Task plan generation method and system based on large model

Abstract

The invention discloses a task plan generation method and system based on a large model, and relates to the technical field of intelligent information scheduling. Comprising a data pre-processing module, a man-hour evaluation module, a task modeling module, a large model reasoning module and a performance analysis module. By obtaining enterprise order data, process man-hour data and system configuration data, hierarchical processing is performed on tasks, process complexity and correction man-hours are calculated, and man-hour demands are generated in combination with equipment utilization and cooperation number of people; the orders are divided into a deduction task plan group, a process dependency graph is constructed, and a large model reasoning module disassembles tasks based on historical task samples to generate sub-task working hours; task allocation is carried out in combination with the special ability and the overall ability of the employees, the processing duration and the optimal value are calculated, and an optimal task plan is screened; the method realizes accurate man-hour prediction, reasonable resource allocation, employee load balancing and task execution efficiency improvement, and is suitable for scenes such as software development, project management and production scheduling.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent information scheduling, in particular to a task plan generation method and system based on a large model. BACKGROUND

[0002] In recent years, the development of artificial intelligence, especially large model technology, has provided new methods for task plan generation. Large models can automatically master the execution mode, sequence and priority distribution of tasks by learning a large amount of historical task data and context information, and generate more reasonable and efficient task plans in complex environments.

[0003] With the development of information technology, enterprises will generate a large number of complex tasks in the fields of software development, project management and production scheduling. Traditional task planning methods mainly rely on manual experience or fixed rules, such as Gantt chart, critical path method (CPM) and priority scheduling method, to decompose, sequence and allocate resources to tasks.

[0004] However, the existing methods usually estimate the task working hours or complexity through experience, lacking consideration of task procedure complexity, collaboration number and equipment utilization rate, etc., which may lead to inaccurate task scheduling. When allocating tasks, the special ability and overall ability of employees and the dependency relationship between tasks are not fully considered, which may result in low resource utilization efficiency, unbalanced employee workload and insufficient task execution efficiency. In addition, traditional methods are difficult to intelligently optimize according to historical data and task characteristics in dynamic environments, so that the task plan often lacks executability and optimization effect when facing complex dependency relationships or urgent deadlines. SUMMARY

[0005] In view of the problems existing in the above background art, a task plan generation method and system based on a large model are proposed.

[0006] The purpose of the present application can be achieved by the following technical solutions: on the one hand, a task plan generation method based on a large model is provided, comprising:

[0007] S1: obtaining business order data, process and working hour data, and company system configuration data from an enterprise database to obtain task original data; performing hierarchical processing on the obtained task original data to obtain a set of task plan groups; obtaining each task plan group to construct a task execution frame to obtain a task plan;

[0008] S2: obtaining the task execution sequence in the task plan to obtain the total task processing time and the total task value; comparing the total task processing time with the task processing period in the business order data, if the total task processing time is less than the task processing period, storing the task plan as a candidate task plan; comparing each total task value of the task plan in the candidate task plan set, selecting the maximum total task value as the optimal task plan.

[0009] As a preferred embodiment of the present application, the process of obtaining the task original data and performing hierarchical processing to obtain the task plan group set is as follows:

[0010] The business order data is disassembled, the company system configuration data is disassembled, and the process and working hour data are disassembled; after hierarchical processing, the order is allocated to different task plan groups according to the working hour demand, daily production capacity and delivery deadline urgency of the order, and the task plan group set is obtained.

[0011] As a preferred embodiment of the present application, the process of obtaining the task original data and performing hierarchical processing to obtain the task plan group set is as follows:

[0012] According to the process dependency relationship, the tasks in each task group are divided into process level nodes; the process level nodes are connected according to the enterprise equipment and shift constraints to obtain the process order of the process level nodes as a process dependency graph; each node in the process dependency graph is bound with fixed number of collaborative people, specified production line or machine type and required working hours of the task; the priority of the corresponding node is valued according to the priority, delivery deadline urgency and working hour length of the task to obtain the comprehensive priority value of each node; the task execution frame is obtained through the above process; based on the obtained task execution frame, the scheduling rule is used to generate the sequence to obtain the task plan.

