Project implementation scheme generation method and system, terminal equipment and medium

Abstract

The invention belongs to the technical field of data processing, and discloses a project implementation scheme generation method and system, terminal equipment and a medium, and the method comprises the steps: selecting a corresponding original intelligent learning model according to an application scene of historical project data, carrying out the extraction, processing and labeling processing of the historical project data, and obtaining historical vector data, adjusting model parameters of the original intelligent learning model by using the historical vector data to obtain a target learning model; analyzing the target project data by using a target learning model to obtain key data, and matching the key data with historical vector data by using a preset matching mechanism to obtain target matching data; and after the key data and the target matching data are merged to obtain the merged data, the target learning model is utilized to predict the merged data to obtain the target project implementation scheme, and the high-quality target project implementation scheme is efficiently and accurately generated.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular to a method, system, terminal device and medium for generating a project implementation plan. Background Art

[0002] In the traditional project management field, project implementation plans are based on the project manager's experience to sort out milestones, critical paths, risk identification and implementation plans. This requires a lot of energy from the project manager and is heavily dependent on people's accumulated experience, which is inefficient.

[0003] Although the pre-training-fine-tuning paradigm has become the mainstream technical approach in modern AI, large general-purpose AI models (such as GPT (Generative Pre-trained Transformer) and BERT (Bidirectional Encoder Representations from Transformers)) have powerful language understanding and generation capabilities through training on large-scale unlabeled data. However, these models often lack specific domain expertise and cannot directly meet the application needs of vertical industries. GPT is a series of generative pre-trained language models based on the Transformer architecture developed by OpenAI; BERT is a bidirectional encoder representation model developed by Google.

[0004] While traditional project management tools exist (such as Microsoft Project and Jira), they primarily rely on rules-driven or simple data analysis methods, making them difficult to adapt to complex, dynamic decision-making needs. Microsoft Project, developed by Microsoft, is a professional project management software primarily used for complex project planning, resource allocation, and progress tracking. Jira is an agile project management tool.

[0005] Therefore existing technology still needs to be improved and improved. Summary of the Invention

[0006] The present application provides a method, system, terminal device and medium for generating a project implementation plan, aiming to solve the problem of low efficiency and poor quality of generating project implementation plans in the prior art.

[0007] In a first aspect, an embodiment of the present application provides a method for generating a project implementation plan, comprising: After selecting a corresponding original intelligent learning model according to the application scenario of the historical project data, the model parameters of the original intelligent learning model are adjusted using the historical vector data to obtain a target learning model; Extracting, processing and annotating the historical project data to obtain historical vector data; After analyzing the target project data using the target learning model to obtain key data, the key data is matched with the historical vector data using a preset matching mechanism to obtain target matching data; After merging the key data and the target matching data to obtain merged data, the target learning model is used to predict the merged data to obtain a target project implementation plan.

[0008] In some embodiments, the extracting, processing, and annotating the historical project data to obtain historical vector data further includes: The historical vector data is imported into a vector database to construct a matching mechanism corresponding to the data type of the historical vector data, thereby obtaining the preset matching mechanism.

[0009] In some embodiments, the matching of the key data with the historical vector data using a preset matching mechanism further includes: The successfully matched data are sorted according to their relevance to the target project implementation plan, and the matching data in the target sorting order are selected as the target matching data.

[0010] In some embodiments, after analyzing the target project data using the target learning model to obtain key data, matching the key data with the historical vector data using a preset matching mechanism to obtain target matching data includes: Decomposing the target project data according to target key categories to obtain key data; After identifying and classifying the data intent of the key data, vectorizing the classified key data to obtain vectorized key data; Calculating the similarity between the historical vector data and the vectorized key data, and using the historical vector data corresponding to the highest similarity as the target matching data; Among them, key data include: project requirements, time requirements and quality standards.

[0011] In some embodiments, adjusting the model parameters of the original intelligent learning model using the historical vector data includes: Selecting a corresponding loss function according to the task type of the historical vector data; After using the original intelligent learning model to predict the project implementation plan based on the historical vector data to obtain a predicted output value, using the loss function to compare the predicted output value with the actual output value, and adjusting the model parameters according to the comparison result; The historical project data includes: project process, task decomposition, time factors, quality assurance plan, and risk identification and response strategy; the task types include: development tasks, testing tasks and document writing tasks.

[0012] In some embodiments, the extracting, processing, and annotating the historical project data includes: extracting, based on the target project data, associated data related to the target project data from the historical project data; After cleaning, standardizing, and structuring the associated data, the processed associated data are labeled according to data type, data intent, or context relevance to obtain the historical vector data; Among them, the data intentions include: demand analysis, time planning, line evaluation and design communication plan.

