Expert extraction scheme compilation method and system based on operation planning optimization

Abstract

The invention provides a compilation method and system of an expert extraction scheme based on operation planning optimization, and relates to the technical field of material professional operation planning optimization, and the method comprises the steps: carrying out the preprocessing of collected expert extraction multi-source data, and obtaining a standardized data set; based on the standardized data set, configuring parameters of the multi-target operation optimization model; solving the multi-target operation optimization model by using an optimization algorithm, generating an optimal solution data set, and generating an initial expert extraction scheme based on the optimal solution data set; and performing compliance verification on the initial expert extraction scheme, receiving a manual fine adjustment instruction based on a visual interface, automatically verifying the adjusted compliance and evaluation indexes by the system, and compiling a final expert extraction scheme. According to the invention, through multi-system automatic docking and model automatic solving, the compilation efficiency of an expert extraction scheme is improved; global optimization is realized based on a multi-target operation optimization model, the response period of demand change is shortened, and a high-quality expert extraction scheme is compiled.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of material professional operational optimization, and particularly relates to a method and system for compiling an expert extraction scheme based on operational optimization. BACKGROUND

[0002] With the rapid development of the economic society, the compliance requirements in the process of bidding expert extraction of large enterprises are becoming more and more strict. Not only should experts with professional technical ability and matching bidding work be extracted, but also the demand of bidding workload should be met to ensure the timeliness and accuracy of bidding. Therefore, the enterprises have established a corresponding expert database management system to classify experts according to their professional background, industry experience and past performance, so as to ensure that experts with corresponding qualifications can be matched to different levels of projects. At the same time, in order to effectively apply the expert database in actual operation, random extraction is realized through information technology means, and the situation of the bidding supplier and the potential conflict of interest are fully considered to ensure the fairness and professionalism of the bidding process. The existing expert database currently covers seven expert categories of bidding, qualification verification, on-site supervision, legal protection, authorized procurement, compliance inspection and document review, and reserves tens of thousands of experts. However, in actual operation, the compilation of the expert extraction scheme still faces three core pain points: first, the current bidding expert extraction scheme compilation still relies on manual operation, which not only consumes time and effort, but also is prone to imperfections or errors due to human factors; second, the expert extraction scheme considers complex factors, and if the scheme is not reasonably compiled, there are problems such as insufficient expert resources, audit risks and poor bidding results; third, enterprises have higher requirements for the transparency and fairness of public resource transactions, and the market also has an urgent need for more efficient and convenient bidding processes.

[0003] Therefore, in the existing compilation of the expert extraction scheme, the manual compilation method is low in efficiency, highly dependent on manpower and prone to operation errors; the extraction method based on the computer considers single factors, lacks multi-dimensional data fusion and global optimization capabilities, and has insufficient compliance risk control, so that manual conflict checking is prone to omissions and it is difficult to obtain a scientific and fair expert extraction scheme. SUMMARY

[0004] In order to overcome the deficiencies of the prior art, the present application provides a method for compiling an expert extraction scheme based on operational optimization, which comprises: The collected expert extraction multi-source data is preprocessed to obtain a standardized data set; the expert extraction multi-source data at least includes package demand data and expert data; Based on the standardized data set, parameters of a multi-objective operational optimization model are configured, the parameters of the multi-objective operational optimization model include weights of a comprehensive objective function and threshold values of constraint conditions; Solving the multi-objective operation optimization model by using an optimization algorithm to generate an optimal solution dataset, and generating an initial scheme extracted by experts based on the optimal solution dataset; Performing compliance verification on the initial scheme extracted by experts, performing effect evaluation on the initial scheme extracted by experts passing the compliance verification, and receiving manual fine-tuning instructions based on a visual interface, automatically verifying the adjusted compliance and evaluation indexes, and compiling a final expert extraction scheme.

[0005] Preferably, the optimization algorithm is an improved genetic algorithm. Solving the multi-objective operation optimization model by using an optimization algorithm to generate an optimal solution dataset, including: Using a binary and real number hybrid coding method to encode candidate solutions of the multi-objective operation optimization model to generate chromosomes; Based on the coding method, an initial population containing multiple chromosomes is initialized and generated as a current population; Taking a comprehensive objective function of the multi-objective operation optimization model as a fitness function, the fitness value of each chromosome in the current population is calculated; According to the fitness value, a selection operation is performed on the current population, and a cross operation using an adaptive cross probability and a mutation operation using a local mutation strategy are sequentially performed on the selected chromosomes to generate a new offspring population; The offspring population is updated as the current population, and it is determined whether the current population meets a convergence condition; If not, return to the step of calculating the fitness of the current population and continue iteration; If yes, the solution corresponding to the chromosome with the highest fitness in the current population is determined as the optimal solution dataset.

[0006] Preferably, the optimization algorithm is a particle swarm optimization algorithm. Solving the multi-objective operation optimization model by using an optimization algorithm to generate an optimal solution dataset, including: Initializing a particle swarm, and a position vector of each particle is used to represent a solution of the multi-objective operation optimization model, and a velocity vector represents a parameter adjustment direction; Taking a comprehensive objective function of the multi-objective operation optimization model as a fitness function, the fitness value of each particle in the current particle swarm is calculated; According to the fitness value, the individual historical optimal solution of each particle and the global historical optimal solution of the entire particle swarm are updated; According to the individual historical optimal solution and the global historical optimal solution, the velocity vector and the position vector of each particle are updated; The updated particle swarm is taken as the current particle swarm, and it is determined whether a preset convergence condition is met; If not, return to perform the step of calculating the current particle swarm fitness value, and continue iteration. If yes, determine the solution corresponding to the particle position vector with the highest fitness value in the current particle swarm as the optimal solution dataset.

[0007] Preferably, the multi-source data comes from at least one of an expert database system, a bidding and tendering system, a bid evaluation and bid evaluation base management system, and a supplier association system. The multi-source data includes at least one of bid package demand, expert information, base information, and bid unit association information. The collection of multi-source data is achieved through real-time docking via an API interface or batch file import.

[0008] Preferably, the comprehensive objective function of the multi-objective operational optimization model includes a multi-objective function of maximizing an evaluation matching degree function, minimizing a resource waste function, and minimizing an evaluation risk function. The constraint conditions include at least one of the following: professional matching constraint, expert level constraint, base capacity constraint, expert load constraint, compliance avoidance constraint, bid evaluation group number constraint, first-level expert mandatory constraint, and bid package-bid evaluation group unique correspondence constraint. The calculation formula of the comprehensive objective function is as follows:

[0009]

[0010]

[0011]

[0012] wherein, is the comprehensive objective function, is the maximum evaluation matching degree function, is the minimum resource waste function, is the minimum evaluation risk function, is the evaluation matching weight, is the resource waste weight, is the evaluation risk weight. is the professional matching degree of the expert and the bid package, , , is the importance weight of the bid package , , is the level coefficient of the expert , is whether the expert participates in the bid evaluation of the bid package , indicates participation, indicates non-participation, is the total number of experts, is the total number of bid packages; is the actual number of groups of occupancy of the base , is the maximum capacity of the base , is the total number of bases; is the risk value of the expert , , is the total number of bid packages in which the expert participated this time.

[0013] Preferably, the initial expert extraction scheme is subjected to compliance verification, and the initial expert extraction scheme that passes the compliance verification is subjected to effectiveness evaluation, including: Based on the compliance rule library, the initial expert extraction scheme is subjected to rule verification, and the initial expert extraction scheme is subjected to data consistency verification with the standardized data set, and a comprehensive verification result is generated; If the comprehensive verification result indicates that there is a violation item, then according to the type of the violation item, the corresponding abnormal processing process is triggered; If the comprehensive verification result indicates that the initial expert extraction scheme passes the verification, then the initial expert extraction scheme that passes the compliance verification is subjected to effectiveness evaluation; Wherein, the effectiveness evaluation includes: Calculate a plurality of evaluation indexes of the initial expert extraction scheme that passes the verification, and the evaluation indexes include one or more of a matching degree index, a resource utilization rate index, and a risk index; Based on the evaluation indexes, an evaluation report is generated to determine whether each index meets a preset threshold.

[0014] Preferably, based on the visual interface, a manual fine-tuning instruction is received, and the system automatically verifies the adjusted compliance and evaluation indexes, and compiles a final expert extraction scheme, including: Through the visual interface, a manual fine-tuning instruction for the initial expert extraction scheme that passes the compliance verification is received, and the manual fine-tuning instruction includes one or more of expert replacement, group adjustment, and time adjustment; In response to the manual fine-tuning instruction, a corresponding adjustment operation is performed to generate an adjusted expert extraction scheme; The adjusted expert extraction scheme is subjected to compliance verification and evaluation index recalculation; When the verification and recalculation results confirm that the adjusted expert extraction scheme is compliant and the evaluation indexes meet the standards, the adjusted expert extraction scheme is determined as the final expert extraction scheme.