[0013] As a preferred embodiment of the present application, the process of disassembling the business order data, disassembling the company system configuration data and disassembling the process and working hour data is as follows:

[0014] The business order data is disassembled: based on the segmented interval discrimination algorithm, the order quantity is compared with the preset interval threshold value, and the batch interval to which the order belongs is output; according to the process category mapping relationship in the database, the type and process category are corresponded by using the hash mapping retrieval algorithm, and the corresponding process complexity coefficient is output according to the existing process category complexity matrix; for the delivery period, the minimum and maximum normalization method is used to obtain the delivery deadline urgency factor;

[0015] The company system configuration data is disassembled, and the number of collaborative operation people is found from the fixed number table of the system according to the product type and quantity interval; if the order contains multiple process links, the fixed number of people for each link is found respectively;

[0016] The process and time data are decomposed to obtain the time requirement: the total order time requirement is output by multiplying the single-piece time, quantity and process complexity; the total order time requirement is corrected by the equipment utilization rate to obtain the total order time requirement;

[0017] Based on the hierarchical processing result of the obtained task original data, the capacity and daily demand are calculated to obtain the capacity per shift, daily capacity and order required days.

[0018] As a preferred embodiment of the present application, the specific process of obtaining the task plan group set is as follows:

[0019] The total time of each order, the daily processable time of the production line and the delivery urgency factor of the order are extracted; the orders are sorted from high to low according to the delivery urgency sorting algorithm; the total time of the order and the daily capacity of the production line are matched by using the iterative distribution algorithm; if the total time of the order is less than the residual daily capacity value of a certain production unit, the order is distributed to the task plan group; if the total time of the order is greater than the available daily capacity value, the order is split into several sub-tasks and distributed to different task plan groups; the task plan group set is formed by statistical calculation after distribution.

[0020] As a preferred embodiment of the present application, the specific process of obtaining the task execution sequence in the task plan to obtain the total task processing time and the total task value is as follows:

[0021] Each task in the task execution sequence is decomposed to obtain the task quantity of each sub-task; the employee's special ability value and overall ability value are obtained based on the obtained task quantity; the task quantity is matched with the business type in the task and the employee's special ability value and overall ability value; the processing time and the task value of each task are obtained based on the processing time and the task value of each task; the total task processing time and the total task value are obtained by statistical calculation of the processing time and the task value of all tasks in the task plan.

[0022] As a preferred embodiment of the present application, the specific process of obtaining the processing time and the task value of each task is as follows:

[0023] The attribute information flow of the sub-task node is input into the large model reasoning engine, and the large model performs inference calculation based on the historical task samples in the enterprise database to output the task quantity value of the sub-task node;

[0024] The historical task processing data is obtained from the enterprise database, the obtained indexes are input into the large model reasoning engine combined with the employee's personal history, and the special ability value and overall ability value of each employee are output; then the task quantity and the employee's ability value are mapped, and the predicted efficiency of the employee to complete the sub-task is calculated;

[0025] Finally, an optimization model is established to minimize the total working hours and balance the staff load: each sub-task node must be assigned to at least one staff who meets the ability threshold; the total assigned working hours of the staff cannot exceed the available working hours after the correction of the utilization rate;

[0026] The task volume is matched and distributed in combination with the business type in the task and the special ability value and the overall ability value of the staff: the task processing duration assigned to the staff is obtained according to the matching efficiency of the staff; the task value is obtained on the basis of the processing duration in combination with the task priority.

[0027] As a preferred embodiment of the present application, the calculation process of the process complexity coefficient is:

[0028] The constituent factors of the process complexity coefficient are defined as a plurality of sub-factors unique to information technology projects, including: a code complexity factor, a requirement frequency factor, an integration complexity factor, a dependency maturity factor, a migration risk factor, a security compliance factor, an automated test difficulty factor, a deployment and operation complexity factor, and a performance scalability requirement factor; the sensitivity of each sub-factor is preset according to the different influences of the sub-factors on complexity; the credibility of the values of each sub-factor is obtained according to historical data as a data confidence parameter;

[0029] Each sub-factor sensitivity parameter is subjected to exponential transformation to obtain an uncorrected exponential factor, and the uncorrected exponential factor is combined with the data confidence parameter to obtain a corrected exponential factor; for a certain process, all the corrected exponential factors are multiplied to obtain a comprehensive complexity amplification factor of the process;

[0030] After obtaining the comprehensive complexity amplification factor, the standard working hours T i and the average working hours T avg are combined to obtain the basic complexity coefficient of the process; the basic complexity coefficient is corrected according to the configuration of the number of collaborative workers to output a corrected complexity coefficient; the corrected complexity coefficient is combined with the equipment utilization rate to obtain a utilization complexity coefficient; and the utilization complexity coefficients of all processes are obtained through hierarchical superposition to obtain the process complexity coefficient.

