Multi-end cooperative intelligent scheduling platform and method for flexible labor management

Abstract

The application discloses a multi-end cooperative intelligent scheduling platform and method for flexible labor management, and particularly relates to the technical field of labor management, and comprises the following steps: a preset scheduling rule algorithm performs work scheduling according to flexible labor requirements, and generates a scheduling arrangement form each time; in each scheduling period, a feature set affecting scheduling is extracted, the fairness and rationality of work arrangement are evaluated, and an evaluation report is generated; according to the evaluation report, a feedback signal is generated to determine whether to optimize the scheduling rule algorithm. In the optimization process, an optimization scheme is designed through the feedback signal, the scheduling algorithm is iteratively trained, the scheduling algorithm is better adapted to the labor requirements, the work distribution is ensured to be more fair and reasonable, and the optimized algorithm is applied to scheduling in the next period; the application can scientifically evaluate whether the task distribution is fair and reasonable, can perform targeted optimization training on unreasonable parts in the task distribution, and gradually realizes that the task distribution in the labor environment is more scientific and efficient.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of labor management, more particularly, the present application relates to a multi-end collaborative intelligent scheduling platform and method for flexible labor management. BACKGROUND

[0002] With the rapid development of economy and the increasingly fierce market competition, the demand for labor in enterprises is becoming more flexible and diverse. Flexible labor mode has gradually become an important way of human resource management and is widely used in e-commerce, logistics, service and other industries. Traditional labor management mode is usually based on long-term labor relationship, which cannot effectively respond to temporary, phased or dynamic human resource demand. This rigid labor management method is prone to the following problems: fixed labor mode is prone to waste of labor resources when demand fluctuates greatly, and may cause labor shortage during peak period, which is difficult to meet the flexible and efficient management needs of enterprises. Traditional scheduling method is usually based on manual experience, lacks scientificity and intelligent support, and is prone to unfair or unreasonable task allocation, affecting labor efficiency and employee satisfaction. When facing dynamic task demand, the existing scheduling system lacks real-time adjustment capability, which leads to task allocation lag and affects the overall operation effect. Therefore, the multi-end collaborative intelligent scheduling platform and method for flexible labor management are proposed to solve the above problems. SUMMARY

[0003] To achieve the above purpose, the present application provides the following technical solutions:

[0004] The intelligent scheduling method for multi-end collaboration of flexible labor management comprises the following steps:

[0005] A scheduling rule model is preset, and the preset scheduling rule model is used for work scheduling according to the flexible labor demand to obtain a work scheduling arrangement form each time;

[0006] In each scheduling period, based on all work scheduling arrangement forms in the scheduling period, feature extraction is performed to extract relevant features affecting labor scheduling, and an influence feature set is obtained;

[0007] According to the influence feature set, the result of each work scheduling arrangement in the scheduling period is evaluated for fairness and rationality, and an evaluation report is generated;

[0008] Different types of feedback signals are generated according to the evaluation report, and whether to optimize and train the preset scheduling rule model is determined according to the type of feedback signal;

[0009] During the optimization training, an optimization training scheme is generated according to the type of the feedback signal, the preset scheduling rule model is optimized through iterative training, so that the scheduling rule model is gradually adapted to the scheduling demand in the working environment, and the optimized scheduling rule model is applied to the labor scheduling in the next period, so as to ensure that the work allocation is more fair and reasonable.

[0010] In a preferred embodiment, the extraction of the related features affecting the labor scheduling means respectively extracting a fairness feature group and a rationality feature group, and the fairness feature group and the rationality feature group together constitute the influence feature set.

[0011] In a preferred embodiment, the evaluation report includes a scheduling period fairness index obtained by performing fairness evaluation and a scheduling period rationality index obtained by performing rationality evaluation.

[0012] In a preferred embodiment, the scheduling period fairness index acquisition logic is as follows:

[0013] For each worker i, the work load and the work duration are respectively defined as:

[0014] ; represents a preset difficulty coefficient corresponding to the workload in the time period t;

[0015] ; represents the work duration in the time period t, represents whether the worker i works in the time period t, if the worker i works in the time period t, then is valued at 1, if the worker i does not work in the time period t, then is valued at 0, and T represents the total number of time periods;

[0016] The balance target of the overall workload and the work duration is obtained as follows:

[0017] First, the overall workload and the work duration are calculated respectively, and the total workload and the total work duration are respectively denoted as and : ; ; n represents the total number of workers;

[0018] Then, the balance target of each worker, i.e., the target workload and the target work duration, is calculated as follows:

[0019] The target workload is: ; represents the target workload of each worker;

[0020] The target work duration is: ; denotes the target working time length of each worker;

[0021] The working load deviation and working time length deviation of each worker are respectively:

[0022] ; denotes the working load deviation of worker i;

[0023] ; denotes the working time length deviation of worker i;

[0024] The worker preference difference is introduced for adjustment:

[0025] ; denotes the preference or resistance feedback of worker i at time period t, if the preference works at time period t, then is valued as 1, if the resistance works at time period t, then is valued as 0, denotes the preference difference value of worker i, denotes the preset preference proportion coefficient of worker i at time period t;

[0026] The scheduling cycle fairness index calculation formula is:

[0027] ; , are preset non-zero trade-off coefficients, denotes the total rest time of worker i, denotes the minimum allowed rest time, F denotes the scheduling cycle fairness index, and C is a preset non-zero constant.