[0013] In some embodiments, using the target learning model to predict the combined data to obtain a target project implementation plan includes: Using the target learning model to perform contextual understanding and logical reasoning on the combined data to generate the target project implementation plan; Among them, the target project implementation plan includes: milestones, key implementation paths, risk identification and response methods, communication plans and project implementation plans.

[0014] In a second aspect, an embodiment of the present application provides a system for generating a project implementation plan, including: An adjustment module is configured to select a corresponding original intelligent learning model based on the application scenario of the historical project data, extract, process, and annotate the historical project data to obtain historical vector data, and use the historical vector data to adjust the model parameters of the original intelligent learning model to obtain a target learning model; A matching module is used to analyze the target project data using the target learning model to obtain key data, and then use a preset matching mechanism to match the key data with the historical vector data to obtain target matching data; The prediction module is used to merge the key data and the target matching data to obtain merged data, and then use the target learning model to predict the merged data to obtain a target project implementation plan.

[0015] In a third aspect, an embodiment of the present application provides a terminal device comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the method for generating a project implementation plan as described above.

[0016] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method for generating a project implementation plan as described above.

[0017] Compared with the existing technology, the present application provides a method, system, terminal device and medium for generating a project implementation plan. The method extracts, processes and annotates historical project data, adjusts the selected original intelligent learning model according to the obtained historical vector data, uses the obtained target learning model to analyze the target project data to obtain key data, and uses a preset matching mechanism to match the key data with the historical vector data to obtain target matching data, then merges the key data and the target matching data to obtain merged data, and inputs the merged data into the target learning model for prediction, thereby quickly generating a real-time, high-quality target project implementation plan, effectively improving the quality and efficiency of project management. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0019] Figure 1 A flow chart of a method for generating a project implementation plan provided in this application; Figure 2 A flow chart for obtaining historical vector data in the method for generating a project implementation plan provided in this application; Figure 3 A flow chart for adjusting model parameters in a method for generating a project implementation plan provided by the present application; Figure 4 A flow chart for obtaining target matching data in the method for generating a project implementation plan provided in this application; Figure 5 A structural diagram of the system for generating the project implementation plan provided in this application.

[0020] Reference numerals: 10 - adjustment module; 20 - matching module; 30 - prediction module. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments.

[0022] The components of the embodiments of the present application generally described and illustrated in the drawings herein may be arranged and designed in a variety of different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed application, but rather merely represents selected embodiments of the present application. All other embodiments obtained by those skilled in the art based on the embodiments of the present application without creative effort are within the scope of protection of the present application.

[0023] Hereinafter, the terms "including", "having" and their cognates used in various embodiments of the present application are intended only to indicate specific features, numbers, steps, operations, elements, components or combinations of the aforementioned items, and should not be understood as excluding the existence of one or more other features, numbers, steps, operations, elements, components or combinations of the aforementioned items or adding the possibility of one or more features, numbers, steps, operations, elements, components or combinations of the aforementioned items. In addition, the terms "first", "second", "third" and the like are only used to distinguish descriptions and should not be understood as indicating or implying relative importance.

[0024] Unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by those skilled in the art to which the various embodiments of the present application belong. The terms (such as those defined in commonly used dictionaries) will be interpreted as having the same meaning as in the context of the relevant technical field and will not be interpreted as having an idealized meaning or an overly formal meaning unless clearly defined in the various embodiments of the present application.

[0025] The present application provides a method, system, terminal device, and medium for generating a project implementation plan. The method for generating the project implementation plan extracts, processes, and annotates historical project data, adjusts the selected original intelligent learning model based on the obtained historical vector data, analyzes the target project data using the obtained target learning model to obtain key data, and uses a preset matching mechanism to match the key data with the historical vector data to obtain target matching data. The key data and target matching data are then merged to obtain merged data, which is then input into the target learning model for prediction, thereby quickly generating a real-time, high-quality target project implementation plan, effectively improving the quality and efficiency of project management.

[0026] The following describes the design scheme of the method for generating the project implementation plan through some specific examples.

[0027] See also Figure 1 , this embodiment of the application provides a method for generating a project implementation plan, including steps S100-S300: S100. After selecting the corresponding original intelligent learning model according to the application scenario of the historical project data, the historical project data is extracted, processed and annotated to obtain historical vector data, and the model parameters of the original intelligent learning model are adjusted using the historical vector data to obtain the target learning model.