[0015] Preferably, based on the visual interface receiving manual fine-tuning instructions, the system automatically checks the adjusted compliance and evaluation indicators, and after compiling the final expert extraction scheme, the method further comprises: outputting the final expert extraction scheme in one of Excel, PDF or XML preset formats; synchronously generating a whole-process log, which is used to record operation records and parameter information in the whole process from data collection to scheme confirmation; packaging and archiving the final expert extraction scheme, the effect evaluation report and the whole-process log to an enterprise document management system.

[0016] Based on the same inventive concept, the present application also provides an expert extraction scheme compiling system based on operational optimization, which comprises: a data set obtaining module, configured to pre-process collected expert extraction multi-source data to obtain a standardized data set; the expert extraction multi-source data at least includes package demand data and expert data; a model parameter configuration module, configured to configure parameters of a multi-objective operational optimization model based on the standardized data set, wherein the parameters of the multi-objective operational optimization model include weights of a comprehensive objective function and threshold values of constraint conditions; an initial scheme generating module, configured to solve the multi-objective operational optimization model by using an optimization algorithm to generate an optimal solution data set, and generate an expert extraction initial scheme based on the optimal solution data set; a final scheme compiling module, configured to perform compliance checking on the expert extraction initial scheme, perform effect evaluation on the expert extraction initial scheme passing the compliance checking, and based on a visual interface receiving manual fine-tuning instructions, the system automatically checks the adjusted compliance and evaluation indicators to compile a final expert extraction scheme.

[0017] Preferably, the optimization algorithm is an improved genetic algorithm; The initial scheme generating module is specifically configured to: encode candidate solutions of the multi-objective operational optimization model by using a binary and real number mixed coding mode to generate chromosomes; initialize to generate an initial population containing a plurality of the chromosomes as a current population based on the coding mode; take the comprehensive objective function of the multi-objective operational optimization model as a fitness function to calculate fitness values of each chromosome in the current population; perform selection operation on the current population according to the fitness values, and in turn perform crossover operation by using an adaptive crossover probability and mutation operation by using a local mutation strategy on the selected chromosomes to generate a new offspring population; update the offspring population to the current population, and judge whether the current population meets a convergence condition; If not, return to the step of calculating the fitness of the current population and continue iteration; If yes, determine the solution corresponding to the chromosome with the highest fitness in the current population as the optimal solution dataset.

[0018] Preferably, the optimization algorithm is a particle swarm optimization algorithm; The initial scheme generation module is specifically configured to: Initialize a particle swarm, and a position vector of each particle is used to represent a solution of the multi-objective operational optimization model, and a velocity vector represents a parameter adjustment direction; Take the comprehensive objective function of the multi-objective operational optimization model as a fitness function, and calculate the fitness value of each particle in the current particle swarm; According to the fitness value, update the individual historical optimal solution of each particle and the global historical optimal solution of the entire particle swarm; According to the individual historical optimal solution and the global historical optimal solution, update the velocity vector and the position vector of each particle; Take the updated particle swarm as the current particle swarm, and judge whether a preset convergence condition is met; If not, return to the step of calculating the fitness of the current particle swarm and continue iteration; If yes, determine the solution corresponding to the position vector of the particle with the highest fitness in the current particle swarm as the optimal solution dataset.

[0019] Preferably, the multi-source data comes from at least one of an expert library system, a bidding and tendering system, a bid evaluation and bid evaluation base management system, and a supplier association system; The multi-source data includes at least one of bid package demand, expert information, base information, and bid unit association information; The collection of the multi-source data is realized through real-time docking via an API interface or batch file import.

[0020] Preferably, the comprehensive objective function of the multi-objective operational optimization model includes multi-objective functions of maximizing an evaluation matching degree function, minimizing a resource waste function, and minimizing an evaluation risk function; The constraint conditions include at least one of the following: professional matching constraint, expert level constraint, base capacity constraint, expert load constraint, compliance avoidance constraint, bid evaluation group number constraint, first-level expert mandatory constraint, and bid package-bid evaluation group unique correspondence constraint; The calculation formula of the comprehensive objective function is as follows:

[0021]

[0022]

[0023]

[0024] wherein, is a comprehensive objective function, is a maximum bid matching degree function, is a minimum resource waste function, is a minimum bid risk function, is a bid matching weight, is a resource waste weight, is a bid risk weight; is an expert and a professional matching degree with a bid package , , is an importance weight of a bid package , , is a grade coefficient of an expert , is whether an expert participates in the bid evaluation of a bid package , indicates participation, indicates non-participation, is the total number of experts, is the total number of bid packages; is the actual number of occupied groups of a base , is the maximum capacity of a base , is the total number of bases; is a risk value of an expert , , is the total number of bid packages that an expert participates in this time.

[0025] Preferably, the final scheme compiling module is specifically used for: performing rule verification on the initial scheme extracted by the expert based on a compliance rule library, and performing data consistency verification on the initial scheme extracted by the expert and a standardized data set, to generate a comprehensive verification result; if the comprehensive verification result indicates that there is a rule violation item, triggering a corresponding exception handling process according to the type of the rule violation item; if the comprehensive verification result indicates that the initial scheme extracted by the expert passes the verification, performing effect evaluation on the initial scheme extracted by the expert that passes the compliance verification; wherein, the effect evaluation includes: calculating a plurality of evaluation indexes of the initial scheme extracted by the expert that passes the verification, the evaluation indexes including one or more of a matching degree index, a resource utilization rate index and a risk index; Generate an evaluation report based on the evaluation indicators, and determine whether each indicator meets the preset threshold.

[0026] Preferably, the final scheme compiling module is specifically used for: Through the visual interface, the user's manual fine-tuning instruction for extracting an initial scheme by the expert through compliance verification is received, and the manual fine-tuning instruction includes one or more of expert replacement, grouping adjustment, and time adjustment; In response to the manual fine-tuning instruction, a corresponding adjustment operation is performed to generate an adjusted expert extraction scheme; The adjusted expert extraction scheme is subjected to compliance verification and evaluation indicator recalculation; When the verification and recalculation results confirm that the adjusted expert extraction scheme is compliant and the evaluation indicators meet the standards, the adjusted expert extraction scheme is determined as the final expert extraction scheme.

[0027] Preferably, the system further includes a final scheme output module for: Outputting the final expert extraction scheme in a preset format of Excel, PDF, or XML; Synchronously generating a full-process log, which is used to record operation records and parameter information in the full process from data collection to scheme confirmation; Packing the final expert extraction scheme, the effect evaluation report, and the full-process log into an enterprise document management system.

[0028] Based on the same inventive concept, the present application also provides an electronic device, comprising: at least one processor and a memory; the memory and the processor are connected through a bus; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, a method for compiling an expert extraction scheme based on operational optimization is realized.

[0029] Based on the same inventive concept, the present application also provides a readable storage medium, which has an execution program stored thereon, and the execution program, when executed, realizes a method for compiling an expert extraction scheme based on operational optimization.

[0030] Compared with the closest prior art, the present application has the following beneficial effects: The application provides a method for compiling an expert extraction scheme based on operational optimization, comprising: preprocessing collected expert extraction multi-source data to obtain a standardized data set; the expert extraction multi-source data at least includes label package demand data and expert data; based on the standardized data set, parameters of a multi-objective operational optimization model are configured, the parameters of the multi-objective operational optimization model include weights of a comprehensive objective function and threshold values of constraint conditions; an optimization algorithm is used to solve the multi-objective operational optimization model to generate an optimal solution data set, and based on the optimal solution data set, an initial expert extraction scheme is generated; the initial expert extraction scheme is subjected to compliance verification, the initial expert extraction scheme passing the compliance verification is subjected to effect evaluation, and based on a visual interface, a manual fine-tuning instruction is received, the system automatically verifies the adjusted compliance and evaluation indexes, and a final expert extraction scheme is compiled. Through automatic connection of multiple systems and automatic solution of the model, the compiling efficiency of the expert extraction scheme is improved; global optimization is realized based on the multi-objective operational optimization model, the response cycle of demand changes is greatly shortened, and a high-quality expert extraction scheme is compiled. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1 A flowchart of the method for compiling an expert extraction scheme based on operational optimization provided by the application is shown in the figure; Figure 2 A flowchart of data preprocessing provided by the application is shown in the figure; Figure 3 A structural diagram of a multi-objective operational optimization model provided by the application is shown in the figure; Figure 4 A specific flowchart of the method for compiling an expert extraction scheme based on operational optimization provided by the application is shown in the figure; Figure 5 A data flow conversion diagram of the method for compiling an expert extraction scheme based on operational optimization provided by the application is shown in the figure; Figure 6 A structural diagram of a system for compiling an expert extraction scheme based on operational optimization provided by the application is shown in the figure; Figure 7 A specific architecture diagram of the system for compiling an expert extraction scheme based on operational optimization provided by the application is shown in the figure; Figure 8 A schematic diagram of an electronic device provided by the application is shown in the figure. DETAILED DESCRIPTION

[0032] Embodiments of the application are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by reference to the accompanying drawings are exemplary and are intended to explain the application, and cannot be understood as a limitation of the application.