[0028] In a preferred embodiment, the acquisition logic of the scheduling cycle rationality index is:

[0029] The task delivery timeliness is used to measure whether the task can be completed within the predetermined time, and the calculation formula is as follows:

[0030] ; N denotes the total number of working tasks, denotes the deadline of the kth task, denotes the actual completion time of the kth task, and TI denotes the task delivery timeliness;

[0031] The worker load balance is used to evaluate the rational allocation degree of the working time and load of each worker, and the calculation formula is as follows:

[0032] ; denotes the actual working time of the ith worker, denotes the total time required for the task assigned to the ith worker, LB denotes the worker load balance, and C is a preset non-zero constant;

[0033] The rational matching degree of personnel scheduling is evaluated by matching the task requirements and worker scheduling, and the calculation formula is as follows:

[0034] denotes the similarity between the skill of the ith worker and the skill required by each scheduled task, denotes the experience value of the ith worker, are preset non-zero measurement coefficients, and PP denotes matching adaptability;

[0035] The distances of task delivery timeliness TI, worker load balance LB, and matching adaptability PP to ideal solution and negative ideal solution are obtained, the distance to ideal solution is JL1, the distance to negative ideal solution is JL2, and the calculation formula of scheduling period rationality index is:

[0036] Y denotes the scheduling period rationality index.

[0037] In a preferred embodiment, the experience value of the ith worker is obtained by the following logic:

[0038] V denotes the total number of historical scheduling of the ith worker, and v denotes the index of the number of historical scheduling, denotes a preset adjustment factor, denotes the quality score of the vth scheduling of the ith worker, denotes the historical average quality score of the ith worker.

[0039] In a preferred embodiment, generating different types of feedback signals according to the evaluation report means:

[0040] Using fuzzy reasoning, the scheduling period fairness index and the scheduling period rationality index are used as input data of fuzzy reasoning, and the output result of fuzzy reasoning is the type of feedback signal. The type of feedback signal includes type one, type two, type three, and type four. Only when the type of feedback signal is type one, the preset scheduling rule model is not optimized and trained.

[0041] In a preferred embodiment, when optimizing and training, generating an optimization training scheme according to the type of feedback signal means:

[0042] ​​The training data quantity of a type of feedback signal is Z, and the training data quantity of the corresponding feedback signal of the type in the range of [2Z, 4Z] is selected to form a training set for optimization training, and in the training set for optimization training, all training data quantities are not less than a preset minimum training quantity threshold.

[0043] In a preferred embodiment, the multi-end collaborative intelligent scheduling platform for flexible labor management comprises:

[0044] The work scheduling module has a preset scheduling rule model inside, and the preset scheduling rule model is used for work scheduling according to the flexible labor demand to generate a work scheduling arrangement sheet each time;

[0045] The feature extraction module extracts relevant features affecting labor scheduling based on all work scheduling arrangement sheets in each scheduling period to obtain an influence feature set in each scheduling period;

[0046] The evaluation module evaluates the fairness and rationality of the result of each work scheduling arrangement in the scheduling period according to the influence feature set to generate an evaluation report;

[0047] The feedback module generates different types of feedback signals according to the evaluation report, and determines whether to optimize the preset scheduling rule model according to the type of the feedback signal;

[0048] The optimization module generates an optimization training scheme according to the type of the feedback signal during optimization training, optimizes the preset scheduling rule model through iterative training to gradually adapt to the scheduling demand in the labor environment, and delivers the optimized scheduling rule model to the work scheduling module to apply the optimized scheduling rule model to the labor scheduling in the next period to ensure that the work allocation is more fair and reasonable.

[0049] Technical effects and advantages of the present application:

[0050] The present application can quickly analyze and match the input labor demand and worker features based on the convolutional neural network intelligent algorithm, automatically generate a task allocation scheme, and significantly reduce the time cost of traditional manual scheduling.

[0051] The application can scientifically evaluate whether the task allocation is fair through the fairness index, avoid over-concentration of tasks on a few people, reduce the dissatisfaction of employees, and generate a rationality index of task allocation by considering the skill level, historical experience and actual demand of workers in the scheduling scheme, to ensure that the allocation meets the actual demand of the post and the task difficulty. The system can analyze the fairness and rationality index through fuzzy reasoning, and can optimize the unreasonable part of task allocation, so as to gradually realize more scientific and efficient task allocation in the labor environment.