[0028] Historical project data includes: project processes, task breakdown, time factors, quality assurance plans, and risk identification and response strategies. Project processes refer to the various project phases and their dependencies; task breakdown is the process of breaking large tasks into smaller units and defining their priorities and schedules; time factors refer to the time distribution required to complete tasks in historical projects; quality assurance plans are specific measures to ensure the quality of project delivery; and risk identification and response strategies are the risks that may be encountered during project implementation and their solutions.

[0029] Data extraction refers to extracting effective information related to project management from historical project data as the basis for subsequent fine-tuning of the AI model. For example, the text description in the project plan is extracted into a task list, time nodes, and dependencies; risk records are converted into structured risk types, probability of occurrence, and response measures. Processing refers to cleaning, normalizing, and standardizing data to remove noise and redundant information. Manual labeling is the manual classification and labeling of data to make it more clearly reflect the laws of project management. For example, marking which tasks are on the critical path and marking which risks are the most common or most serious.

[0030] For example, data preparation begins by collecting relevant data or empirical data accumulated within the company regarding project management as historical project data (including project processes, task breakdowns, time factors, quality assurance plans, and risk identification and response strategies). From a machine language perspective, historical project data can be categorized into the following types: natural language text (e.g., project requirements documents, meeting minutes, task descriptions, and user queries); structured data (e.g., task lists, schedules, resource matrices, and budget information); and semi-structured data (e.g., email exchanges and chat logs).

[0031] Then, the historical project data is processed through data extraction, processing, and manual labeling, and then high-dimensional vector representation is performed to obtain historical vector data (i.e., represented by a string of numbers). In addition, a base model (i.e., the original intelligent learning module) is selected, that is, a suitable general AI large model is selected according to the application scenario of the historical project data. For example, if stronger text generation capabilities are required, the GPT series model can be selected, or if more attention is paid to text encoding and semantic understanding, models such as BERT or RoBERTa can be selected. Among them, RoBERTa: Robustly Optimized BERT Approach‌ is a natural language processing model. The annotated data is represented by a vector through Sentence-BERT encoding, wherein Sentence-BERT encoding is a natural language processing technology based on BERT (Bidirectional Encoder Representations from Transformers), which is used to convert sentences or text fragments into high-dimensional vector representations).

[0032] Afterwards, the historical vector data is stored in a vector database for quick retrieval and recall of relevant information.

[0033] Secondly, the original intelligent learning module is fine-tuned using historical vector data. That is, by defining the target task, designing the loss function in the model, and adjusting the hyperparameters, the target learning model is obtained after adjustment, so that the general AI large model can better understand and generate solutions related to the target project management field, that is, it can understand the terms and concepts in the target project management (such as milestones, critical paths, risk management, etc.), and can automatically generate scientific and reasonable project implementation plans based on the needs of the input new project.

[0034] It can be understood that in this application, by selecting an appropriate original intelligent learning model based on the application scenario, and extracting, processing, and annotating historical project data to obtain historical vector data, the original intelligent learning model parameters are adjusted using the historical vector data. This allows the selection of an appropriate intelligent learning model and adaptive adjustments, thereby improving the accuracy and pertinence of the generated project implementation plan. Furthermore, this not only provides a preliminary understanding of natural language and structured data, but also reduces the computing resources and time required to train the model from scratch.

[0035] In one implementation, see Figure 2 , extract, process and annotate historical project data, including: S101. Extracting associated data related to the target project data from the historical project data based on the target project data; S102: After cleaning, standardizing, and structuring the associated data, the processed associated data are labeled according to data type, data intent, or context relevance to obtain historical vector data.

[0036] Among them, data intent includes: demand analysis, time planning, line evaluation and design communication plan.

[0037] For example, after obtaining the target project data, we first extract data from historical project data. This involves extracting relevant data from the historical project data. For example, we extract data from internal resources such as historical project documents, meeting minutes, task management systems, progress reports, and risk assessment forms. We then extract the text descriptions in project plans into task lists, timelines, and dependencies.

[0038] Then, the associated data is cleaned, standardized, and structured to remove noise and redundant information and obtain processed associated data.

[0039] Data cleaning involves removing redundant information (such as irrelevant notes and duplicate task descriptions), correcting erroneous data (such as inconsistent time field formats and inconsistent numerical units), and supplementing missing values (through statistical analysis or inference based on data from similar projects). Standardization involves unifying data formats (such as dates to YYYY-MM-DD and time units to hours) and standardizing terminology (such as unifying "demand analysis" and "demand research" into "demand definition"). Structuring involves converting unstructured data (such as free text descriptions) or semi-structured data into structured data (such as tables) to facilitate subsequent analysis. Risk records are converted into structured risk types, probability of occurrence, and response measures. For example, "Task A needs to be completed within 3 days and may be affected by insufficient resources" can be converted into: "Task Name: Task A; Estimated Duration: 3 Days; Risk Factor: Insufficient Resources."