[0033] With the rapid development of digital and intelligent technology, various industries are actively promoting the digital and intelligent transformation. In the field of bidding, preparing the expert extraction scheme through intelligent means has become an important trend to improve work efficiency, reduce error rate and enhance fairness. Through intelligent algorithms and models, intelligent preparation of expert extraction scheme can consider various factors such as the professional background, experience and regional distribution of experts, so as to generate a more scientific and fair expert extraction scheme. At the same time, intelligent preparation of projects can automatically check the consistency and accuracy of data to avoid human errors. The intelligent preparation of expert extraction scheme effectively promotes the digital and intelligent development of the bidding industry.

[0034] Embodiment 1: The present application provides a kind of based on the preparation method of expert extraction scheme of operation optimization, specifically, Figure 1 The flowchart of the preparation method of expert extraction scheme based on operation optimization provided by the embodiment of the present application is shown in the figure, and includes the following steps: S101: the expert extraction multi-source data collected is preprocessed, and a standardized data set is obtained;Expert extraction multi-source data at least includes label package demand data and expert data; S102: based on the standardized data set, the parameters of the multi-objective operation optimization model are configured, and the parameters of the multi-objective operation optimization model include the weight of the comprehensive objective function and the threshold of the constraint condition; S103: the multi-objective operation optimization model is solved by using optimization algorithm, and the optimal solution data set is generated, and the expert extraction initial scheme is generated based on the optimal solution data set; S104: the expert extraction initial scheme is checked for compliance, the expert extraction initial scheme that passes the compliance check is evaluated, and the manual fine-tuning instruction is received based on the visual interface, the adjusted compliance and evaluation index are automatically checked by the system, and the final expert extraction scheme is prepared.

[0035] The present application improves the preparation efficiency of expert extraction scheme by automatically docking multiple systems and automatically solving models. Based on the multi-objective operation optimization model, global optimization is realized, the response cycle of demand change is greatly shortened, and high-quality expert extraction scheme is prepared.

[0036] The present application first collects expert extraction multi-source data, and pre-processes the expert extraction multi-source data to obtain a standardized data set. The expert extraction multi-source data at least includes label package demand data and expert data.

[0037] The multi-source data comes from at least one of the following: expert database system, bidding system, bid evaluation base management system, and supplier association system; the multi-source data includes at least one of the following: bid package requirements, expert information, base information, and bidding unit association information; the collection of multi-source data is achieved through real-time API interface connection or batch file import.

[0038] In one specific implementation, the user selects the project ID (Identifier) ​​of the plan to be prepared on the system interface. The system automatically triggers the "multi-source data collection" task, which connects to the expert database system, bidding system, bid evaluation base management system, and supplier association system through the API (Application Programming Interface) to collect the project's bid package requirements, expert information, base information, and bidding unit association information.

[0039] Among them, 1) the expert database system collects information such as expert ID, professional category, level, available time, recent bidding records, list of associated suppliers, and region; 2) Bidding and tendering system: Connects to the bidding and tendering management system to collect information such as project ID, number of bid packages, professional requirements for each bid package, project level, and list of bidding units; 3) Bidding Base Management System: Collects information such as base ID, location, maximum number of bidding groups that can be accommodated, number of groups currently occupied, and supporting facilities; 4) Supplier Association System: Connects with enterprise CRM (Customer Relationship Management) system and business registration information system to collect the list of shareholders, senior executives and related companies of bidding units, which is used to expand the avoidance rules.

[0040] Data preprocessing includes: cleaning multi-source data to obtain cleaned multi-source data, which includes at least one of missing value imputation, outlier marking, and duplicate value removal; and standardizing the cleaned multi-source data to obtain standardized multi-source data, which includes at least one of unified data encoding, unified format encoding, and data conflict detection.

[0041] like Figure 2 The diagram shown illustrates the data preprocessing flow provided by this invention. In one specific embodiment, the data preprocessing process includes: 1) Data cleaning: The system automatically handles missing values ​​in the raw data, such as filling in default values ​​when the available time of experts is missing; outliers, such as marking expert levels that are out of range as pending verification; duplicate values, deleting duplicate expert records; and generates a cleaning report, which includes the amount of data processed and the processing method.

[0042] 2) Data standardization: the system unifies data coding and format according to preset standards, such as expert professional coding, bid package professional coding, and time format, to ensure data consistency.

[0043] 3) Data conflict detection: the system detects expert available time conflicts, base capacity conflicts, and association relationship conflicts, marks high-risk data such as saturated bases and associated bid units, and feeds back the detection results to the user.

[0044] After obtaining the standardized data set, parameters of the multi-objective operational optimization model are configured based on the standardized data set, including weights of the comprehensive objective function and threshold values of the constraint conditions.

[0045] Specifically, the comprehensive objective function of the multi-objective operational optimization model includes a multi-objective function of maximizing the bid evaluation matching degree function, minimizing the resource waste function, and minimizing the bid evaluation risk function. The constraint conditions include at least one of the following: professional matching constraint, expert level constraint, base capacity constraint, expert load constraint, compliance avoidance constraint, bid evaluation group number constraint, first-level expert mandatory constraint, and bid package-bid evaluation group unique correspondence constraint.

[0046] The calculation formula of the comprehensive objective function is as follows:

[0047]

[0048]

[0049]

[0050] wherein, is the comprehensive objective function, is the maximum bid evaluation matching degree function, is the minimum resource waste function, is the minimum bid evaluation risk function, is the bid evaluation matching weight, is the resource waste weight, is the bid evaluation risk weight; is the professional matching degree of the expert and the bid package, , , is the importance weight of the bid package , , is the level coefficient of the expert , is whether the expert participates in the bid evaluation of the bid package , represents participation, represents non-participation, is the total number of experts, is the total number of bid packages; is the actual number of occupied groups of the base , is the maximum capacity of the base , is the total number of bases; is the risk value of the expert , , is the total number of bid packages in which the expert participates this time.

[0051] The constraint conditions include: Constraint 1, professional matching constraint:

[0052] wherein, is the professional matching degree of the expert and the bid package , is the minimum matching degree threshold; Constraint 2, expert level constraint:

[0053] wherein, is the level coefficient of the expert , is the lowest expert level coefficient required by the bid package ; Constraint 3, base capacity constraint:

[0054] wherein, is the actual number of occupied groups of the base , is the maximum capacity of the base ; Constraint 4, expert load constraint:

[0055] wherein, is the number of times the expert has participated in bid evaluation in the current period, is whether the expert participates in the bid evaluation of the bid package , is the upper limit of the number of times of bid evaluation; Constraint 5, compliance avoidance constraint:

[0056] where, is the set of experts associated with the bidding unit, i.e. the experts In the high-risk set, it is mandatory to not participate in the evaluation of all the packages ; Constraint 6, number of panelists constraint:

[0057] where, is the assignment of the evaluation task of the package to the panel , denotes assignment, denotes non-assignment, is the participation of the expert in the evaluation of the package , is the minimum number of panelists, is the maximum number of panelists; Constraint 7, first-level expert mandatory constraint:

[0058] where, is the assignment of the evaluation task of the package to the panel , denotes assignment, denotes non-assignment, is the participation of the expert in the evaluation of the package , denotes if the expert is a first-level expert, denotes first-level expert, denotes non-first-level expert; Constraint 8, package-panel unique correspondence constraint:

[0059] denotes that the evaluation task of each package is assigned to only one panel, is the assignment of the evaluation task of the package to the panel , denotes assignment, denotes non-assignment, is the total number of panels.

[0060] As Figure 3As shown, the structure diagram of the multi-objective operational optimization model provided by the application is shown. Through user configuration parameters and constraint conditions, a comprehensive objective function is obtained, and the improved genetic algorithm is further used for solving, and an optimal solution dataset is output. In a specific embodiment, the parameter configuration of the multi-objective operational optimization model includes: 1) target function weight configuration: the user configures the weight of the comprehensive objective function through a visual interface, and the system provides "default configuration" and "custom configuration"; 2) constraint condition configuration: the user configures the threshold of the constraint condition, including professional minimum matching degree T, upper limit of expert monthly bid evaluation times, and expert number range of bid evaluation group; 3) parameter verification: the system automatically verifies whether the configured parameters are reasonable, and prompts the user to adjust if not.

[0061] After the parameters of the multi-objective operational optimization model are configured, the optimization algorithm is further used to solve the multi-objective operational optimization model to generate an optimal solution dataset, and an initial scheme of expert extraction is generated based on the optimal solution dataset.