[0031] As a preferred embodiment of the present application, the specific process of obtaining the total working hours of the order is:

[0032] The total available time in the running log is extracted, the actual running time is divided by the total available time by using a ratio calculation model, and an initial utilization rate value is obtained;Based on the initial utilization rate value, the downtime maintenance time is obtained and classified, the planned maintenance and unplanned downtime are distinguished, the logical discrimination algorithm is used to correct the unplanned downtime, the initial utilization rate is corrected, and the corrected utilization rate value is obtained;The corrected utilization rate result is normalized by using the minimum and maximum normalization method to obtain the final utilization rate.

[0033] Another aspect of the application provides a large model-based task plan generation system, comprising: a data preprocessing module, a work time evaluation module, a task modeling module, a large model reasoning module and an efficiency analysis module;

[0034] The data preprocessing module is used for obtaining business order data, process and work time data, company system configuration data, performing cleaning, standardization and hierarchical decomposition to output task raw data;

[0035] The work time evaluation module is used for receiving task raw data, calculating corrected total work time and process complexity coefficient based on complexity factor, work time and equipment utilization rate to form work time demand result;

[0036] The task modeling module is used for dividing orders into deduction task plan groups according to work time demand, daily capacity and delivery deadline urgency, and constructing process dependency graph to generate task execution framework;

[0037] The large model reasoning module is used for decomposing the task execution framework, combining historical task samples, and obtaining task quantity values of subtasks by large model reasoning to generate task nodes;

[0038] The efficiency analysis module is used for distributing task node sets, calculating task processing time and task merit value, and obtaining candidate task plans by statistics, and outputting the optimal task plan by screening.

[0039] Compared with the prior art, the application has the following advantages:

[0040] 1、The application constructs a multi-dimensional factor system of process complexity coefficient, and combines standard work time, collaborative operation number and equipment utilization rate for correction, which more objectively quantifies the task complexity, improves the accuracy of work time prediction, avoids the task scheduling distortion problem caused by single complexity calculation dimension in the prior art, and makes the task plan generation result fit the actual production and development environment of the enterprise.

[0041] 2、The present application introduces a large model reasoning mechanism, decomposes the task, intelligently matches the task with the employee's special ability value and the overall ability value, and generates a task plan that meets the deadline constraint and has the highest overall optimal value. The allocation of task resources is realized before the delivery cycle is guaranteed, the overall task execution efficiency is improved, the employee load balancing is optimized, and the problems of low execution efficiency and insufficient resource utilization caused by relying on manual experience or fixed rule allocation in the prior art are solved. BRIEF DESCRIPTION OF DRAWINGS

[0042] For the convenience of those skilled in the art, the present application will be further described below with reference to the accompanying drawings.

[0043] Figure 1 The step flowchart of the present application is shown in the figure.

[0044] Figure 2 The principle diagram of the present application is shown in the figure. DETAILED DESCRIPTION

[0045] The technical solutions of the present application will be described below in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0046] It should be understood that the terms "include" and "contain" used in the specification and claims of the present disclosure indicate the presence of the described features, whole, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, whole, steps, operations, elements, components and / or their sets.

[0047] It should also be understood that the terms used in the present disclosure are only for the purpose of describing specific embodiments and are not intended to limit the present disclosure. As used in the present disclosure and claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include the plural forms. It should be further understood that the term "and / or" used in the present disclosure and claims means any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0048] Referring to Figure 1 The present application provides a large model-based task plan generation method, which includes:

[0049] S1: Obtain the business order data of the enterprise and process it to obtain the task execution sequence, which is denoted as task plan; specifically:

[0050] S11: The raw task data includes business order data, process and working hour data, and company policy configuration data obtained from the enterprise database. Business order data includes order number, product type, product quantity, delivery cycle and priority. Process and working hour data includes standard working hours per piece, process complexity coefficient and equipment utilization rate. Company policy configuration data includes setting a fixed number of collaborative workers according to product type, process link and quantity range.