[0040] Secondly, the processed associated data is annotated, that is, the processed associated data is classified and labeled according to data type, data intent or context relevance to obtain historical vector data.

[0041] Among them, classification labeling, or type labeling, is to mark project management experience according to data categories. For example: task type: development task, testing task, document writing task, etc.; risk level: high risk, medium risk, low risk; quality assurance measures: code review, unit testing, user acceptance testing, etc. Intent labeling is to label the actual meaning behind each piece of data. For example: "Task A needs to be completed within 3 days" can be labeled as "task time prediction"; "Possible problem of insufficient resources" can be labeled as "risk identification." The following association labeling is to label the logical relationship between data. For example: labeling the dependency between Task A and Task B (such as "Task B can only start after Task A is completed"); labeling the correlation between a risk factor and a specific task (such as the correlation between "insufficient resources" and "Task A").

[0042] In one implementation, see Figure 3 , using historical vector data to adjust the model parameters of the original intelligent learning model, including: S103. Select a corresponding loss function according to the task type of the historical vector data; S104. After using the original intelligent learning model to predict the project implementation plan based on the historical vector data to obtain the predicted output value, the predicted output value is compared with the actual output value using the loss function, and the model parameters are adjusted according to the comparison result.

[0043] Among them, task types include: development tasks, testing tasks, and document writing tasks, etc.

[0044] Exemplarily, the process of adjusting the parameters of the original intelligent learning model is as follows: First, define the type of target task, that is, clarify the specific task that the adjusted original intelligent learning model needs to complete, such as task priority assessment or resource conflict detection. Then, start adjusting the parameters of the original intelligent learning model, mainly including: fine-tuning of supervised learning method and fine-tuning of reinforcement learning method.

[0045] Fine-tuning supervised learning methods involves first designing corresponding loss functions based on the task types of historical project data (including development tasks, testing tasks, and document writing tasks), including cross-entropy loss, mean squared error, and sequence generation loss. The labeled historical project data (i.e., historical vector data) is then divided into training, validation, and test sets. Each example in the training set is an "input-output" pair, where the input is key project data (such as requirements, time constraints, and quality requirements), and the output is the target solution (such as milestones, critical paths, and risk response measures).

[0046] When designing a loss function based on the task type of the historical project data, if the historical project data involves multi-classification or binary classification problems, cross-entropy loss can be used, as it is particularly suitable for classification tasks and has good convergence properties. If the problem involves numerical prediction, mean squared error can be used to measure the average of the squared differences between the predicted value and the true value. If the output is a text-based solution (such as a communication solution), sequence generation loss can be used, which uses negative log-likelihood loss to optimize the generated results.

[0047] The training set is then used to update the parameters of the base model. Model performance is evaluated on the validation set, and hyperparameters (such as learning rate and batch size) are adjusted. The original intelligent learning model is then used to predict the project implementation plan based on the historical vector data from the test set, generating predicted output values. A loss function is then used to quantify the gap between the predicted output values and the true output values from the model training. This gap is then used to further adjust the hyperparameters in the loss function to fine-tune the base model.

[0048] The fine-tuning of the reinforcement learning method is to design the corresponding loss function based on the task type of historical project data, evaluate the effectiveness of the generated solution by simulating different project scenarios, and adjust the model parameters, namely hyperparameters (such as learning rate or batch size, etc.) based on the feedback.

[0049] In one implementation method, historical project data is extracted, processed, and annotated to obtain historical vector data, and then the following steps are further included: The historical vector data is imported into a vector database to construct a matching mechanism corresponding to the data type of the historical vector data, thereby obtaining a preset matching mechanism.

[0050] For example, first, a vector database is established in advance to store a large amount of basic information of historical vector data. The basic information includes task templates, milestone cases, risk records, communication templates, etc. Among them, task templates refer to the best practices for completing similar tasks; milestone cases refer to common milestone settings in different types of projects; risk records refer to risks encountered in previous projects and their solutions; communication templates refer to communication document examples for different types of projects.