[0062] In this embodiment, the optimization algorithm can be an improved genetic algorithm or a particle swarm optimization algorithm.

[0063] When the optimization algorithm is an improved genetic algorithm, the optimization algorithm is used to solve the multi-objective operational optimization model to generate an optimal solution dataset, including: A binary and real number hybrid coding method is used to code the candidate solution of the multi-objective operational optimization model to generate a chromosome; Based on the coding method, an initial population containing a plurality of the chromosomes is initialized and generated as a current population; The comprehensive objective function of the multi-objective operational optimization model is used as a fitness function to calculate the fitness value of each chromosome in the current population; The selection operation is performed on the current population according to the fitness value, and the crossover operation with adaptive crossover probability and the mutation operation with local mutation strategy are sequentially performed on the selected chromosomes to generate a new offspring population; The offspring population is updated as the current population, and it is judged whether the current population meets the convergence condition; If not, return to the step of calculating the fitness of the current population and continue iteration; If yes, the solution corresponding to the chromosome with the highest fitness in the current population is determined as the optimal solution dataset.

[0064] When the optimization algorithm is a particle swarm optimization algorithm, the optimization algorithm is used to solve the multi-objective operational optimization model to generate an optimal solution dataset, including: Initialize the particle swarm, and the position vector of each particle is used to represent a solution of the multi-objective operational optimization model, and the velocity vector represents the parameter adjustment direction; Taking a comprehensive objective function of the multi-objective operational optimization model as the fitness function, the fitness value of each particle in the current particle swarm is calculated; According to the fitness value, the individual historical optimal solution of each particle and the global historical optimal solution of the entire particle swarm are updated; According to the individual historical optimal solution and the global historical optimal solution, the speed vector and the position vector of each particle are updated; The updated particle swarm is taken as the current particle swarm, and it is judged whether the preset convergence condition is met; If not, the step of calculating the fitness value of the current particle swarm is returned to continue iteration; If yes, the solution corresponding to the position vector of the particle with the highest fitness value in the current particle swarm is determined as the optimal solution dataset.

[0065] In a specific embodiment, model solving and initial scheme generation include: 1) model parameter input: the system inputs the "standardized dataset" and the user-configured parameters into the operational optimization model; 2) improved genetic algorithm solving: the system starts the improved genetic algorithm, solves according to the process of "coding→initialization of population→calculation of fitness→crossing→mutation→iteration", and displays the fitness value change curve in real time during the iteration process for the convenience of user monitoring; 3) generation of optimal solution dataset: when the algorithm meets the convergence condition, the iteration is stopped after 100 times or the fitness value change is less than or equal to 0.001, and the "optimal solution dataset" is generated, which contains expert-tag association, grouping, and base allocation information; 4) initial scheme generation: the system generates the initial scheme based on the "optimal solution dataset" according to the enterprise standard template.

[0066] After obtaining the initial scheme of expert extraction, the initial scheme is further optimized. That is, the initial scheme of expert extraction is subjected to compliance verification, the initial scheme of expert extraction passing the compliance verification is subjected to effect evaluation, and based on the visual interface, a manual fine-tuning instruction is received, and the system automatically verifies the adjusted compliance and evaluation indexes to compile a final expert extraction scheme.

[0067] Specifically, the initial scheme of expert extraction is subjected to compliance verification, and the initial scheme of expert extraction passing the compliance verification is subjected to effect evaluation, including: performing rule verification on the initial scheme of expert extraction based on a compliance rule library, and performing data consistency verification on the initial scheme of expert extraction and the standardized dataset to generate a comprehensive verification result; if the comprehensive verification result indicates that there are irregular items, triggering the corresponding exception handling process according to the type of the irregular items; if the comprehensive verification result indicates that the initial scheme of expert extraction passes the verification, the initial scheme of expert extraction passing the compliance verification is subjected to effect evaluation; The effect evaluation includes: The multiple evaluation indexes of the expert extraction initial scheme passing the compliance verification are calculated, and the evaluation indexes include one or more of a matching degree index, a resource utilization rate index, and a risk index. An evaluation report is generated based on the evaluation indexes, and it is determined whether each index reaches a preset threshold.

[0068] In a specific embodiment, the process of compliance verification includes: 1) rule verification: the system calls the "compliance rule library" to automatically detect whether the initial scheme conforms to 12 types of compliance rules, such as expert level matching and avoidance rules. If there are rule violation items, mark the violation content and type. 2) data consistency verification: the system compares the initial scheme data with the "standardized data set" to ensure that there is no data tampering or inconsistency, such as whether the available time of the expert is consistent with the scheme time. 3) rule violation processing: if it is a minor violation, such as exceeding the budget of travel costs, the system prompts the user "whether to accept the scheme"; if it is a major violation, such as violating the avoidance rule, the constraint parameters are adjusted and then the solution is recalculated.

[0069] In some optional embodiments, a machine learning classification model can also be used for compliance verification. A random forest classification model is trained through historical compliance / violation scheme data. The model input is the feature parameters of the scheme, and the output is the compliance probability. When the compliance probability is greater than or equal to 0.95, it is determined to be compliant, otherwise it is determined to be in violation. This model is suitable for enterprises with complex and frequently changing compliance rules (such as the financial industry, which needs to add "funds-related avoidance" and "regulatory-related avoidance" rules). The model can be iteratively trained through new data to automatically adapt to rule changes.

[0070] The process of effect evaluation includes: 1) effect evaluation: the system calculates the matching degree index, the resource utilization rate index, and the risk index of the scheme, generates an "evaluation report", and marks whether the indexes meet the standards. 2) evaluation result judgment: if all indexes meet the standards, proceed to the next step; if the indexes do not meet the standards (such as resource utilization rate < 70%), the system prompts the user "whether to adjust the model parameters", and the user confirms to reconfigure the parameters.

[0071] Further, based on the visual interface, a manual fine-tuning instruction is received, and the system automatically verifies the adjusted compliance and evaluation indexes to compile a final expert extraction scheme. Specifically, through the visual interface, a manual fine-tuning instruction for the expert extraction initial scheme passing the compliance verification is received, and the manual fine-tuning instruction includes one or more of expert replacement, grouping adjustment, and time adjustment. In response to the manual fine-tuning instruction, an adjusted expert extraction scheme is generated by performing the corresponding adjustment operation. The adjusted expert extraction scheme is subjected to compliance verification and evaluation index recalculation. When the verification and recalculation results confirm that the adjusted expert extraction scheme is compliant and the evaluation indexes meet the standards, the adjusted expert extraction scheme is determined as the final expert extraction scheme.

[0072] In one specific embodiment, the process of manual fine-tuning and final confirmation includes: 1) manual fine-tuning: the user performs expert replacement, grouping adjustment, and time adjustment through the visual interface, and the system automatically checks the compliance and index changes after adjustment to ensure that the adjusted scheme is still compliant. 2) Final scheme confirmation: after the user confirms that the scheme is correct, the "final expert extraction scheme" is generated.

[0073] After the final expert extraction scheme is prepared, the method further includes scheme output and archiving. Specifically, it includes outputting the final expert extraction scheme in one of the preset formats of Excel, PDF or XML; synchronously generating a full-process log, which is used to record operation records and parameter information in the whole process from data collection to scheme confirmation; and packaging the final expert extraction scheme, the effect evaluation report and the full-process log into the enterprise document management system.

[0074] In one specific embodiment, 1) scheme output: the system supports user selection of Excel, PDF, XML format output of the final scheme, while providing online preview function; 2) log generation: the system automatically generates "full-process log" to record all operations and parameters from data collection to scheme confirmation, ensuring audit traceability; 3) archiving: the system archives the final scheme, evaluation report and full-process log to the enterprise document management system, and synchronously to the bidding system, completing the scheme preparation.

[0075] Figure 4 The specific flowchart of the expert extraction scheme preparation method based on operational optimization provided by the present application is provided. First, data collection is performed through the expert database system, the bidding system, the base management system and the supplier contact system and other interface systems; second, the solution is obtained through the operational optimization model, then the scheme is generated and verified, and then the scheme is evaluated and adjusted, and finally the prepared final expert extraction scheme is output.

[0076] Figure 5 The data flow diagram of the expert extraction scheme preparation method based on operational optimization provided by the present application is provided. 1, data collection and preprocessing; 2, model parameter configuration; 3, model solution, obtain initial scheme; 4, full-process compliance verification; 5, scheme evaluation and manual fine-tuning; 6, scheme output and archiving.

