise at least one of user consultation behavior features, user browsing behavior features and user login behavior features, and the project basic attribute features comprise at least one of project type and project life cycle features.

3. The method of claim 1, wherein, In the case where the preset push condition is met, the method further comprises: pushing a project planning visualization component to the user to which the project belongs, the project planning visualization component comprising an entrance triggering the multi-round dialogue; in response to the entrance of the multi-round dialogue being triggered, calling a dialogue model to enter the multi-round dialogue with the user to which the project belongs.

4. The method of claim 1, wherein, Carrying out multi-round dialogue with the user to which the project belongs based on a dialogue model, and carrying out project planning demand intention identification based on the multi-round dialogue, comprises: generating a push script for sending to the user to which the project belongs in the multi-round dialogue based on a script generation sub-model of the dialogue model; obtaining user scripts sent by the user to which the project belongs in the multi-round dialogue; detecting project planning related keywords based on an intention recognition sub-model of the dialogue model, and carrying out project planning demand intention identification based on the keyword detection result, a user portrait of the user to which the project belongs, a historical project planning strategy and a project planning knowledge base.

5. The method of claim 1, wherein, The method further comprises: identifying a guide stage corresponding to the user to which the project belongs, the guide stage being that the project questionnaire has not been filled in, the project questionnaire has been filled in, or the project planning strategy corresponding to the project questionnaire has been generated; Based on the preset prompt rules corresponding to the guidance phase, prompt information representing the guidance phase of the user to whom the project belongs is sent to the project planning customer service.

6. The method of claim 1, wherein, Based on the dialogue model, a questionnaire is generated to guide the dialogue, including: At least one data service system is invoked based on the questionnaire pre-filling engine to extract multi-source user data of the users to which the project belongs; After standard formatting and de-identification of the multi-source user data, it is filled into the project survey questionnaire template to obtain a pre-filled questionnaire. Based on the dialogue model's speech generation sub-model, a questionnaire-guided dialogue is generated, and the questionnaire-guided dialogue and the pre-filled questionnaire are sent to the multi-turn dialogue to guide the user of the project to continue filling in the pre-filled questionnaire.

7. The method of claim 1, wherein, Before analyzing the project questionnaire based on the project planning model and generating the corresponding project planning strategy, the method further includes: Add the project planning customer service representative as a member of the conversation, so that the user of the project can have a conversation with the project planning customer service representative in the multi-round conversation. Send product information charts for various types of products to the multi-turn dialogue; Send the one-click authorization visualization component to the multi-turn dialogue; In response to the user of the project authorizing the generation of a project planning strategy by using the one-click authorization visualization component, the process proceeds to the step of analyzing the project questionnaire based on the project planning model to generate the project planning strategy corresponding to the project questionnaire.

8. The method of claim 7, wherein, After adding the project planning customer service representative as a member of the conversation, the method further includes: The query portal of the expert knowledge base is sent to the multi-round dialogue, enabling the project planning customer service to query project-related knowledge through the query portal and send the query results of the project-related knowledge to the multi-round dialogue. And / or, add expert users as dialogue members to the multi-turn dialogue.

9. The method of claim 1, wherein, Based on the dialogue model, after conducting a dialogue with the user of the project to recommend the target product, the method further includes: Monitor at least one project-related event of the user to whom the project belongs; Determine the event classification of the project-related events, including: global impact event level, local impact event level, or operational impact event level; Based on the project-related events and their corresponding event classifications, the fluctuation curves of the project's preset indicators within a future preset period are predicted. Input the fluctuation curve of the aforementioned indicator into the future event prediction model to predict future events of the project. Input the project's future events, the project's questionnaire, the project's related events, and the project's planning strategy into the project strategy migration model to generate an updated project planning strategy. The updated project planning strategy and the candidate recommended products are calculated based on the product recommendation model, and the updated target products are selected based on the feature similarity. Based on the dialogue model, a dialogue is conducted with the user of the project to recommend the updated target product.

10. A product recommendation device based on project planning, characterized by, The device includes: The monitoring module is configured to monitor whether user behavior data of a user to whom the project belongs satisfies a preset pushing condition. The identification module is configured to, when the preset pushing condition is satisfied, perform multi-round dialogue with the user to whom the project belongs based on a dialogue model, and perform project planning requirement intention identification based on the multi-round dialogue. The first generation module is configured to, when the project planning requirement intention is identified, generate a questionnaire guide dialogue based on the dialogue model, and cause the user to whom the project belongs to fill a project questionnaire based on the questionnaire guide dialogue, the project questionnaire comprising at least project requirement information and user information of the user to whom the project belongs. The second generation module is configured to analyze the project questionnaire based on a project planning model, and generate a project planning strategy corresponding to the project questionnaire, the project planning strategy comprising at least a project total target, a project phased target, a project resource storage mode, a product recommendation requirement, and a project disaster recovery strategy. The calculation module is configured to calculate a feature similarity between the project planning strategy and a candidate recommended product based on a product recommendation model, and filter out a target product based on the feature similarity. The recommendation module is configured to perform dialogue with the user to whom the project belongs for recommending the target product based on the dialogue model.

11. An electronic device, comprising: The device comprises a processor and a memory storing computer program instructions; and the processor implements the method according to any one of claims 1-9 when executing the computer program instructions.

12. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer program instructions, and the computer program instructions are executed by a processor to implement the method according to any one of claims 1-9.

13. A computer program product, characterised in that, The instructions in the computer program product are executed by a processor of an electronic device, so that the electronic device can perform the method according to any one of claims 1-9.

Images