[0051] After obtaining historical vector data, import it into a vector database (such as Milvus (an open-source vector database system focused on efficiently handling large-scale similarity search tasks), Faiss (an efficient vector similarity search library particularly suitable for large-scale in-memory vector retrieval), or Pinecone (a hosted vector search engine service that provides out-of-the-box similarity search functionality)). Establish a matching mechanism corresponding to the data type of the historical vector data. Specifically, design recall paths corresponding to the various types of historical vector data (based on the characteristics of project management scenarios, historical vector data includes natural language text, structured data, and semi-structured data). Specifically, design a dedicated recall strategy (i.e., a preset matching mechanism) for each data type: The recall path for natural language text is to match the text using semantic retrieval technology (such as BERT-based sentence embedding). For example, input "Please assign the highest priority task to Team A" to recall the related task descriptions and historical decision records. The recall path for structured data is to build an index (such as an inverted index or vector index) to speed up queries. For example, input the task ID or person in charge information to quickly locate related task records. The recall path for semi-structured data is to use key fields (such as sender, subject, and timestamp) as the basis for recall. For example, input "Discussion on resource conflicts in the past week" to recall related email chains.

[0052] To fully utilize multi-source data, we need to design cross-channel fusion mechanisms, including joint recall and context-awareness. Joint recall involves simultaneously recalling data from multiple channels and combining the results based on weights. Context-awareness dynamically adjusts the priority of each channel based on the context of the user's query. For example, if a user is querying about resource allocation, the resource matrix in the structured data will be prioritized.

[0053] It can be understood that in this application, by designing a matching mechanism corresponding to the data type of the historical vector data, multi-channel recall of data is achieved so that data matching the historical vector data can be quickly obtained.

[0054] S200: After analyzing the target project data using the target learning model to obtain key data, the key data is matched with the historical vector data using a preset matching mechanism to obtain target matching data.

[0055] Exemplarily, after obtaining the target learning model, the target learning model decomposes and analyzes the input target project data to obtain key data, including project requirements, time requirements, quality standards and other constraints.

[0056] Then, a pre-defined matching mechanism is used to perform a parallel, multi-angle match between the key data and the historical vector data. This involves a multi-channel recall of the historical vector data. For example, using cosine similarity or other distance metrics, the vector database is searched for historical data that is most similar to the input key data to obtain the target matching data. Furthermore, in another embodiment, the matching method can also be used to recall target matching data based on task type, time factors (such as duration), and / or quality requirements.

[0057] It can be understood that in this application, by matching key data with historical vector data, it is possible to recall the key data from the vector database so as to output data with high similarity to the historical vector data.

[0058] In one implementation, see Figure 4 Step S200: After analyzing the target project data using the target learning model to obtain key data, the key data is matched with the historical vector data using a preset matching mechanism to obtain target matching data, including: S210. Decompose target project data according to target key categories to obtain key data; S220: After identifying and classifying the data intent of the key data, vectorize the classified key data to obtain vectorized key data; S230: Calculate the similarity between the historical vector data and the vectorized key data, and use the historical vector data corresponding to the highest similarity as the target matching data.

[0059] For example, the specific process of obtaining target matching data is as follows: First, users input target project data into the target learning model using natural language descriptions or structured forms. For example, a natural language description might read: "We need to develop an e-commerce website, which must be completed within three months, with a quality standard of over 95% availability and a budget of less than 100,000 yuan." A structured form might include fields such as 'Project Name,' 'Requirement Description,' 'Time Requirements,' 'Quality Standards,' etc.

[0060] Then, the target learning model breaks down the target project data to obtain key data, including: project requirements: extracting specific task goals (such as "developing an e-commerce website"); time requirements: identifying the start and end times or stage time nodes of the project (such as "complete within 3 months"); quality standards: parsing quality-related constraints (such as "95% availability"), as well as other constraints: such as budget constraints, resource allocation, etc.

[0061] Next, the user's intent is automatically identified based on the input data. These intent types include requirements analysis, timeline planning, line assessment, and design communication plan. Requirements analysis clarifies the core objectives and scope of the project; timeline planning sets milestones and implementation paths; risk assessment identifies potential risks and provides response strategies; and the communication plan is a communication plan within and outside the design team. A pre-trained classification model (such as a BERT-based classifier) is then used to classify the input data's intent and invoke the corresponding submodule based on the classification results.

[0062] Furthermore, the classified key data is vectorized and can also be converted into vector representation through Sentence-BERT encoding to obtain the vectorized key data. The similarity between the historical vector data and the vectorized key data is calculated, and the historical vector data corresponding to the highest similarity is returned as the target matching data.

[0063] In one implementation method, key data is matched with historical vector data using a preset matching mechanism, and then the method further includes: The successfully matched data are sorted according to their relevance to the target project implementation plan, and the matching data in the target sorting order are selected as the target matching data.