[0051] The calculation process for the process complexity coefficient is as follows:

[0052] The constituent factors of the process complexity coefficient are defined as multiple sub-factors unique to information technology projects, with each sub-factor taking values ​​in the range [0, 1]. Specifically, they include:

[0053] Code complexity factor s1: This factor characterizes the complexity of the code implementation. The metrics include the complexity of the algorithm within the module, the number of branches and loops, and the cognitive burden on the developer. For simple CRUD operations, the value is 0.1; for numerous business logic branches, the value is 0.6; and for complex algorithms or concurrency control, the value is 0.9.

[0054] Demand frequency factor s2: This factor indicates whether the demand definition is clear and how frequently it changes; a higher value is used when the demand is vague or changes frequently, and a lower value is used when the demand is clear and changes less.

[0055] Integration complexity factor s3: describes the degree of integration between the system to be developed and external systems, third-party APIs, or heterogeneous platforms; the more external systems that need to be connected and the greater the differences, the higher the value of this factor.

[0056] Dependency Maturity Factor s4: Reflects the stability of third-party libraries or external dependent components. This factor is an inverse indicator; the lower the maturity of the dependency, the higher its complexity. In specific calculations, s4 is set to 1 - mat, where mat is the maturity index of the dependency.

[0057] Migration risk factor s5: reflects the scale and compatibility risks of data migration involved in the task; if large-scale data migration or core data model changes are involved, this factor will have a higher value.

[0058] Security and compliance factor s6: This factor is used to characterize the difficulty of a task in terms of security and compliance; it takes a high value when sensitive data needs to be processed, strict security levels need to be met, or specific industry compliance audits need to be conducted.

[0059] Automated testing difficulty factor s7: This factor describes the testability of a task in the automated testing phase. If the system modules are difficult to modularize or highly dependent on external systems, the difficulty of automated testing increases, and the value of this factor will rise.

[0060] Deployment and maintenance complexity factor s8: This factor is used to characterize the difficulty of system deployment and maintenance; if the task involves multi-environment deployment, rolling upgrades, zero-downtime switching or complex rollback mechanisms, the value of this factor will be higher.

[0061] Performance scalability requirement factor s9: This factor describes the technical challenges of a system in high-concurrency, low-latency, or distributed consistency scenarios. The more stringent the performance and scalability requirements, the higher the value of this factor.

[0062] To further reflect the different impacts of various sub-factors on complexity, this invention sets a sensitivity parameter β for each factor. k >0, k=1,2,…,9; where higher sensitivity is set for key factors, and sensitivity close to 1.0 is set for secondary factors; and the confidence level of each sub-factor value is obtained based on historical data, which is used as the data confidence parameter c. k ∈[0,1].

[0063] By applying an exponential transformation to each sub-factor and each factor with a sensitivity parameter, the uncorrected exponential factor g is obtained. k :g k =exp(β) k ×s k ); In obtaining the uncorrected exponential factor g k Based on the data confidence parameter c k The corrected exponential factor g is obtained. k ':

[0064] For a given process i, the overall complexity amplification factor P is obtained by multiplying all the corrected exponential factors. i :

[0065] In obtaining the comprehensive complexity amplification factor P i Then, combined with the standard working time T for this process. i With average working hours T avg The basic complexity coefficient of the process is obtained.

[0066] The basic complexity coefficient is adjusted based on the number of people working collaboratively, using the following formula: Output corrected complexity coefficient in, The number of collaborative workers allocated according to the system; then, combined with the equipment utilization rate, the operation complexity coefficient is obtained, and then calculated using the formula. We obtain, where R i The utilization rate of the equipment used in this process belongs to [0, 1].

[0067] Then, by hierarchically superimposing the operational complexity coefficients of all processes, the process complexity coefficient is obtained. Where i = 1, 2, ..., n.

[0068] S12: Perform layered processing on the acquired raw task data, specifically breaking it down as follows:

[0069] Commercial order data is broken down into order quantities. Based on a segmented interval discrimination algorithm, the order quantities are compared with preset interval thresholds to output the corresponding batch intervals. Product types are also analyzed. Based on the process category mapping relationship in the database, a hash mapping retrieval algorithm is used to complete the correspondence between types and process categories. The corresponding process complexity coefficients are output based on the existing process category complexity matrix. For delivery cycles, minimum and maximum normalization methods are used to map different delivery lengths to the [0, 1] interval to obtain the delivery urgency factor.

