A Regional Classification Method and System in a Crowdsourcing Environment with Multi-Level Task Allocation

Abstract

The present invention relates to the technical field of task allocation, and in particular, to a method and system for regional classification in a crowdsourcing environment with multi-level task allocation. The method includes obtaining historical data of crowdsourcing workers; using a random forest model to predict the number of workers in each grid after grid division based on the historical data of crowdsourcing workers; obtaining a density ratio based on the predicted number of workers, and dividing the task area into low-ratio, medium-ratio, and high-ratio areas, wherein based on the goal of maximizing cost-benefit, a delay strategy is adopted in the low-ratio area, a balance strategy is adopted in the medium-ratio area, and a collaboration strategy is adopted in the high-ratio area; cross-grid area recruitment is used for the remaining tasks to complete the task allocation. The present invention provides a multi-level task allocation strategy for the regional classification method in a crowdsourcing environment. The solution divides the entire sensed area after gridification into low-ratio, medium-ratio, and high-ratio areas through the corresponding density ratio, fully considering the characteristics of different areas.

Description

Technical Field

[0001] The present invention relates to the technical field of task allocation, and in particular to a method and system for regional classification in a crowdsourcing environment with multi-level task allocation. Background Art

[0002] Task allocation is a crucial technical issue in mobile crowdsensing systems. The core lies in how to effectively schedule workers to complete various tasks while finding the best balance between resource utilization and service quality. With the increase in task complexity, traditional task allocation methods face many challenges, such as low allocation efficiency, long response time, and insufficient resource allocation. To address these challenges, a variety of emerging task allocation strategies have emerged in the research field. However, in a dynamically changing environment and complex task scenarios, these methods are still insufficient, lacking flexibility and intelligent processing capabilities. In traditional crowdsourcing systems, task allocation often relies on fixed rules and static information, making it difficult to achieve the best resource allocation when facing changes in the number of workers and task requirements. Especially when the spatio-temporal characteristics of tasks have a more obvious impact on resource scheduling, traditional methods are difficult to effectively cope with this challenge. Therefore, there is an urgent need for a new task allocation framework that can dynamically adjust strategies to adapt to changes in participants and tasks, thereby improving the overall execution efficiency. Summary of the Invention

[0003] To solve the above-mentioned problems, the present invention provides a method and system for regional classification in a crowdsourcing environment with multi-level task allocation. Based on the regional ratio classification of the predicted number of workers, the task area is divided into low-ratio, medium-ratio, and high-ratio areas, and corresponding strategies are adopted for different classified areas to allocate tasks within the area. After the task allocation in each area is completed, the remaining cost is used for task allocation across grid areas, thereby ensuring the task completion rate while ensuring a relatively high task sensing quality.

[0004] In the first aspect, a method for regional classification in a crowdsourcing environment with multi-level task allocation provided by the present invention adopts the following technical solutions:

[0005] A method for regional classification in a crowdsourcing environment with multi-level task allocation includes:

[0006] Obtain historical data of crowdsourcing workers;

[0007] Use a random forest model to predict the number of workers in each grid after grid division based on the historical data of crowdsourcing workers;

[0008] Based on the predicted number of workers, obtain the density ratio, and divide the task area into low-ratio, medium-ratio, and high-ratio regions. Among them, based on the goal of maximizing cost-effectiveness, adopt a delay strategy in the low-ratio region, a balance strategy in the medium-ratio region, and a collaboration strategy in the high-ratio region;

[0009] For the remaining tasks, adopt cross-grid area recruitment to complete the task allocation.

[0010] Furthermore, the use of the random forest model to predict the number of workers in each grid after grid division based on the historical data of crowdsourcing workers includes dividing the entire sensing area into grids, and using the random forest algorithm to predict the number of workers in each grid. Among them, the attribute set of crowdsourcing workers is , where respectively represent the horizontal and vertical coordinate positions of the worker, represents the perceived quality ability level of the crowdsourcing worker.

[0011] Furthermore, the obtaining of the density ratio based on the predicted number of workers includes using the ratio of the predicted number of workers in each grid to the number of tasks in each grid after the actual task is released as the density ratio, and dividing the entire sensing area into low-ratio, medium-ratio, and high-ratio regions through the density ratio.

[0012] Furthermore, the adoption of the delay strategy in the low-ratio region includes using the binary integer programming method to select the most cost-effective workers based on the perceived quality of the workers and the distance between the tasks, so as to maximize the cost-effectiveness ratio within each grid in the low-ratio region, expressed as:

[0013] ,

[0014] where represents the worker 's perceived quality, represents the cost of assigning the worker to the task . Among them, takes a value of 1 or 0. When the value is 1, it represents that the task is assigned to the worker , and when the value is 0, it represents that the assignment is not successful.