[0064] For example, when recalling target matching data, the successfully matched data needs to be sorted by relevance, ensuring that the solutions that best meet the current project requirements are ranked first. The sorting scheme can be based on various indicators, such as the similarity between the data and the input information, the historical success rate of the data, and the timeliness of the data (newer data first).

[0065] Then, the top N historical vector data with the highest matching degree are output as matching data.

[0066] S300: After merging the key data and the target matching data to obtain merged data, the target learning model is used to predict the merged data to obtain a target project implementation plan.

[0067] Among them, the target project implementation plan includes: milestones, key implementation paths, risk identification and response methods, communication plans and project implementation plans.

[0068] For example, after obtaining the target matching data, the AI large model needs to be inferred. First, input preparation is performed, merging the key data and target matching data to generate the merged data, thus forming a complete contextual input for the AI large model. For example, the user input is: "Develop an e-commerce website, complete in 3 months, 95% quality." The recalled target matching data is: "Similar projects are typically completed in 3 phases: requirements analysis, development implementation, and testing and launch." The merged data is then: "Develop an e-commerce website, complete in 3 months, 95% quality. Similar projects are typically completed in 3 phases: requirements analysis, development implementation, and testing and launch."

[0069] The combined data is then fed into the target learning model for prediction, thereby generating a detailed project execution plan, namely the target project implementation plan.

[0070] It can be understood that in this application, the key data and target matching data are first merged to obtain the merged data, which is then input into the target learning model for prediction, thereby quickly, accurately and efficiently generating a detailed target project implementation plan.

[0071] In one implementation method, a target learning model is used to predict the combined data to obtain a target project implementation plan, including: The target learning model is used to perform contextual understanding and logical reasoning on the merged data to generate a target project implementation plan.

[0072] For example, after obtaining the merged data, the inference phase begins: the merged data is fed into the target learning model for prediction, which involves contextual understanding and logical reasoning. The adjusted AI model, drawing on its internal knowledge graph and semantic understanding capabilities, analyzes the merged data in context to capture key information (such as task type, objectives, constraints, etc.) and understand the implicit meaning of the input data, avoiding incorrect reasoning due to ambiguity. For example, if the input data contains ambiguous statements (such as "complete as soon as possible"), the model will infer the specific deadline based on the context.

[0073] Then, after completing the context understanding, the AI model enters the logical reasoning stage: The process of logical reasoning includes: first, task decomposition: dividing the overall task into multiple subtasks and determining the dependencies between the subtasks; second, resource assessment: evaluating the required resources (such as manpower, equipment, and time, etc.) according to task requirements; third, constraint checking: verifying whether the generated plan meets the actual constraints (such as budget constraints, staffing, etc.); fourth, risk prediction: identifying potential risk points and proposing corresponding countermeasures.

[0074] Finally, based on the results of contextual understanding and logical reasoning, the AI big model generates the final project implementation plan, namely the target project implementation plan, which at least includes: milestones, key implementation paths, risk identification and response methods, communication plans, and project implementation plans.

[0075] Milestones are significant milestones or milestones during project execution. Each milestone typically corresponds to a specific goal or deliverable, such as the completion of requirements analysis, system design, or preliminary testing. Milestones are crucial indicators of project progress and are primarily used for schedule control: by setting milestones, complex projects can be broken down into manageable phases for easier monitoring and adjustment; team motivation: completing each milestone boosts team confidence and motivation; and risk management: identifying potential issues promptly to avoid project delays.

[0076] The critical path (Critical Path) refers to the chain of tasks that determines the total duration of a project. Any delay in tasks on the critical path will directly impact the overall project completion time. The Critical Path involves determining the dependencies of all tasks in the project, analyzing the time required for tasks, and identifying the longest task chain. For example, if Task A takes 3 days and Task B takes 5 days, and Task B depends on A, then A + B constitute part of the Critical Path. The Critical Path is primarily used to optimize resource allocation: focusing on prioritizing tasks on the critical path; time management: ensuring that critical tasks are completed on time to avoid affecting the overall schedule; and risk warning: focusing on potential risk points on the critical path.

[0077] Risk identification refers to the early discovery of factors that may have a negative impact on the project; risk response refers to the development of corresponding mitigation measures or emergency plans. Risk identification is the analysis of input data, external environment and internal processes to identify possible risk points. For example, technical implementation difficulties, insufficient resources, changes in market demand, etc. Risk response includes risk avoidance, risk mitigation, risk transfer and risk acceptance. Among them, risk avoidance: taking measures to eliminate the source of risk; risk mitigation: reducing the probability or intensity of risk occurrence; risk transfer: transferring risk responsibility to a third party (such as purchasing insurance); risk acceptance: accepting risks within a controllable range. Risk identification and response methods are mainly used to ensure project success: by identifying and responding to risks in advance, reducing the impact of uncertainty on the project; decision support: providing project managers with a clear basis for risk assessment to help make more informed decisions; and cost control: avoiding additional cost expenditures due to unexpected problems.