[0052] The scheduling platform in the application can better view and receive the scheduling operation arrangement form through the cooperation of cloud, mobile and PC terminals, so that the enterprise managers and dispatchers can greatly improve the flexibility and convenience of management. The task scheduling table and execution state can be synchronized to the worker terminal in real time, and the worker can receive task instructions through the mobile terminal, thereby improving the transparency and employee experience of labor, and the platform supports efficient cooperation between multiple roles such as enterprise managers and labor personnel, and promotes smooth docking of task allocation and execution.

[0053] According to the fairness and rationality evaluation results of the scheduling period, the application generates a class, a class, a class, a class of feedback signals, and dynamically adjusts the optimization direction. For example, for the feedback signal of insufficient fairness, the system focuses on optimizing the balance of task allocation. Through historical data and periodic feedback, the system can optimize the scheduling rule model after each scheduling period if necessary, so that it gradually adapts to the complex and variable labor environment and realizes self-optimization.

[0054] The flexible labor management mode in the application can dynamically adjust the labor scale according to actual needs, avoid the redundant cost of fixed labor relations, reduce the waste of human resources, and through scientific and reasonable task allocation, workers can more efficiently complete the work, reduce the additional expenses caused by task delay or failure. The scheduling platform helps enterprises optimize labor strategies, further reduces management costs and improves efficiency, and greatly shortens the response time of task allocation through intelligent scheduling, so that enterprises can respond to market changes and unexpected demands more quickly. The intelligent and data-driven scheduling method makes the labor management of enterprises more refined, and improves the overall operation efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0055] In order to facilitate the understanding of those skilled in the art, the application will be further described below in conjunction with the drawings;

[0056] Figure 1 The schematic diagram of the intelligent scheduling method for multi-terminal cooperation of flexible labor management in the application.

[0057] Figure 2 The schematic diagram of the intelligent scheduling platform for multi-terminal cooperation of flexible labor management in the application. DETAILED DESCRIPTION

[0058] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work belong to the scope of protection of the present application.

[0059] With reference to Figure 1 - Figure 2 The following embodiments are obtained:

[0060] Embodiment 1: The flexible labor management in the present application refers to a non-traditional labor management mode. Enterprises flexibly allocate human resources and assign tasks according to actual post requirements and business changes. It has the characteristics of flexible labor mode, various management modes, and efficient cost control. Post demand orientation includes: dynamic post demand: enterprises determine post demand according to business changes (such as project start, emergency tasks, etc.), and do not have fixed labor relations. Task-oriented labor: arrange personnel by task, and the post is closely linked with task demand. Through preset rules or intelligent algorithms, the task is assigned to the most suitable personnel.

[0061] Multi-terminal collaboration refers to the integration of cloud, mobile and PC collaboration methods to achieve intelligent scheduling and efficient management of flexible labor. For example, the intelligent scheduling system for multi-terminal collaboration of flexible labor management in the present application is deployed in the cloud, the mobile terminal is the individual information collection port of the worker, and is used to send the work scheduling form to the target worker from the cloud, the PC terminal is the terminal for the enterprise to publish the work task demand, and is used to display the work scheduling form generated by the cloud to the enterprise. The main functions of the system deployed in the cloud include post demand and human resource matching, task allocation, scheduling feedback in each scheduling cycle, and decision whether to optimize and train the preset scheduling rule model.

[0062] More specifically, the present application provides an intelligent scheduling method for multi-terminal collaboration of flexible labor management, comprising the following steps:

[0063] A scheduling rule model is preset, and the preset scheduling rule model is used for work scheduling according to the flexible labor demand to obtain a work scheduling arrangement form each time; this is the basic setting of the system, which means that the scheduling process is based on predefined rules. These rules take into account the flexibility of labor demand (such as skill requirements, etc.), and generate a work scheduling table based on this. The scheduling rules may include work hour limits, etc., with the purpose of achieving a preliminary work arrangement. Ensure that the system can generate a scheduling plan that meets the actual demand according to the established rules. The scheduling rule model can be obtained by training based on a convolutional neural network (CNN), and the model is deployed in the cloud.

[0064] In each scheduling cycle, based on all work scheduling arrangement forms within the scheduling cycle, feature extraction is performed to extract relevant features that affect labor scheduling, obtaining an impact feature set; this step identifies those factors (i.e. features) that have a significant impact on the results of work scheduling by analyzing all work scheduling arrangement forms within each scheduling cycle. These features may include the efficiency of work scheduling, the workload of employees, task completion, scheduling flexibility, etc. By extracting and summarizing these key features, the system can provide necessary data support for subsequent evaluation and optimization.