[0070] The company's policy configuration data was broken down, and the number of collaborative workers N was retrieved from the fixed number of employees table based on product type and quantity range. fixed If the order involves multiple process steps, find out the fixed number of people required for each step.

[0071] Decompose the process and time data, and calculate the time requirement: output the total time requirement H of the order by multiplying the time per piece by the quantity and process complexity. total The total working hours H of the equipment order are obtained by adjusting the equipment utilization rate. total ′, specifically:

[0072] The total available time and actual running time in the operation log are extracted. A ratio calculation model is used to divide the actual running time by the total available time to obtain the initial utilization rate. Based on this initial utilization rate, a correction is performed: downtime maintenance time in the records is obtained and categorized, distinguishing between planned maintenance and unplanned downtime. A logical discrimination algorithm is used to deduct and correct unplanned downtime, thus correcting the initial utilization rate to obtain the corrected utilization rate. Finally, the corrected utilization rate result is subjected to minimum and maximum normalization methods, limited to the interval [0, 1], to obtain the final utilization rate R. i ;

[0073] Based on the obtained utilization rate, the utilization rate R of the corresponding equipment for each process is obtained. i The original total working hours H total Divide by utilization rate R i The corrected total working hours H are obtained. total ′;

[0074] Based on the hierarchical processing of the acquired raw task data, capacity is calculated and matched with daily demand, including capacity per shift: Cap. shift =N fixed ×H shift Daily production capacity: Cap day =Caps hift Number of shifts; Number of days required for the order: The integer symbol.

[0075] Based on the order's time requirements, daily production capacity, and delivery urgency, the orders are assigned to different simulation task planning groups, and a set of simulation task planning groups is obtained; specifically:

[0076] Extract the total working hours H for each order total The system considers the daily processing time of the production line and the urgency factor of order delivery dates. A sorting algorithm based on urgency is used to rank orders from highest to lowest urgency. An iterative allocation algorithm is then applied to the ranked orders to match the total processing time of the orders with the daily production capacity of the production line: if the total processing time of the orders is less than the remaining daily capacity of a production unit, all orders are allocated to that task planning group; if the total processing time of the orders is greater than the available daily capacity, the orders are split into several sub-tasks and allocated to different task planning groups. After the above allocation calculations, a set of projected task planning groups is statistically generated.

[0077] It should be noted that during the allocation process, the urgency of the delivery date is taken into account when adjusting the allocation priority. Orders with high urgency are given priority to be arranged within the available timeframe.

[0078] S13: Obtain the task execution frame for each simulation task plan group. Specifically, this involves: breaking down the tasks within each task group into process-level nodes based on process dependencies; node attributes include working hours, number of personnel, and equipment requirements; connecting the process-level nodes according to enterprise equipment and shift constraints to obtain the process sequence of the process-level nodes, which serves as the process dependency graph; binding each node in the process dependency graph with: a fixed number of collaborators, a specified production line or machine model, and the required working hours for the task; assigning priority values ​​to the corresponding nodes based on task priority, delivery urgency, and working hours.

[0079] The delivery urgency factor and working time length are normalized to map data of different dimensions to the [0, 1] interval. Then, an empirical rule-based judgment method is used to combine the task priority, normalized delivery urgency, and normalized working time length to obtain the comprehensive priority value for each node. The steps are as follows:

[0080] For nodes with high task priority, their overall priority value is directly increased; for nodes with high delivery urgency, additional weight is added in the calculation; for task nodes with long working hours, the priority is appropriately adjusted according to empirical rules; the task execution frame is obtained through the above process.

[0081] S14: Based on the obtained task execution frame, use scheduling rules to generate the order, prioritizing tasks with high priority and tight delivery time, and then tasks with short working time. Generate the execution order and time window of tasks within each group, which is recorded as the task plan.