[0078] An implementation plan is a specific action plan developed to achieve project goals, including task breakdown, resource allocation, and time scheduling. Task breakdown involves dividing the overall project into multiple subtasks; resource allocation involves determining the personnel, equipment, and technical support required for each task; and time scheduling involves setting start and end times for each task. For example, requirements analysis could be completed in the first week, system design could be completed in the second week, and coding could begin in the third week. The implementation plan's primary functions include guiding execution by providing team members with clear work guidelines; coordinating resources by ensuring the appropriate allocation of resources across tasks; and tracking progress by comparing actual progress against planned progress to promptly correct any deviations.

[0079] A communication plan refers to the strategies and mechanisms developed to ensure a smooth flow of information within and outside the project team. A communication plan includes: communication targets, communication methods, communication frequency, and information delivery. Specifically, communication targets: Identify the relevant parties that need to be communicated with, including team members, customers, suppliers, etc.; communication methods: Select appropriate communication tools and methods (such as meetings, emails, instant messaging tools, etc.); communication frequency: Set a schedule for regular communication (such as weekly meetings, daily stand-up meetings, etc.); and information delivery: Ensure that important information is accurately and promptly delivered to relevant parties. For example, daily stand-up meetings can be used to update task progress, and weekly reports can be used to report project status. The main purpose of a communication plan is to promote collaboration: Ensure information sharing among team members to reduce misunderstandings and conflicts; Improve efficiency: Resolve problems quickly through efficient communication mechanisms; and Enhance transparency: Keep all relevant parties informed of project progress and enhance trust.

[0080] Furthermore, in one implementation method, the performance of the target learning model is tested in a real project management scenario, its indicators such as accuracy and response speed are evaluated, and the model parameters or data set are further optimized based on the test results.

[0081] See also Figure 5 , this application embodiment provides a system for generating a project implementation plan, including: The adjustment module 10 is used to select the corresponding original intelligent learning model according to the application scenario of the historical project data, extract, process and annotate the historical project data to obtain historical vector data, and use the historical vector data to adjust the model parameters of the original intelligent learning model to obtain the target learning model; the matching module 20 is used to use the target learning model to analyze the target project data to obtain key data, and then use the preset matching mechanism to match the key data with the historical vector data to obtain target matching data; the prediction module 30 is used to merge the key data and the target matching data to obtain merged data, and then use the target learning model to predict the merged data to obtain the target project implementation plan. Exemplarily, the method for generating a project implementation plan is applied to a system for generating a project implementation plan. Then, the implementation process of the system for generating a project implementation plan is as follows: First, the corresponding AI big model (i.e., the original intelligent learning model) is selected according to the application scenario of the collected historical project data. After the historical project data is extracted, processed, and labeled, the selected original intelligent learning model is adjusted according to the obtained historical vector data. The target project data is analyzed using the obtained target learning model to obtain key data such as project requirements, time requirements, quality standards, and other constraints. The key data is matched with the historical vector data using preset matching mechanisms such as cosine similarity or other distance measurement methods to obtain target matching data.

[0082] Then, the key data and target matching data are merged to obtain the merged data, which is input into the target learning model for prediction to obtain the target project implementation plan, including milestones, key implementation paths, risk identification and response methods, communication plans, and project implementation plans.

[0083] It can be understood that in this application, a mature general AI big model is adopted as the basic architecture, and historical project data in the project management field is used to carry out targeted fine-tuning of the base model in order to better understand and process natural language instructions and structured data related to project management. In addition, by accumulating historical project data and experience as the core elements of big model reasoning, the intelligence of the reasoning engine is further improved, which feeds back to the business, reduces dependence on the experience of project managers, realizes real-time decision-making and prediction functions, and effectively improves the quality and efficiency of project management.

[0084] The present application also provides a terminal device. Exemplarily, the terminal device includes a processor and a memory, wherein the memory stores a computer program, and the processor runs the computer program to enable the terminal device to execute the functions of each module in the above-mentioned project implementation plan generation method or the above-mentioned project implementation plan generation system.

[0085] The processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, including at least one of a central processing unit (CPU), a graphics processing unit (GPU), a network processor (NP), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The general-purpose processor can be a microprocessor or any conventional processor, etc., and can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of this application.