[0065] According to the impact feature set, the results of each work scheduling arrangement within the scheduling cycle are evaluated for fairness and reasonableness, and an evaluation report is generated; in this step, the system analyzes each work scheduling arrangement based on the impact features extracted in the previous step to evaluate their fairness and reasonableness. Fairness evaluation may consider whether the workload and work time allocation of each employee are reasonably treated; while reasonableness evaluation may include whether resources are effectively utilized during task execution. Ensure that the scheduling of work is fair and reasonable in terms of expected goals, identify potential problems or unreasonable aspects in scheduling.

[0066] Different types of feedback signals are generated according to the evaluation report, and whether to optimize the preset scheduling rule model is determined according to the type of feedback signal; different types of feedback signals may be generated in the evaluation report according to different fairness and reasonableness results. For example, feedback signals may be two types: "increase fairness", three types: "increase reasonableness", four types: "increase reasonableness and fairness" or one type: "fairness, reasonableness meet requirements"; according to these feedback signals, the system decides whether to optimize the scheduling rules. Ensure timely feedback to the scheduling algorithm according to the actual evaluation results, avoid continuing to schedule according to inappropriate rules, and thus promote system optimization.

[0067] During the optimization training, an optimization training plan is generated according to the type of feedback signal, and the preset scheduling rule model is optimized through iterative training, so that it gradually adapts to the scheduling demand in the working environment, and the optimized scheduling rule model is applied to the labor scheduling in the next period to ensure that the work allocation is more fair and reasonable. This stage is the self-improvement process of the entire system. According to the feedback signal obtained in the previous step, the system will generate an optimization training plan, and through multiple iterative training of the scheduling rule model, it will better adapt to the actual labor demand. Through continuous training and optimization, the scheduling rules become more accurate and flexible, and can play a better role in the next period of labor scheduling. The ultimate goal is to ensure that the work allocation in the scheduling process is fair and reasonable, that is, on the basis of meeting the labor demand, the work burden of each employee is balanced as much as possible, and excessive work or unfair work allocation is avoided. At the same time, the efficiency of scheduling and the utilization rate of resources should be maximized. Improve the sustainability and employee satisfaction of the system, improve the overall work efficiency and resource utilization rate through fair and reasonable scheduling arrangements.

[0068] Extracting relevant features that affect labor scheduling means extracting a fairness feature group and a rationality feature group respectively, and the fairness feature group and the rationality feature group together constitute the influence feature set. The evaluation report includes the scheduling period fairness index obtained by performing fairness evaluation and the scheduling period rationality index obtained by performing rationality evaluation.

[0069] The extraction of the fairness feature group is to ensure that in the process of labor scheduling, each employee or worker obtains fair treatment and task allocation, and to avoid favoritism, discrimination or unequal treatment. Generally, fairness evaluation will focus on the following aspects: fairness of work burden: whether the workload of employees is evenly distributed, avoiding some employees from bearing too much or too little work. Fairness of scheduling time: whether the worker shift, work length and other arrangements are balanced. Work hour allocation: whether there are long-term work, excessive overtime and other unfair phenomena.

[0070] The rationality feature group focuses on evaluating the actual feasibility and efficiency of the scheduling scheme to ensure that the labor scheduling meets the demand while being efficient and economical. The rationality feature group usually focuses on the following aspects: reasonable allocation of work tasks: whether the task allocation meets the skill level, experience and other requirements of the workers, avoiding overly complex or easy task arrangements. Matching of production / service demand: whether the scheduling can meet the actual requirements of production planning, customer demand, etc., avoiding excessive or insufficient labor input. Optimization of resource utilization: whether to effectively utilize existing resources, avoiding waste or shortage.

[0071] A fairness index F is defined, which reflects the fairness of work allocation when scheduling work among multiple workers. Generally, the higher the fairness, the closer the index is to 1; the lower the fairness, the lower the index. The logic for obtaining the fairness index of the scheduling period is as follows:

[0072] For each worker i, the work load and the work duration are defined as follows:

[0073] ; represents the preset difficulty coefficient corresponding to the work amount in the time period t;

[0074] ; represents the work duration in the time period t, represents whether the worker i works in the time period t, if working in the time period t, takes the value of 1, if not working in the time period t, takes the value of 0, and T represents the total number of time periods;

[0075] Obtain the balance target of the overall work load and work duration:

[0076] First, calculate the overall work load and work duration, and let the total work amount and total work duration be and respectively: ; ; n represents the total number of workers;

[0077] Then, calculate the balance target of each worker, i.e., the target work load and target work duration:

[0078] Target work load: ; represents the target work load of each worker;

[0079] Target work duration: ; represents the target work duration of each worker;

[0080] The work load deviation and work duration deviation of each worker are as follows:

[0081] ; represents the work load deviation of worker i;

[0082] ; represents the work duration deviation of worker i;