[0082] S2: Obtain the task plan for analysis, specifically:

[0083] S21: Obtain the task execution order within the task plan, and decompose each task according to the task execution order to obtain the task quantity of each subtask, specifically:

[0084] The task is structured and broken down according to the execution order of the task plan: When the task belongs to software development IT tasks, the breakdown includes functional modules such as requirements analysis, interface development, database design, front-end implementation, unit testing, system integration testing, and deployment; when the task belongs to system operation and maintenance IT tasks, the breakdown dimensions include steps such as environment configuration, service deployment, script execution, monitoring indicator setting, disaster recovery switching, and fault rollback. It should be noted that each sub-task unit forms an independent task node, with node attributes including: sub-task name, estimated working hours, resource dependencies, and execution sequence constraints. After task parsing, the original sequential tasks are mapped to obtain sub-task nodes.

[0085] The attribute information stream of the obtained sub-task nodes is input into the large model inference engine, which then performs inference calculations based on historical task samples in the enterprise database: H sub =H node ×W complex Output the task quantity value H of the subtask node sub Among them, Hn ode W represents the base time value for the task node. complex To derive the complexity weights based on task complexity, the large model obtains them based on the correlation between the distribution of working hours and task complexity in historical task samples.

[0086] S22: Task allocation is based on the obtained task volume, including processing historical task processing data to obtain employee specific ability values ​​and overall ability values, specifically:

[0087] Retrieve historical task processing data from the enterprise database and extract indicators such as task completion time, defect rate, rework rate, and cross-domain task completion rate for each employee;

[0088] The acquired metrics, combined with employee resumes, are input into the large-scale model inference engine to output each employee's specialized competency score, ZXL. i1 Compared to the overall ability value ZTL i1 Among them, the specialized ability value corresponds to the competence in a certain type of task, and is calculated as follows: Among them, S i,j W is the performance score of employee i1 in the j-th task of this type. complex，j Let n be the importance weight of the j-th task. d This represents the number of tasks of this type; the overall capability value is obtained by weighted fusion of all specialized capabilities: Where, β d Preset contribution weights for different task domains;

[0089] Then based on the task volume H sub Mapping the values ​​to employee competency scores, and calculating the predicted efficiency E of the employee in completing the subtask. i1,j : Wherein, γ is the experience correction coefficient, which is used to adjust according to the employee's performance fluctuations over the past 3 months;

[0090] Finally, with the goal of minimizing total working hours and balancing employee workload, an optimization model min Z is established: Among them, H p,i Set the planned working hours limit for employee i1; each subtask node must be assigned to at least one employee who meets the capability threshold; the total allocated working hours for an employee must not exceed their available working hours after the utilization rate is adjusted.

[0091] S23: Increase the task quantity H sub,j With business type D within the task j1 Combined with employees' specialized ability scores ZXL i1 (d) Match and assign tasks to the overall capability value ZTLi1; based on the matching efficiency of employee i1, obtain the task processing time T assigned to employee i1. i1,j : Where λ is the preset adjustment weight of overall capacity on efficiency; based on the processing time and the task priority Pj, the task excellence value U is obtained. i1,j : Where η is a comprehensive adjustment factor used to balance urgency and fairness in task allocation. If U i1,j The larger the value, the higher the matching degree and priority value of employee i1 in executing task j.

[0092] Based on the processing time T of each task i1,j With task merit value U i1,j The processing time T for all tasks within the task plan. i1,j and task processing excellence value U i1,jThe system calculates the total task processing time and overall task performance. It then compares the total task processing time with the task processing cycle in the business order data. If the total task processing time is less than the task processing cycle, the task plan is stored as a candidate task plan. Finally, it compares the overall task performance of each task plan in the candidate task plan set and selects the plan with the highest overall task performance as the optimal task plan.

[0093] Please see Figure 2 As shown, another aspect of the present invention provides a task plan generation system based on a large model, including: a data preprocessing module, a time assessment module, a task modeling module, a large model inference module, and a performance analysis module;

[0094] The data preprocessing module is used to acquire business order data, process and working hour data, and company policy configuration data, and to clean, standardize, and decompose them into layers to output the raw task data.

[0095] The time assessment module receives the original task data, calculates and corrects the total working hours and process complexity coefficients based on the complexity factor, working hours and equipment utilization rate, and forms the working hour requirement result.

[0096] The task modeling module is used to divide orders into task planning groups based on working hours, daily capacity and delivery urgency, and to build process dependency graphs to generate task execution frameworks.

[0097] The large model inference module is used to decompose the task execution framework and, in conjunction with historical task samples, infer the task quantity value H of the sub-tasks from the large model. sub Generate a set of task nodes.