[0077] Existing technical solutions: (1) Traditional manual scheme preparation At present, most enterprises, especially small and medium-sized projects, still use the traditional manual scheme preparation as the basic mode, which relies on business personnel to manually complete scheme preparation. The specific process is as follows: 1) Demand collection: business personnel obtain project demand documents from the bidding management department, and obtain information such as the number of material bid packages, professional requirements for each bid package, evaluation start time, and proposed evaluation base; 2) Expert screening: log in to the expert database management system, filter experts who meet the conditions according to "professional category" and "expert level", and manually record the names, contact information, and recent available time of experts; 3) Conflict investigation: query the bidding unit information through the supplier management system, manually exclude experts who have "employment association" "ownership association" "project cooperation association" with the bidding unit, such as experts who have worked for the bidding unit in the past three years; 4) Resource balancing: manually calculate the total number of experts needed according to the capacity of the proposed evaluation base, if the number of experts after screening is insufficient, return to the "expert screening" step to expand the scope, such as lowering the expert level requirement; 5) Finalize the plan: organize the selected expert list, evaluation grouping, and base allocation information into an Excel table, submit it to the audit department for compliance audit, and form the final plan after passing the audit.

[0078] (2) Simple random selection plan The random selection plan of evaluation experts is to realize the "random selection" of experts through a computer system, replacing part of the manual operation, and the specific implementation logic is as follows: 1) System architecture: composed of "input module" "random selection module" "conflict exclusion module" "output module", among which "input module" supports business personnel to manually enter bid package professional requirements, number of experts, and blacklisted experts; 2) Core function: "Random selection module" based on "pseudo-random number algorithm", randomly selects a specified number of experts from the expert database according to "professional category"; "Conflict exclusion module" has simple rules such as excluding blacklisted experts and excluding more than two experts from the same unit, automatically filtering experts who do not meet the conditions; 3) Plan output: if the number of randomly selected experts meets the demand, output the expert list and grouping suggestion; if the number is insufficient, the system prompts "expand the professional scope or lower the level requirement", and the business personnel manually adjust and reselect.

[0079] (3) Extraction tool plan based on basic rules By developing an evaluation expert extraction auxiliary tool, "fixed rules" are added based on simple random selection to further improve the compliance of the plan, and the specific implementation is as follows: 1) Data docking: semi-automatically dock with the enterprise's internal expert database system and supplier system, which can automatically obtain the "professional category" "level" "associated supplier list" of experts, without manual input.

[0080] 2) Rule configuration. Built-in 3 types of fixed rules: ① Professional matching rule: calculate the matching degree through "professional keyword coincidence degree", the matching degree of expert specialty and bid package demand ≥80%, that is, it meets the professional matching rule; ② Regional rule: preferentially select experts in the same city as the bid base to reduce travel costs; ③ Avoidance rule: automatically exclude experts who have direct association with the bidding unit.

[0081] 3) Scheme generation: after the business personnel enter the number of bid packages and the number of experts, the system filters experts according to "rule priority" (professional matching > region > avoidance), generates a preliminary scheme, and if there is a situation of insufficient number of experts, the system automatically relaxes the "regional rule", such as expanding to experts in surrounding cities, and reselecting.

[0082] The defects of the above existing technical solutions are: (1) Low efficiency and high dependence on manual operation, prone to operation errors In traditional manual scheme preparation, expert screening, conflict checking, and resource balancing all need to be done manually. It takes 1-2 working days to prepare a large project containing 20 bid packages, which is much lower than the business demand. At the same time, when manually counting the available time of experts and calculating the base capacity, data entry errors are easy to occur. While the simple random selection scheme and the basic rule tool scheme reduce some manual operations, but the business personnel still need to manually enter the bid package demand and adjust the screening range when there is a shortage of experts, and after the scheme is generated, the base resources and expert load information need to be manually checked.

[0083] (2) Single consideration factor, lack of multi-dimensional data fusion and global optimization capability The existing technology only covers the basic matching of "expert-bid package", ignoring key dimensions: ① Expert load (such as if a same expert's monthly bid evaluation times exceed 3 times, it is easy to lead to the decline of bid evaluation quality); ② Bid evaluation base capacity, such as a base can accommodate a maximum of 5 bid evaluation groups, the existing technology does not consider the risk of base overload); the simple random selection system uses local random screening and cannot achieve global resource optimization; the basic rule tool selects according to fixed priority and is prone to the problem of professional matching meeting the requirements but wasting base resources.

[0084] In view of the above technical defects, the present application provides an expert extraction scheme preparation method based on operational optimization, which solves the following technical problems: (1) Through multi-system automatic connection, data automatic preprocessing, and scheme automatic generation, "end-to-end automation" of expert extraction scheme is realized, and the large project scheme preparation time is shortened from 1-2 days to 30 minutes, while the manual operation error rate is reduced to below 1%; (2) Construct a multi-dimensional data model of fusing experts (profession, level, load, availability), packages (demand, level), bidding units (correlation), and evaluation bases (capacity, location), and realize the global optimization of "minimum resource waste, maximum evaluation matching degree, and compliance constraint satisfaction" through operational optimization algorithm; (3) Design an optimization model and rule base of "parameterized configuration", support business personnel to adjust the target function weight and constraint condition parameter through visual interface, without development intervention, and shorten the response demand change cycle to within 1 hour.

[0085] The technical key points of the expert extraction scheme preparation method based on operational optimization provided by the application are as follows: (1) Multi-objective and multi-constrained dynamic operational optimization model A three-objective comprehensive function including "maximizing evaluation matching degree, minimizing resource waste, and minimizing risk" is constructed, and the user can dynamically configure the weight; at the same time, 6 kinds of rigid constraints (profession, level, base, load, avoidance, and grouping) are set, covering the full-dimensional compliance requirements of expert extraction; the core problem of single consideration factor and no global optimization in the prior art is solved, which is the key to the global optimization of the scheme, and the design of the objective function and the setting of the constraint condition are all different from the simple rules of the prior art.

[0086] (2) Technical key points and points to be protected 2: efficient solving mechanism based on improved genetic algorithm "Binary + real number hybrid coding" is used to represent expert-package association and grouping information, and "adaptive crossover probability" and "local mutation" are designed to improve the model solving efficiency and convergence accuracy; the problems of "slow solving and low accuracy" in the prior art are solved, and the algorithm improvement is targeted at the complexity characteristics of the expert extraction scheme, and the solving speed and accuracy are superior to those of the traditional algorithm.

[0087] The beneficial effects of the expert extraction scheme preparation method based on operational optimization provided by the application are as follows: (1) Efficiency is significantly improved, and labor cost is greatly reduced Through intelligent preparation technologies such as automatic docking of multiple systems, automatic solving of models, and automatic generation of schemes, the preparation time of the expert extraction scheme of the procurement project is shortened from 1-2 days of traditional manual operation to within 30 minutes, and the efficiency is improved by more than 95%. At the same time, 100% automation is realized in data acquisition, conflict investigation, and compliance verification, and business personnel only need to participate in parameter configuration and manual fine-tuning, labor cost is reduced by 80%, data preprocessing error rate is effectively reduced, compliance verification pass rate is significantly improved, and repetitive labor is reduced.

[0088] (2) Scheme quality optimization, realizing global resource optimal configuration Through the multi-objective optimization model, the matching degree of experts and the professional package is improved, and high-level and experienced experts are preferentially selected, so that the evaluation quality is significantly improved. According to the simulation test, the audit pass rate of the evaluation report is effectively controlled. Through overall optimization of base resources and expert load, the utilization rate of base resources is improved to 80%, avoiding the problem of saturation of some bases and idling of some bases; the balance degree of expert load is improved by 30%, and the difference in the number of monthly evaluation of the same expert is significantly reduced, reducing the evaluation errors caused by expert fatigue.

[0089] (3) Compliance risk is controllable, and meets the audit traceability requirements By built-in 12 types of compliance rule library, the rules such as “conflict avoidance”, “grade mismatch” and “base overload” are automatically excluded, completely eliminating the problem of “conflict checking omission” in manual preparation, and reducing the compliance risk. A “full-process log” is generated to record data sources, parameter configuration, solution process and adjustment records. The log cannot be tampered with, and the audit can quickly trace the scheme generation logic, meeting the requirements of transparency and auditability of central enterprises. The expert extraction scheme preparation process is solidified into standardized steps to avoid the randomness of manual operation.

[0090] Embodiment 2 Based on the same inventive concept, the application also provides an expert extraction scheme preparation system based on operational optimization, the system structure is as shown in Figure 6 The system comprises: A data set obtaining module 601 is configured to preprocess collected expert extraction multi-source data to obtain a standardized data set; the expert extraction multi-source data at least comprises package demand data and expert data; A model parameter configuration module 602 is configured to configure parameters of a multi-objective operational optimization model based on the standardized data set, wherein the parameters of the multi-objective operational optimization model comprise weights of a comprehensive objective function and threshold values of constraint conditions; An initial scheme generation module 603 is configured to solve the multi-objective operational optimization model by using an optimization algorithm to generate an optimal solution data set, and generate an initial expert extraction scheme based on the optimal solution data set; A final scheme preparation module 604 is configured to perform compliance verification on the initial expert extraction scheme, perform effect evaluation on the initial expert extraction scheme that passes the compliance verification, and receive a manual fine-tuning instruction based on a visual interface. The system automatically verifies the adjusted compliance and evaluation indexes, and prepares a final expert extraction scheme.