[0083] The worker preference difference is introduced to adjust. Assuming that worker i is more inclined to the early shift and resists the night shift, the preference adjustment factor of worker i can be obtained by calculating the difference between the actual working period and the preference:

[0084] ; represents the preference or resistance feedback of worker i at time period t, if the preference works at time period t, then is valued at 1, if the resistance works at time period t, then is valued at 0, represents the preference difference value of worker i, represents the preset preference proportion coefficient of worker i at time period t;

[0085] If some workers do not get enough rest time, the fairness index needs to be adjusted. If the rest time is too short, the adjustment factor increases, so the scheduling period fairness index calculation formula is:

[0086] ; , are preset non-zero weighting coefficients to adjust the influence of rest time and worker preference on the fairness index, represents the total rest time of worker i, represents the minimum allowed rest time, is used to measure the sufficiency of worker rest time. F represents the scheduling period fairness index, the value of F is closer to 1, indicating that the scheduling is more fair, and closer to 0, indicating that the scheduling is unfair.

[0087] The logic for obtaining the scheduling period rationality index is:

[0088] The task delivery timeliness is measured by whether the task can be completed within the scheduled time. The calculation formula is as follows:

[0089] ; N represents the total number of work tasks, represents the deadline of the kth task, represents the actual completion time of the kth task, and TI represents the task delivery timeliness. The task delivery timeliness measures the closeness of the actual completion time of each task to the deadline. The value is closer to 1, indicating that the proportion of tasks completed on time is higher, and it is more timely.

[0090] The worker load balance is evaluated by the rational allocation degree of the working time and load of each worker. The calculation formula is as follows:

[0091] ; represents the actual working time of the ith worker, denotes the total time required for the task assigned to the ith worker, LB denotes the worker load balance; measures whether the load distribution among workers is balanced, the value closer to 1 indicates that the actual working time of the worker and the estimated time of the assigned task are less different, and the load is more balanced.

[0092] The matching adaptability is evaluated by matching the task requirements and the worker scheduling, and the calculation formula is as follows:

[0093] denotes the similarity between the skill of the ith worker and the skill required by each scheduled task, such as cosine similarity, denotes the experience value of the ith worker, are preset non-zero measurement coefficients, and PP denotes the matching adaptability, which measures the matching degree of the worker's ability and the task requirement, and the higher the value, the more the worker's ability can meet the task requirement.

[0094] The distances of the task delivery timeliness TI, the worker load balance LB, and the matching adaptability PP to the ideal solution and the negative ideal solution are obtained, the distance to the ideal solution is JL1, and the distance to the negative ideal solution is JL2, and the calculation formula of the scheduling period rationality index is:

[0095] Y denotes the scheduling period rationality index, the scheduling period rationality index measures the closeness to the ideal target, and the value closer to 1 indicates that the task allocation of the scheduling period is more reasonable and fair. The larger the scheduling period rationality index Y, the closer the current scheduling rule is to the ideal target, and the task allocation is more fair and reasonable. The closer the index is to 1, the higher the load balance of the worker, the timeliness of the task delivery, and the matching degree of the ability. When the rationality index is large, the scheduling can fully meet the actual needs of the labor environment, improve the overall management efficiency and resource utilization.

[0096] Ideal solution: refers to the comprehensive state when all evaluation indexes (TI, LB, and PP) reach the optimal value. The ideal solution represents the performance of the system in the most ideal case. Calculation method: assuming that the ideal value of each index is: task delivery timeliness (TI) ideal value: maximum value (usually 1, indicating that all tasks are completed on time). The ideal value of the worker load balance (LB) is the maximum value (usually 1, indicating that the load of all workers is completely balanced). The ideal value of the matching adaptability (PP) is the maximum value (usually 1, indicating that the task and the worker are completely matched). According to the ideal value of the index, the Euclidean distance between the current state (TI, LB, and PP) and the ideal solution is JL1.

[0097] ​​Negative ideal solution: refers to the comprehensive state when all evaluation indexes (TI, LB, PP) reach the worst value. The negative ideal solution represents the performance of the system in the most undesirable case. Calculation method: assuming that the negative ideal value of each index is: task delivery timeliness (TI) negative ideal value: minimum value (usually 0, indicating that all tasks are not completed on time). Worker load balance (LB) negative ideal value: minimum value (usually 0, indicating that all worker loads are extremely unbalanced). Matching adaptability (PP) negative ideal value: minimum value (usually 0, indicating that the task and the worker are completely mismatched). According to the negative ideal value of the index, the Euclidean distance between the current state (TI, LB, PP) and the negative ideal solution is calculated, which is JL2.