[0098] The performance analysis module combines employees' specific ability values ​​and overall ability values ​​to allocate task node sets, calculate task processing time and task excellence value, statistically obtain candidate task plans, and filter and output the optimal task plan.

[0099] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A task plan generation method based on a large model, characterized in that, include: S1: Obtain the raw task data by acquiring business order data, process and working hour data, and company policy configuration data from the enterprise database; The original task data is processed in layers to obtain a set of inferred task plans; Each simulation task plan group is used to construct a task execution frame, thereby obtaining the task plan; S2: Obtain the task execution order within the task plan to get the total task processing time and the total task efficiency value; compare the total task processing time with the task processing cycle in the business order data; if the total task processing time is less than the task processing cycle, store the task plan as a candidate task plan. The overall merit value of each task plan in the candidate task plan set is compared, and the task plan with the largest overall merit value is selected as the optimal task plan.

2. The task plan generation method based on a large model according to claim 1, characterized in that, The process of performing layered processing on the acquired raw task data to obtain the deduced task plan set is as follows: Commercial order data, company system configuration data, and process and working hour data are broken down. After layered processing, orders are allocated to different simulation task planning groups based on the working hour requirements, daily production capacity, and delivery urgency of the orders, resulting in a simulation task planning group set.

3. The task plan generation method based on a large model according to claim 1, characterized in that, The specific process of obtaining the task plan by constructing the task execution frame for each simulation task plan group is as follows: Based on process dependencies, tasks within each task group are broken down into process-level nodes. These process-level nodes are then connected according to enterprise equipment and shift constraints to obtain the process sequence, which serves as the process dependency graph. Each node in the process dependency graph is bound to: a fixed number of collaborators, a specified production line or machine model, and the required man-hours for the task. Priority values ​​are assigned to the corresponding nodes based on task priority, delivery urgency, and man-hour duration to obtain the comprehensive priority value for each node. Through the above process, the task execution frame is obtained. Based on the obtained task execution frame, the task plan is generated by using scheduling rules to produce the order.

4. The task plan generation method based on a large model according to claim 2, characterized in that, The process of breaking down commercial order data, company policy configuration data, and process and working hour data: The business order data is broken down as follows: Based on the segmented interval discrimination algorithm, the number of orders is compared with the preset interval threshold, and the batch interval to which they belong is output; according to the process category mapping relationship in the database, the hash mapping retrieval algorithm is used to complete the correspondence between type and process category, and the corresponding process complexity coefficient is output according to the existing process category complexity matrix. For the delivery cycle, the delivery urgency factor is obtained by using the minimum and maximum normalization methods; The company's policy configuration data is broken down, and the number of collaborative workers is obtained from the fixed number of employees table of the policy based on product type and quantity range; If the order involves multiple process steps, find out the fixed number of people required for each step. The time requirement is obtained by breaking down and calculating the process and time data: the total time requirement for the order is output by multiplying the time per piece by the quantity and the process complexity. The total working hours of the equipment order are obtained by adjusting the equipment utilization rate. Based on the results of hierarchical processing of the acquired raw task data, the production capacity and daily demand are calculated to obtain the production capacity per shift, daily production capacity, and the number of days required for an order.

5. The task plan generation method based on a large model according to claim 2, characterized in that, The specific process of obtaining the simulation task plan set: Extract the total working hours, daily processing hours of the production line, and the urgency factor of the order for each order; sort the orders from high to low urgency using a sorting algorithm based on urgency; use an iterative allocation algorithm to match the total working hours of the orders with the daily capacity of the production line: if the total working hours of the orders are less than the remaining daily capacity of a certain production unit, then all orders are allocated to that task planning group. If the total working hours of an order exceed the available daily production capacity, the order will be split into several sub-tasks and assigned to different task planning groups. After allocation and calculation, a set of simulation task plans is formed.

6. The task plan generation method based on a large model according to claim 1, characterized in that, The specific process of obtaining the task execution order within the task plan to get the total task processing time and overall task performance value is as follows: Each task is broken down into subtasks according to the task execution order to obtain the task quantity of each subtask; based on the obtained task quantity, tasks are allocated to obtain the employee's specific ability value and overall ability value; the task quantity is matched with the business type within the task and the employee's specific ability value and overall ability value; based on the processing time and task excellence value of each task; the processing time and task excellence value of all tasks in the task plan are statistically analyzed to obtain the total task processing time and total task excellence value.