[0091] Preferably, the optimization algorithm is an improved genetic algorithm. The initial scheme generation module is specifically configured to: A binary and real number hybrid coding method is used to encode candidate solutions of the multi-objective operational optimization model to generate a chromosome; Based on the encoding mode, an initial population containing a plurality of chromosomes is initialized as a current population; A comprehensive objective function of the multi-objective operational optimization model is taken as a fitness function to calculate a fitness value of each chromosome in the current population; A selection operation is performed on the current population according to the fitness value, and a crossover operation with adaptive crossover probability and a mutation operation with a local mutation strategy are sequentially performed on the selected chromosomes to generate a new offspring population; The offspring population is updated as the current population, and it is determined whether the current population meets a convergence condition; If not, the step of calculating the fitness of the current population is returned to continue iteration; If yes, a solution corresponding to a chromosome with the highest fitness in the current population is determined as an optimal solution dataset.

[0092] Preferably, the optimization algorithm is a particle swarm optimization algorithm; The initial scheme generation module is specifically configured to: Initialize a particle swarm, and a position vector of each particle is used to represent a solution of the multi-objective operational optimization model, and a velocity vector represents a parameter adjustment direction; A comprehensive objective function of the multi-objective operational optimization model is taken as a fitness function to calculate a fitness value of each particle in the current particle swarm; According to the fitness value, an individual historical optimal solution of each particle and a global historical optimal solution of the entire particle swarm are updated; According to the individual historical optimal solution and the global historical optimal solution, a velocity vector and a position vector of each particle are updated; The updated particle swarm is taken as the current particle swarm, and it is determined whether a preset convergence condition is met; If not, the step of calculating the fitness value of the current particle swarm is returned to continue iteration; If yes, a solution corresponding to a position vector of a particle with the highest fitness in the current particle swarm is determined as an optimal solution dataset.

[0093] Preferably, the multi-source data comes from at least one of an expert library system, a bidding and tendering system, a bid evaluation and bid evaluation base management system, and a supplier association system; The multi-source data includes at least one of package demand, expert information, base information, and bid unit association information; The multi-source data is collected through real-time docking by an API interface or batch file import.

[0094] Preferably, the comprehensive objective function of the multi-objective operational optimization model includes multi-objective functions of maximizing an evaluation matching degree function, minimizing a resource waste function, and minimizing an evaluation risk function; The constraint conditions include at least one of the following: professional matching constraint, expert level constraint, base capacity constraint, expert load constraint, compliance avoidance constraint, bid evaluation group number constraint, first-level expert forced constraint, bid package-bid evaluation group unique correspondence constraint; The calculation formula of the comprehensive target function is as follows:

[0095]

[0096]

[0097]

[0098] wherein, is a comprehensive target function, is a maximum bid evaluation matching degree function, is a minimum resource waste function, is a minimum bid evaluation risk function, is a bid evaluation matching weight, is a resource waste weight, is a bid evaluation risk weight; is the professional matching degree of an expert and a bid package , , is the importance weight of a bid package , , is the level coefficient of an expert , is whether an expert participates in the bid evaluation of a bid package , indicates participation, indicates non-participation, is the total number of experts, is the total number of bid packages; is the actual number of occupied groups of a base , is the maximum capacity of a base , is the total number of bases; is the risk value of an expert , , is the total number of bid evaluation bid packages participated in by an expert this time.

[0099] Preferably, the final scheme preparation module is specifically used for: The initial scheme extracted by the expert is subjected to rule checking based on a compliance rule library, and data consistency checking is performed on the initial scheme extracted by the expert and a standardized data set, to generate a comprehensive checking result; If the comprehensive checking result indicates that there are rule violation items, corresponding exception handling processes are triggered according to the types of the rule violation items; If the comprehensive checking result indicates that the initial scheme extracted by the expert passes the checking, an effect evaluation is performed on the initial scheme extracted by the expert that passes the compliance checking. The effect evaluation includes: A plurality of evaluation indexes of the initial scheme extracted by the expert that passes the checking are calculated, and the evaluation indexes include one or more of a matching degree index, a resource utilization rate index, and a risk index. An evaluation report is generated based on the evaluation indexes, and it is determined whether each index reaches a preset threshold.

[0100] Preferably, the final scheme preparation module is specifically configured to: Through a visual interface, a user's manual fine-tuning instruction for the initial scheme extracted by the expert that passes the compliance checking is received, and the manual fine-tuning instruction includes one or more of expert replacement, grouping adjustment, and time adjustment. In response to the manual fine-tuning instruction, a corresponding adjustment operation is performed, and an adjusted expert extraction scheme is generated. The adjusted expert extraction scheme is subjected to compliance checking and re-computation of evaluation indexes. When the results of the checking and the re-computation confirm that the adjusted expert extraction scheme is compliant and the evaluation indexes meet the threshold, the adjusted expert extraction scheme is determined as the final expert extraction scheme.

[0101] Preferably, the system further includes a final scheme output module configured to: The final expert extraction scheme is output in a preset format of one of Excel, PDF, or XML. A whole-process log is synchronously generated, and the whole-process log is used to record operation records and parameter information in a whole process from data collection to scheme confirmation. The final expert extraction scheme, the effect evaluation report, and the whole-process log are packaged and archived to an enterprise document management system.

[0102] It can be understood that the preparation system of the expert extraction scheme based on operational optimization and the preparation method of the expert extraction scheme based on operational optimization cooperatively realize intelligent preparation of the expert extraction scheme, wherein the platform system is a carrier of technology, and the intelligent preparation technology is a core logic of the system.

[0103] The preparation system of the expert extraction scheme based on operational optimization provided by the present application will be described in detail in combination with a specific implementation process. Figure 7As shown, the specific architecture diagram of the expert extraction scheme preparation system based on operational optimization provided by the present application. The expert extraction scheme preparation system based on operational optimization adopts a "layered architecture + modular design", which is divided into 6 layers, and each layer is connected through a standardized interface. The data flow is "data acquisition layer-data preprocessing layer-operational optimization model layer-scheme generation and verification layer-scheme evaluation and adjustment layer-result output layer", and the specific composition is as follows: (1) Data acquisition layer Realize the automatic collection of multi-source data, provide complete and accurate basic data for subsequent models, and the composition modules include: 1) Expert library docking module: Docking the enterprise expert library system, collecting expert ID, professional category, level, available time, recent evaluation record, associated supplier list, and information such as the region to which it belongs; 2) Bidding system docking module: Docking the bidding management system, collecting project ID, number of bid packages, professional requirements of each bid package, project level, and bid unit list information; 3) Evaluation base management docking module: Docking the evaluation base management system, collecting base ID, location, maximum number of evaluation groups that can be accommodated, current number of occupied groups, and supporting facilities information; 4) Supplier association system docking module: Docking the enterprise CRM system and business information system, collecting the shareholder, senior management, and associated enterprise list of the bidding unit, which is used to expand the avoidance rules; Through API interface and external system real-time docking, support "real-time collection" and "timely synchronization", and the collected data is temporarily stored in "data buffer pool" to provide input for the data preprocessing layer.

[0104] In some optional embodiments, the collected data can also be input in the form of "batch Excel import". If the enterprise external system does not open the API interface, the user can export data from each system and upload the data through the system Excel batch import module. The system automatically parses the Excel table; it is suitable for enterprises whose external systems have not realized digital upgrade and cannot provide API interface, or temporary projects.

[0105] (2) Data preprocessing layer Clean, standardize, and detect conflicts on the collected raw data to ensure that the data quality meets the model requirements, and the composition modules include: 1) Data cleaning module: First, handle missing values, and when the expert available time is missing, fill in "available on weekdays from 9:00 to 17:00" by default; handle abnormal values; second, when the expert level exceeds "level 1-3", mark it as "to be manually verified"; third, delete duplicate expert records; 2) Data standardization module: First, unified data coding, such as expert professional coding according to "National Grid Expert Professional Classification Standard (Q / GDW 12345-2022)", and package professional demand coding according to the same standard to ensure consistency; Second, unified data format, such as converting expert available time to "YYYY-MM-DD HH:MM:SS" format; Third, establish data conflict detection module to detect expert available time conflict, such as expert has been booked for a certain period of time for bid evaluation, marked as unavailable; Fourth, detect base capacity conflict, such as the current base has occupied groups ≥ maximum capacity, marked as saturated; Fifth, detect association conflict, such as the expert and the bidding unit have unrecorded association, marked as high risk; Receive raw data from the data collection layer "data buffer pool", generate "standardized data set" after processing, and store it in the local database of the system to provide input for the operations optimization model layer.