[0086] The memory may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), and electrically erasable programmable read-only memory (EEPROM). The memory is used to store computer programs, and the processor can execute the computer programs accordingly after receiving an execution instruction.

[0087] The present application also provides a computer-readable storage medium for storing the computer program used in the terminal device. For example, the computer-readable storage medium may include, but is not limited to, various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0088] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely illustrative. For example, the flowcharts and structure diagrams in the accompanying drawings show the possible architectures, functions and operations of the devices, methods and computer program products according to the multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and the module, program segment or a part of the code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in an alternative implementation, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the structure diagram and / or flowchart, as well as the combination of boxes in the structure diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or can be implemented using a combination of dedicated hardware and computer instructions.

[0089] In addition, the functional modules or units in the various embodiments of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0090] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, or the part that contributes to the existing technology, or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a smart phone, personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of this application.

[0091] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A method for generating a project implementation plan, characterized in that: include: After selecting the corresponding original intelligent learning model according to the application scenario of the historical project data, the historical project data is extracted, processed and annotated to obtain historical vector data, and the model parameters of the original intelligent learning model are adjusted using the historical vector data to obtain the target learning model; After analyzing the target project data using the target learning model to obtain key data, the key data is matched with the historical vector data using a preset matching mechanism to obtain target matching data; After merging the key data and the target matching data to obtain merged data, the target learning model is used to predict the merged data to obtain a target project implementation plan.

2. The method for generating a project implementation plan according to claim 1, characterized in that: The extracting, processing and marking the historical project data to obtain historical vector data further includes: The historical vector data is imported into a vector database to construct a matching mechanism corresponding to the data type of the historical vector data, thereby obtaining the preset matching mechanism.

3. The method for generating a project implementation plan according to claim 1, characterized in that: The step of matching the key data with the historical vector data using a preset matching mechanism further includes: The successfully matched data are sorted according to their relevance to the target project implementation plan, and the matching data in the target sorting order are selected as the target matching data.

4. The method for generating a project implementation plan according to claim 1, wherein: After analyzing the target project data using the target learning model to obtain key data, the key data is matched with the historical vector data using a preset matching mechanism to obtain target matching data, including: Decomposing the target project data according to target key categories to obtain key data; After identifying and classifying the data intent of the key data, vectorizing the classified key data to obtain vectorized key data; Calculating the similarity between the historical vector data and the vectorized key data, and using the historical vector data corresponding to the highest similarity as the target matching data; Among them, key data include: project requirements, time requirements and quality standards.

5. The method for generating a project implementation plan according to claim 1, wherein: The adjusting the model parameters of the original intelligent learning model by using the historical vector data includes: Selecting a corresponding loss function according to the task type of the historical vector data; After using the original intelligent learning model to predict the project implementation plan based on the historical vector data to obtain a predicted output value, using the loss function to compare the predicted output value with the actual output value, and adjusting the model parameters according to the comparison result; The historical project data includes: project process, task decomposition, time factors, quality assurance plan, and risk identification and response strategy; the task types include: development tasks, testing tasks and document writing tasks.

6. The method for generating a project implementation plan according to claim 1, characterized in that: The extracting, processing and marking of the historical project data includes: extracting, based on the target project data, associated data related to the target project data from the historical project data; After cleaning, standardizing, and structuring the associated data, the processed associated data are labeled according to data type, data intent, or context relevance to obtain the historical vector data; Among them, the data intentions include: demand analysis, time planning, line evaluation and design communication plan.

7. The method for generating a project implementation plan according to claim 1, characterized in that: The method of using the target learning model to predict the combined data to obtain a target project implementation plan includes: Using the target learning model to perform contextual understanding and logical reasoning on the combined data to generate the target project implementation plan; Among them, the target project implementation plan includes: milestones, key implementation paths, risk identification and response methods, communication plans and project implementation plans.

8. A system for generating a project implementation plan, characterized in that: include: An adjustment module is configured to select a corresponding original intelligent learning model based on the application scenario of the historical project data, extract, process, and annotate the historical project data to obtain historical vector data, and use the historical vector data to adjust the model parameters of the original intelligent learning model to obtain a target learning model; A matching module is used to analyze the target project data using the target learning model to obtain key data, and then use a preset matching mechanism to match the key data with the historical vector data to obtain target matching data; The prediction module is used to merge the key data and the target matching data to obtain merged data, and then use the target learning model to predict the merged data to obtain a target project implementation plan.

9. A terminal device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the steps of the method for generating a project implementation plan according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that A computer program is stored, and when the computer program is executed by a processor, the steps of the method for generating a project implementation plan according to any one of claims 1 to 7 are implemented.

Images