[0098] Experience value of the ith worker is a comprehensive evaluation index that measures the performance and ability of the worker in historical scheduling tasks, and the acquisition logic is:

[0099] ; V represents the total number of historical scheduling of the ith worker, indicating the number of scheduling tasks the worker has participated in, which is used to calculate the overall experience value of the worker, and is part of the denominator to calculate the average value, v represents the index of the historical scheduling number, ranging from 1 to V, used to traverse all historical scheduling tasks participated by the worker, represents a preset adjustment factor, controlling the influence intensity of , which determines the sensitivity of the score gap to the experience value calculation. The larger the value, the stronger the response of the experience value to the score deviation. represents the quality score of the vth scheduling of the ith worker, obtained by scoring the historical task completion performance, represents the historical average quality score of the ith worker. is an S-shaped function, with a value tending to 1, indicating that the worker's performance in this scheduling is better than the average level, a value of 0.5 indicating that the worker's performance is at the average level, and a value tending to 0 indicating that the worker's performance in this scheduling is lower than the average level. The experience value will be dynamically adjusted with the increase of the scheduling number and the change of the worker's performance, and can reflect the trend of the worker's ability.

[0100] According to the evaluation report, different types of feedback signals are generated, which means:

[0101] Using fuzzy reasoning, the scheduling period fairness index and the scheduling period rationality index are used as input data for fuzzy reasoning, and the output result of fuzzy reasoning is the type of feedback signal. The types of feedback signals include type 1, type 2, type 3 and type 4. Only when the type of feedback signal is type 1, the preset scheduling rule model is not optimized and trained.

[0102] In the scheduling cycle, the input variables include the fairness index and the rationality index of the scheduling cycle. The first step of fuzzy reasoning is to convert these precise values into fuzzy values in fuzzy sets. The fuzzy set of fairness index: low: indicates poor fairness, with significant imbalance in task allocation. Medium: indicates that fairness is acceptable but can be improved. High: indicates that task allocation is very fair and meets expectations.

[0103] The fuzzy set of rationality index: low: indicates that task allocation lacks rationality, with poor overall scheduling effect. Medium: indicates that there is a lack of rationality, but it is acceptable. High: indicates that task allocation has high rationality and meets requirements.

[0104] Through the preset membership function, the specific values of fairness and rationality index are attributed to the fuzzy set. For example, a trapezoidal or triangular membership function can be used to map input values to fuzzy membership degrees of "low", "medium", and "high".

[0105] Fuzzy rules are rules defined based on experience and logic, describing the relationship between input variables and output results. The rules are related to the combination of fairness and rationality index, mainly including the following:

[0106] If the fairness is high and the rationality is high, the feedback signal is "Class 1: both fairness and rationality meet requirements". If the fairness is medium and the rationality is high, the feedback signal is "Class 2: increase fairness". If the fairness is high and the rationality is medium, the feedback signal is "Class 3: increase rationality". If the fairness is low or the rationality is low, the feedback signal is "Class 4: increase rationality and fairness".

[0107] The fuzzy reasoning engine derives fuzzy output based on fuzzy rules and fuzzy membership degrees of input variables. Common reasoning methods include maximum membership degree method, weighted average method, etc. For all fuzzy combinations of input variables, determine the output of the corresponding rule. Combine the membership degrees of input variables to calculate the contribution of each rule to the output. For example, if the fairness is medium (membership degree is 0.6) and the rationality is high (membership degree is 0.8), the contribution of the rule "fairness medium and rationality high" is 0.6.

[0108] The fuzzy inference generates a fuzzy output, which needs to be converted into a specific feedback signal type. Common de-fuzzification methods include the center-of-gravity method and the maximum membership degree method. The center-of-gravity method calculates the center of gravity of the output fuzzy set as the specific output value. For example, the fuzzy set of signal types can be mapped to the interval of 0 to 1, corresponding to different feedback signal types. The maximum membership degree method selects the result corresponding to the output fuzzy set with the maximum membership degree as the final output. In the present application, the de-fuzzification result directly corresponds to the type of feedback signal, including: Category I: both fairness and rationality meet the requirements and do not need to be optimized. Category II: needs to be optimized to increase fairness. Category III: needs to be optimized to increase rationality. Category IV: needs to be optimized to increase both fairness and rationality.

[0109] According to the de-fuzzification result, a specific feedback signal type is generated, and it is decided whether to optimize the scheduling rule model: when the feedback signal is Category I, the scheduling rule does not need to be optimized, indicating that the existing rule has met the requirements of fairness and rationality. When the feedback signal is Category II, Category III, or Category IV, the system determines the optimization direction (improving fairness, rationality, or both) according to the feedback type and triggers the optimization training process.

[0110] During optimization training, generating an optimization training scheme according to the type of feedback signal means:

[0111] The number of training data for Category I feedback signal is Z, and the number of training data for the corresponding feedback signal type in the range of [2Z, 4Z] is selected to form the training set for optimization training, and in the training set for optimization training, the number of all training data is not less than the preset minimum training number threshold.