[0106] (3) Operations optimization model layer Construct a multi-objective, multi-constrained operations optimization model, solve the global optimal solution through improved algorithm, and the modules include: 1) Objective function construction Objective 1: Maximize bid evaluation matching degree

[0107] In the formula, is the professional matching degree of expert and package , valued at 0-1; is the importance weight of package , valued at 0.5-1; is the grade coefficient of expert , valued at first level = 1, second level = 0.8, third level = 0.6; indicates whether the expert participates in the bid evaluation of package , 0-1 variable, 1 indicates participation, 0 indicates non-participation; Objective 2: Minimize resource waste

[0108] In the formula, is the actual occupied group number of base , is the maximum capacity of base , The smaller the value, the less the resource waste; Objective 3: Minimize bid evaluation risk

[0109] In the formula, is the expert The risk value ranges from 0 to 1, with higher risk values ​​indicating closer correlations. For experts The total number of bid packages participating in this evaluation; Overall objective function:

[0110] In the formula, , , The weights configured for users all range from 0 to 1, and the sum of the three is 1.

[0111] 2) Setting Constraints Constraint 1: Professional Matching Constraint:

[0112] In the formula, T is the minimum matching threshold configured by the user, with a default value of 0.8; Constraint 2: Expert Level Constraint:

[0113] In the formula, For standard package Minimum required expert level coefficient, Level 1 package =1.0, Level 2 package =0.8, Level 3 package =0.6, corresponding one-to-one with the project level; Constraint 3: Base capacity constraint:

[0114] In the formula, base The actual number of occupied groups does not exceed the maximum capacity; Constraint 4: Expert Load Constraint

[0115] In the formula, For experts The number of times the bid has been evaluated this month. The maximum number of times that can be configured for a user is 3 by default, but it is configurable. Constraint 5: Compliance Avoidance Constraint:

[0116] In the formula, For experts The evaluation risk value ranges from 0 to 1. This applies when the expert has a direct relationship with the bidding unit (e.g., holding a position or shareholding). =0.8-1.0, when there is an indirect relationship (such as a relative holding a position). = 0.5-0.7, no correlation = 0-0.4; the set of experts who are "associated with the bidding unit", ≥ 0.8; to any expert within the set of experts , forced not to participate in the evaluation of all packages j, completely avoiding the risk of conflict of interest; Constraint 6: Evaluation team size constraint:

[0117] wherein, is a 0-1 variable, 1 indicates that the evaluation task of the package is assigned to the evaluation team , 0 indicates that it is not assigned, and the number of experts in each evaluation team is controlled to be 3-5; Constraint 7: Primary expert mandatory constraint:

[0118] wherein, is a 0-1 variable, 1 indicates that the expert is a primary expert, and 0 indicates that the expert is not a primary expert, and each evaluation team contains at least one primary expert; Constraint 8: Package-evaluation team unique correspondence constraint:

[0119] wherein, the evaluation task of each package is only assigned to one evaluation team, is the total number of evaluation teams.

[0120] 3) Improved genetic algorithm solution In view of the problems of high model complexity and slow traditional algorithm solution, an improved genetic algorithm is adopted: Coding method: "binary + real number hybrid coding" is adopted, the binary part represents "expert-package" association (0 represents no association, 1 represents association), and the real number part represents the evaluation team assignment of the expert; Fitness function: directly use the comprehensive objective function F, the higher the fitness value, the better the scheme; Genetic operation optimization: the crossover operation adopts "adaptive crossover probability", the crossover probability of individuals with high fitness is low, and the excellent genes are preserved, the mutation operation adopts "local mutation", only the conflict variables are modified, such as the grouping that violates the base capacity constraint; Convergence condition: when the number of iterations reaches 100 times or the fitness value changes by ≤0.001 for 10 consecutive iterations, stop solving; Read parameters from the data preprocessing layer "standardized dataset", receive user-configured weights and constraint thresholds, solve to generate an "optimal solution dataset" containing expert-bid package association, grouping information, and base allocation information, and pass to the scheme generation and verification layer.

[0121] In some alternative embodiments, a particle swarm optimization algorithm can also be selected for solving. Specifically, the particle swarm optimization algorithm represents expert-bid package association and grouping information through particle position, and represents parameter adjustment direction through particle velocity, takes the comprehensive objective function of the present application as the fitness function, and solves the optimal solution through iterative updating of individual extreme value and global extreme value; it is suitable for small and medium-sized projects with bid package quantity ≤10 and expert quantity ≤50, and its solving speed is 30% faster than the improved genetic algorithm, but in large projects (bid package ≥20), the convergence accuracy is slightly lower.

[0122] (4) Scheme generation and verification Generate a standardized scheme based on the "optimal solution dataset" and perform full-process compliance verification to ensure that the scheme is risk-free. The composition module includes the following parts: 1) Automatic scheme generation Generate a scheme according to the enterprise standard template, which includes: project basic information, including project ID and bid package quantity; expert information, including name, ID, specialty, level, and grouping; base information, including base ID, location, and grouping allocation; evaluation time arrangement and avoidance explanation, including excluded experts and reasons.

[0123] 2) Compliance verification ① Rule verification: call the built-in "compliance rule library", which contains 12 types of rules, such as "expert level matching project level" and "no same unit association for experts in the same group"; automatically detect whether the scheme complies with the rules, and if it violates, mark it as a "violation item", such as "insufficient expert level for bid package 3"; ② Data consistency verification: compare the scheme data with the "standardized dataset" to ensure that there is no data tampering, such as consistency between expert available time and scheme time.

[0124] 3) Abnormality handling If the verification finds a violation item, automatically determine the "violation type": for mild violations (such as expert travel cost exceeding the budget), prompt the user whether to accept; for severe violations (such as violating the avoidance rules), return to the operations research optimization model layer to adjust the constraint parameters and solve again; Receive the "optimal solution dataset" from the operations research optimization model layer, generate a scheme, and perform compliance verification. The scheme that passes the verification is passed to the scheme evaluation and adjustment layer, and the scheme that does not pass the verification is returned to the model layer or prompted to the user.

[0125] (5) Scheme evaluation and adjustment Effect evaluation of compliance solution, supporting manual fine-tuning, ensuring that the solution meets the actual business needs, the composition module includes the following parts: 1) Effect evaluation First, calculate the following 3 types of core indicators: ① Matching degree index: Actual value, reflecting the matching quality of experts and packages; ② Resource utilization rate index: , reflecting the efficiency of base resource utilization; ③ Risk index: Actual value, reflecting the compliance risk of the solution; Then generate an evaluation report, indicating whether the indicators meet the standard, and the standard threshold is configured by the user, such as resource utilization rate ≥ 70%.

[0126] 2) Manual fine-tuning Provide a visual interface to support users to make 3 types of adjustments based on the "optimal solution": ① Expert replacement: replace a certain expert with other experts of the same profession and the same level, and the system automatically checks the compliance after replacement; ② Group adjustment: adjust the evaluation group allocation of experts, and the system automatically detects the base capacity constraints; ③ Time adjustment: adjust the evaluation time, and the system automatically updates the available state of the expert; After adjustment, the system automatically recalculates the evaluation indicators; Receive the compliance solution from the solution generation and verification layer, generate an evaluation report, and support manual fine-tuning to pass the final solution to the result output layer.

[0127] (6) Result output layer Output the final solution and related logs, support audit traceability and business application, the composition module includes the following parts: 1) Solution output module After outputting the solution, the user can output the following 3 types of formats: Excel: for business personnel operation; PDF: for approval; XML: for interfacing with the bidding system, and supports online preview.

[0128] 2) Log generation Generate full-process logs, including: data collection time, preprocessing parameters, model weight and constraint configuration, iterative solution process, compliance verification result, manual adjustment record, logs cannot be tampered with, stored in a blockchain node to enhance security; 3) Archiving module Finally, the system archives the final solution, evaluation report, and full-process logs to the enterprise document management system, supporting query by project ID.

[0129] Example 3: Based on the same inventive concept, the present application also provides an electronic device, which can be a computer device, a single-chip microcomputer device, a smart mobile device, etc. The electronic device in the embodiment can include a processor, a memory, a transceiver component, etc. The memory, the processor and the transceiver component are connected through a bus; the memory can be used to store an execution program, and the exemplary execution program can include instructions; the processor is used to execute the instructions stored in the memory. The memory can also be used to store data, which can be called and / or modified when the instructions are executed. Figure 8

[0130] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, and are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions in the readable storage medium to implement a corresponding method flow or a corresponding function, so as to implement the steps of the method for preparing an expert extraction scheme based on operational optimization in the above embodiment.