7. The task plan generation method based on a large model according to claim 6, characterized in that, The specific process for obtaining the processing time and task performance value for each task is as follows: The attribute information stream of the obtained sub-task nodes is input into the large model inference engine, and the large model performs inference calculations based on historical task samples in the enterprise database to output the task quantity value of the sub-task nodes. Historical task processing data is retrieved from the enterprise database. The retrieved metrics are combined with the employee's personal resume and input into the large model inference engine to output the specific ability value and overall ability value of each employee. Then, the task volume is mapped to the employee's ability value, and the predicted efficiency of the employee in completing the sub-task is calculated. Finally, with the goal of minimizing total working hours and balancing employee load, an optimization model is established: each subtask node must be assigned to at least one employee who meets the capacity threshold; the total allocated working hours of an employee must not exceed the available working hours after their utilization rate is corrected. The task volume is matched and allocated by combining the task's business type with the employee's specific ability value and overall ability value: the task processing time allocated to the employee is obtained based on the employee's matching efficiency; and the task priority is combined with the processing time to obtain the task's priority value.

8. The task plan generation method based on a large model according to claim 4, characterized in that, The calculation process for the process complexity coefficient is as follows: The constituent factors of the process complexity coefficient are defined as multiple sub-factors unique to IT projects, including: code complexity factor, requirement frequency factor, integration complexity factor, dependency maturity factor, migration risk factor, security compliance factor, automated testing difficulty factor, deployment and maintenance complexity factor, and performance scalability requirement factor; the sensitivity of each sub-factor is preset according to the different impacts of the sub-factors on complexity; the credibility of the values ​​of each sub-factor is obtained based on historical data and used as a data confidence parameter. Each sub-factor sensitivity parameter is subjected to an exponential transformation to obtain an uncorrected exponential factor. Based on the uncorrected exponential factor, the data confidence parameter is combined to obtain the corrected exponential factor. For a certain process, all the corrected exponential factors are multiplied and synthesized to obtain the comprehensive complexity amplification factor of the process. After obtaining the overall complexity amplification factor, and combining it with the standard time T of this process... i With average working hours T avg The basic complexity coefficient of the process is obtained; the basic complexity coefficient is corrected according to the configuration of the number of people working together, and the corrected complexity coefficient is output; the utilization complexity coefficient is obtained by combining the equipment utilization rate; and the utilization complexity coefficients of all processes are obtained by hierarchical superposition to obtain the process complexity coefficient.

9. The task plan generation method based on a large model according to claim 4, characterized in that, The specific process for obtaining the total working hours of the order is as follows: Extract the total available time and actual running time from the running log, and use a ratio calculation model to divide the actual running time by the total available time to obtain the initial utilization rate value; The initial utilization rate is corrected by: obtaining the recorded downtime maintenance time, classifying it, distinguishing between planned maintenance and unplanned downtime, using a logical discrimination algorithm to deduct and correct the unplanned downtime, correcting the initial utilization rate, and obtaining the corrected utilization rate value; then, the minimum and maximum normalization methods are used on the corrected utilization rate result to obtain the final utilization rate.

10. A task plan generation system based on a large model, characterized in that... The method for generating a task plan based on a large model as described in any one of claims 1-9 includes: a data preprocessing module, a time assessment module, a task modeling module, a large model inference module, and a performance analysis module; The data preprocessing module is used to acquire commercial order data, process and working hour data, and company policy configuration data, and to clean, standardize, and decompose the data into layers to output the original task data. The time assessment module is used to receive the original task data and calculate and correct the total time and process complexity coefficient based on the complexity factor, time and equipment utilization rate to form the time requirement result; The task modeling module is used to divide orders into simulated task planning groups based on working hours, daily capacity and delivery urgency, and to build a process dependency graph to generate a task execution framework; The large model inference module is used to decompose the task execution framework, combine historical task samples, and generate task nodes by inferring the task quantity values ​​of sub-tasks from the large model. The performance analysis module combines employees' specific ability values ​​and overall ability values ​​to allocate task node sets, calculate task processing time and task excellence value, statistically obtain candidate task plans, and filter and output the optimal task plan.

Images