[0112] Specific examples are as follows: during each optimization training, the number of historical training data for Category I feedback signal (fairness and rationality meet the requirements) is Z as the basis data. At the same time, according to the type of feedback signal in the current scheduling period (such as Category II, Category III, or Category IV), the corresponding feedback signal training data is selected, and the number is 2Z to highlight the pertinence and focus on optimizing the deficiencies in the current scheduling period. In this way, by balancing the historical high-quality scheduling data (as the basis for reference) and the data of the current deficiencies (as the optimization focus), the adaptability of the scheduling rule model to the deficient direction can be improved more effectively.

[0113] The number of data for Category I feedback signal Z: Category I feedback signal data represents ideal cases where fairness and rationality meet the requirements in the scheduling period, and is used as the basis data for model learning to maintain the overall balance of the scheduling algorithm. The data number is fixed as Z and is extracted from historical scheduling periods.

[0114] The selection rule of the current feedback signal type: if the feedback signal type is type II (increasing fairness), select the type II feedback signal data related to "increasing fairness"; if the feedback signal type is type III (increasing rationality), select the type III feedback signal data related to "increasing rationality"; if the feedback signal type is type IV (increasing fairness and rationality), select the type IV feedback signal data related to both aspects.

[0115] The data quantity corresponding to the current period feedback signal is 2Z, which is twice the data quantity of type I feedback signal, for strengthening the learning ability of the model on the current deficiency. The data Z of type I feedback signal and the data 2Z of the current feedback signal are combined to form an optimization training set. In the training set: type I feedback signal data accounts for one third (maintaining the basic reference), and current feedback signal data accounts for two thirds (optimization focus). In the optimization training, type I feedback signal data is learned preferentially to maintain the overall fairness and rationality of the scheduling algorithm, and the data learning of the current feedback signal type is strengthened. The current feedback signal data can increase the proportion when the problem is serious in application. The problem being serious refers to a high degree of low scheduling quality.

[0116] Embodiment 2: A multi-end collaborative intelligent scheduling platform for flexible workforce management, deployed in the cloud, comprising:

[0117] A work scheduling module, which has a preset scheduling rule model inside, and the preset scheduling rule model is used to schedule work according to flexible workforce needs to generate a work scheduling arrangement form each time;

[0118] A feature extraction module, which extracts relevant features that affect workforce scheduling based on all work scheduling arrangement forms in each scheduling period to obtain an impact feature set;

[0119] An evaluation module, which evaluates the fairness and rationality of the results of each work scheduling arrangement in the scheduling period according to the impact feature set to generate an evaluation report;

[0120] A feedback module, which generates different types of feedback signals according to the evaluation report, and determines whether to optimize the preset scheduling rule model according to the type of the feedback signal;

[0121] An optimization module, which generates an optimization training scheme according to the type of the feedback signal during optimization training, optimizes the preset scheduling rule model through iterative training to make it gradually adapt to the scheduling needs in the workforce environment, and delivers the optimized scheduling rule model to the work scheduling module to apply the optimized scheduling rule model to workforce scheduling in the next period to ensure that work allocation is more fair and reasonable.

[0122] The above formulas are all de-dimensioned to calculate the numerical values, the formulas are obtained by collecting a large amount of data to simulate the most recent real situation, and the preset parameters in the formulas are set by a person skilled in the art according to the actual situation.