[0131] Embodiment 4: Based on the same inventive concept, the present application also provides a readable storage medium, specifically an electronic device readable storage medium (Memory), which is a memory device in the electronic device and is used to store programs and data. It can be understood that the readable storage medium here can include the built-in storage medium in the electronic device, and of course can also include the expansion storage medium supported by the electronic device. The storage medium provides a storage space, and the storage space stores the operating system of the terminal. Moreover, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more execution programs (including program codes). It should be noted that the storage medium here can be a high-speed RAM memory, or a non-volatile memory, for example, at least one disk memory. One or more instructions stored in the storage medium can be loaded and executed by the processor to implement the steps of the method for preparing an expert extraction scheme based on operational optimization in the above embodiment.

[0132] ​Those skilled in the art will appreciate that embodiments of the application can be devised for a method, a system, or a computer program product. Accordingly, the present application can be embodied in the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) embodying computer readable program code.

[0133] The present application is described in reference to the flowchart and / or block diagrams of the method, apparatus (system) and computer program product according to embodiments of the application. It will be understood that each block of the flowchart and / or block diagrams, and combinations of blocks in the flowchart and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processing device or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0134] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0135] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0136] Finally, it should be noted that the above-described embodiments are merely intended for describing the technical solutions of the present application, but not to limit the scope of protection of the present application. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can make various changes and modifications to the specific embodiments, or equivalents thereof, after learning the spirit of the present application. However, these changes and modifications, or equivalents thereof, should fall within the scope of protection of the appended claims.

Claims

1. A method for developing an expert extraction scheme based on operations research optimization, characterized in that, include: The collected expert data from multiple sources was preprocessed to obtain a standardized dataset; The expert-extracted multi-source data includes at least standard package requirement data and expert data; Based on the standardized dataset, the parameters of the multi-objective operations research optimization model are configured, including the weights of the comprehensive objective function and the thresholds of the constraint conditions. The multi-objective operations research optimization model is solved using an optimization algorithm to generate an optimal solution dataset, and an initial expert extraction scheme is generated based on the optimal solution dataset. The initial expert selection plan is subject to compliance verification. The initial expert selection plan that passes the compliance verification is then evaluated for effectiveness. Manual fine-tuning instructions are received through a visual interface. The system automatically verifies the compliance and evaluation indicators of the adjusted plan and compiles the final expert selection plan.

2. The method according to claim 1, characterized in that, The optimization algorithm is an improved genetic algorithm; The step of solving the multi-objective operations research optimization model using an optimization algorithm to generate an optimal solution dataset includes: A hybrid binary and real number encoding method is used to encode the candidate solutions of the multi-objective operations research optimization model, generating chromosomes; Based on the encoding method, an initial population containing multiple chromosomes is generated and used as the current population; Using the comprehensive objective function of the multi-objective operation research optimization model as the fitness function, the fitness value of each chromosome in the current population is calculated; Based on the fitness value, a selection operation is performed on the current population, and then a crossover operation with adaptive crossover probability and a mutation operation with local mutation strategy are performed on the selected chromosomes in sequence to generate a new offspring population. Update the offspring population to the current population, and determine whether the current population satisfies the convergence condition; If the condition is not met, return to the step of calculating the current population fitness and continue iterating; If the conditions are met, the solution corresponding to the chromosome with the highest fitness in the current population is determined as the optimal solution dataset.

3. The method according to claim 1, characterized in that, The optimization algorithm is the particle swarm optimization algorithm. The step of solving the multi-objective operations research optimization model using an optimization algorithm to generate an optimal solution dataset includes: Initialize the particle swarm, where the position vector of each particle represents a solution of the multi-objective operations research optimization model, and the velocity vector represents the direction of parameter adjustment. Using the comprehensive objective function of the multi-objective operations research optimization model as the fitness function, the fitness value of each particle in the current particle swarm is calculated. Based on the fitness value, update the individual historical best solution of each particle and the global historical best solution of the entire particle swarm. Based on the individual historical optimal solution and the global historical optimal solution, update the velocity vector and position vector of each particle; The updated particle swarm is used as the current particle swarm, and it is determined whether the preset convergence condition is met. If the condition is not met, return to the step of calculating the current particle swarm fitness value and continue iterating; If the conditions are met, the solution corresponding to the position vector of the particle with the highest fitness in the current particle swarm is determined as the optimal solution dataset.

4. The method according to claim 1 or 2, characterized in that, The multi-source data comes from at least one of the following: expert database system, bidding system, bid evaluation base management system, and supplier association system. The multi-source data includes at least one of the following: tender package requirements, expert information, base information, and information related to bidding entities; The collection of multi-source data is achieved through real-time API interface connection or batch file import.

5. The method according to claim 1, characterized in that, The comprehensive objective function of the multi-objective operations research optimization model includes a multi-objective function that maximizes the bid evaluation matching degree function, minimizes the resource waste function, and minimizes the bid evaluation risk function; The constraints include at least one of the following: professional matching constraints, expert level constraints, base capacity constraints, expert workload constraints, compliance avoidance constraints, number of bid evaluation team members constraints, mandatory constraints for first-level experts, and unique correspondence between bid package and bid evaluation team. The formula for calculating the comprehensive objective function is as follows: in, For the comprehensive objective function, To maximize the evaluation matching function, To minimize the resource waste function, To minimize the evaluation risk function, To match weights for evaluation, Weighting for resource waste As a risk weight for bid evaluation; For experts With standard package Professional matching degree , For standard package Importance weights , For experts The grade coefficient, For experts Whether to participate in the bidding package The evaluation of bids, Indicate participation, Indicates non-participation. The total number of experts, The total number of bid packages; As a base The actual number of groups occupied. As a base Maximum capacity, Total number of bases; For experts The risk value, , For experts The total number of bid packages participating in this evaluation.

6. The method according to claim 1, characterized in that, The compliance verification of the initial expert extraction plan, followed by an effectiveness evaluation of the initial expert extraction plan that passes the compliance verification, includes: The initial expert extraction scheme is validated based on the compliance rule base, and the initial expert extraction scheme is validated against the standardized dataset to generate a comprehensive validation result. If the comprehensive verification result indicates that there is a violation, the corresponding exception handling process is triggered according to the type of the violation; If the comprehensive verification result indicates that the initial expert extraction scheme has passed the verification, then the effectiveness of the initial expert extraction scheme that has passed the compliance verification will be evaluated. The effect evaluation includes: Calculate multiple evaluation indicators for the verified expert extraction initial plan, including one or more of the following: matching degree indicator, resource utilization rate indicator, and risk indicator; An evaluation report is generated based on the evaluation indicators to determine whether each indicator has reached the preset threshold.

7. The method according to claim 1, characterized in that, The system receives manual fine-tuning instructions via a visual interface, automatically verifies the compliance and evaluation indicators of the adjusted settings, and compiles the final expert selection plan, including: Through a visual interface, users can receive manual fine-tuning instructions for the initial plan extracted by experts that has passed compliance verification. The manual fine-tuning instructions include one or more of the following: expert replacement, group adjustment, and time adjustment. In response to the manual fine-tuning instruction, the corresponding adjustment operation is executed to generate the adjusted expert extraction plan; The adjusted expert selection scheme was subjected to compliance verification and the evaluation indicators were recalculated. When the verification and recalculation results confirm that the adjusted expert selection scheme is compliant and the evaluation indicators meet the standards, the adjusted expert selection scheme will be determined as the final expert selection scheme.

8. A system for developing expert extraction schemes based on operations research optimization, characterized in that, include: The dataset acquisition module is used to preprocess the collected multi-source data extracted by experts to obtain a standardized dataset. The expert-extracted multi-source data includes at least standard package requirement data and expert data; The model parameter configuration module is used to configure the parameters of the multi-objective operations research optimization model based on the standardized dataset. The parameters of the multi-objective operations research optimization model include the weights of the comprehensive objective function and the thresholds of the constraint conditions. The initial solution generation module is used to solve the multi-objective operations research optimization model using an optimization algorithm, generate an optimal solution dataset, and generate an expert extraction initial solution based on the optimal solution dataset. The final scheme compilation module is used to perform compliance verification on the initial expert extraction scheme, evaluate the effectiveness of the initial expert extraction scheme that has passed the compliance verification, and receive manual fine-tuning instructions based on a visual interface. The system automatically verifies the compliance and evaluation indicators after adjustment and compiles the final expert extraction scheme.

9. An electronic device, characterized in that, include: At least one processor and memory; The memory and processor are connected via a bus; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, a method for developing an expert extraction scheme based on operations research optimization as described in any one of claims 1 to 7 is implemented.

10. A readable storage medium, characterized in that, It contains an execution program, which, when executed, implements a method for compiling an expert extraction scheme based on operations research optimization as described in any one of claims 1 to 7.

Images