[0123] It should be understood that the size of the sequence number of each process described above in various embodiments of the present application does not mean the order of execution, and the execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0124] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0125] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the system, device and unit described above can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0126] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A multi-end collaborative intelligent scheduling method for flexible workforce management, characterized in that, Comprise the following steps: A scheduling rule model is preset, and the preset scheduling rule model is used for work scheduling according to flexible labor demand to obtain a work scheduling arrangement table for each time; In each scheduling period, based on all work scheduling arrangement tables in the scheduling period, feature extraction is performed, relevant features affecting labor scheduling are extracted, and an influence feature set is obtained; According to the influence feature set, the result of each work scheduling arrangement in the scheduling period is evaluated for fairness and rationality, and an evaluation report is generated; According to the evaluation report, different types of feedback signals are generated using fuzzy reasoning, and whether the preset scheduling rule model is optimized and trained is determined according to the type of the feedback signal; During the optimization training, an optimization training scheme is generated according to the type of the feedback signal, and the preset scheduling rule model is optimized through iterative training to gradually adapt to the scheduling demand in the labor environment; The logic for obtaining the scheduling period fairness index is: For each worker i, its workload and work duration are defined as: ; represents a preset difficulty coefficient corresponding to the workload in the time period t; ; denotes the working hours in time period t, denotes whether worker i works in time period t, if worker i works in time period t, then takes the value 1, if worker i does not work in time period t, then takes the value 0, T denotes the total number of time periods; Obtain the balance target of the overall work load and the work duration: Let the total workload and total working hours be denoted by and : ; ; n denotes the total number of workers; Computing target workloads: ; representing a target workload for each worker; Computing target work duration: ; denotes the target work duration of each worker; The work load deviation and the work duration deviation of each worker are respectively: ; represents the workload bias of worker i; ; denotes the work duration deviation of worker i; Adjustment is introduced to the worker preference difference: ; represents the preference or aversion feedback of worker i at time period t, if the preference works at time period t, takes the value of 1 if the aversion works at time period t, takes the value of 0, represents the preference difference value of worker i, represents the preset preference proportion coefficient of worker i at time period t; The scheduling period fairness index calculation formula is: ; , are preset non-zero trade-off coefficients, denotes the total rest time of worker i, denotes the minimum allowed rest time, F denotes the scheduling period fairness index, and C is a preset non-zero constant; The logic for obtaining the scheduling period rationality index is: The timeliness of task delivery is used to measure whether the task can be completed within the scheduled time, and the calculation formula is as follows: N denotes the total number of tasks, Tk denotes the deadline of the kth task, Tk denotes the actual completion time of the kth task, and TI denotes the task delivery timeliness. The worker load balance is used to evaluate the reasonable allocation degree of the work time and load of each worker, and the calculation formula is as follows: ; denotes the actual working time of the ith worker, denotes the total time required for the task assigned to the ith worker, LB denotes the worker load balance, and C is a preset non-zero constant; The matching adaptability of task demand and worker scheduling is used to evaluate the reasonable matching degree of personnel scheduling, and the calculation formula is as follows: ; represents the skill similarity of the ith worker to the skill requirement of each scheduled task, represents the experience value of the ith worker, , are preset non-zero measurement coefficients, and PP represents matching adaptability. The distance from the task delivery timeliness TI, the worker load balance LB, and the matching adaptability PP to the ideal solution and the negative ideal solution is obtained, the distance to the ideal solution is JL1, and the distance to the negative ideal solution is JL2. The scheduling period rationality index calculation formula is: ; Y represents a dispatch cycle rationality index; experience value of the ith worker The acquisition logic is: ; V represents the total number of historical scheduling of the ith worker, and v represents the index of the historical scheduling number, represents a preset adjustment factor, represents the quality score of the vth scheduling of the ith worker, represents the historical average quality score of the ith worker.

2. The multi-end collaboration intelligent scheduling method for flexible workforce management according to claim 1, characterized in that, Extracting relevant features affecting labor scheduling means extracting a fairness feature group and a rationality feature group respectively, and the fairness feature group and the rationality feature group together constitute the influence feature set. 3.The multi-end cooperative intelligent scheduling method for flexible workforce management according to claim 2, wherein, The evaluation report includes the scheduling period fairness index obtained by the fairness evaluation and the scheduling period rationality index obtained by the rationality evaluation.

4. The multi-end collaboration intelligent scheduling method for flexible workforce management according to claim 3, characterized in that, According to the evaluation report, different types of feedback signals are generated using fuzzy reasoning, which means: The scheduling period fairness index and the scheduling period rationality index are used as input data for fuzzy reasoning, the result of fuzzy reasoning is the type of feedback signal, the type of feedback signal includes type 1, type 2, type 3, and type 4. Only when the type of feedback signal is type 1, the preset scheduling rule model is not optimized and trained.

5. The multi-end collaboration intelligent scheduling method for flexible workforce management according to claim 4, characterized in that, During the optimization training, an optimization training scheme is generated according to the type of the feedback signal, which means: The number of training data of type 1 feedback signal is Z, and the number of training data of corresponding feedback signal type in the range of [2Z, 4Z] is selected to form a training set for optimization training, and the number of all training data in the training set for optimization training is not less than the preset minimum training number threshold.

6. A multi-party collaborative intelligent scheduling platform for flexible workforce management for implementing the multi-party collaborative intelligent scheduling method for flexible workforce management according to any one of claims 1 to 5, characterized in that, Comprise: A work scheduling module, which internally presets a scheduling rule model, and the preset scheduling rule model is used for work scheduling according to flexible labor demand to generate a work scheduling arrangement table for each time; The feature extraction module extracts features affecting the work scheduling based on all work scheduling schedules in each scheduling period to obtain an influence feature set; The evaluation module evaluates the fairness and rationality of the result of each work scheduling arrangement in the scheduling period according to the influence feature set, and generates an evaluation report; The feedback module generates different types of feedback signals according to the evaluation report, and determines whether to optimize and train the preset scheduling rule model according to the type of the feedback signal; The optimization module generates an optimization training scheme according to the type of the feedback signal during optimization training, optimizes the preset scheduling rule model through iterative training, makes it gradually adapt to the scheduling demand in the work environment, and delivers the optimized scheduling rule model to the work scheduling module.

Images