[0015] Furthermore, the adoption of the balance strategy in the medium-ratio region includes using a multi-round task-packing assignment method. First, initially pack the tasks in the region, and then assign them according to the emergency priority of the task packages. After the first-round packing assignment is completed, for the task packages that fail to be successfully assigned, unpack and repack them to improve the matching rate and the flexibility of task assignment.

[0016] ​Furthermore, the strategy of cooperation in the high ratio region includes selecting a combination of workers for each task through a greedy algorithm. Specifically, all possible combinations of workers for each task are permuted first, and the total cost and perceived quality of each combination are calculated one by one. When the total cost of a certain combination is the lowest or the perceived quality is the highest under the same cost, the system sets this combination as the current optimal choice. As the allocation progresses, the workers already assigned to a specific task will not be reassigned to other tasks, and the recruitment cost distance of the workers is calculated using the following formula:

[0017]

[0018] where represents the worker and the task point The Haversine distance between them is expressed as:

[0019]

[0020] where:

[0021]

[0022] where is the radius of the earth. In the present invention, the Haversine distance is used for all relevant distance calculations.

[0023] Furthermore, the allocation of the remaining tasks is completed by recruiting across grid regions, which includes, after completing the task allocation in the low ratio, medium ratio, and high ratio regions, analyzing the grid where each unassigned task is located, detecting all its directly adjacent grid regions, screening eligible workers in the adjacent grids according to the urgency of the task, generating an optional set of workers for each task, and achieving global optimal allocation through the linear programming method while maximizing the cost-benefit ratio of the objective function.

[0024] In a second aspect, a regional classification system in a crowdsourcing environment for multi-level task allocation includes:

[0025] A data acquisition module, configured to acquire historical data of crowdsourcing workers;

[0026] A prediction module, configured to use a random forest model to predict the number of workers in each grid after grid division based on the historical data of crowdsourcing workers;

[0027] A policy module, configured to obtain a density ratio based on the predicted number of workers, and divide the task area into low-ratio, medium-ratio, and high-ratio areas, where based on the goal of maximizing cost-effectiveness, a delay strategy is adopted in the low-ratio area, a balancing strategy is adopted in the medium-ratio area, and a collaborative strategy is adopted in the high-ratio area;

[0028] An allocation module, configured to recruit across grid regions for the remaining tasks to complete the task allocation.

[0029] In a third aspect, the present invention provides a computer-readable storage medium, in which multiple instructions are stored, and the instructions are adapted to be loaded and executed by a processor of a terminal device for the area classification method in a crowdsourcing environment with multi-level task allocation.

[0030] In a fourth aspect, the present invention provides a terminal device, including a processor and a computer-readable storage medium, where the processor is used to implement each instruction; the computer-readable storage medium is used to store multiple instructions, and the instructions are adapted to be loaded and executed by the processor for the area classification method in a crowdsourcing environment with multi-level task allocation.

[0031] In summary, the present invention has the following beneficial technical effects:

[0032] (1) The present invention provides a multi-level task allocation strategy for an area classification method in a crowdsourcing environment. The solution divides the entire sensed area after gridification into low-ratio, medium-ratio, and high-ratio areas through the corresponding density ratio, fully considering the characteristics of different areas.

[0033] (2) Different delay, balancing, and collaborative strategies are respectively adopted according to the three ratio areas. After the three ratio areas are divided, the division of cross-grid areas is considered, making full use of resources. While improving the task sensing quality, the task completion rate is also greatly improved. Description of the Drawings

[0034] Figure 1 It is a schematic diagram of task packaging within the distance threshold in the medium-ratio area of Embodiment 1 of the present invention; within;

[0035] Figure 2 It is a schematic diagram of worker recruitment within the cross-grid area of Embodiment 1 of the present invention;

[0036] Figure 3 It is a schematic diagram of the comparison of the sensing quality corresponding to different costs on different data sets in Embodiment 1 of the present invention;

[0037] Figure 4 It is a schematic diagram of the comparison of the task completion rate corresponding to different costs in different cities in Embodiment 1 of the present invention. Detailed Embodiment

[0038] The present invention will be further described in detail below with reference to the accompanying drawings.

[0039] Glossary:

[0040] Density ratio: The ratio of the number of workers in the predicted grid block to the actual number of tasks during the task publishing process is called the density ratio, which is expressed as:

[0041]

[0042] Where represents the predicted number of workers in the grid block, represents the actual number of tasks within the grid block. When it is a low-ratio area, it is a medium-ratio area, it is a high-ratio area.

[0043] Cost-benefit ratio: The cost-benefit ratio (CER) is the ratio between the benefit allocated to each task and the cost. It is used to measure the relative ratio between the quality of the workers allocated to the task and the corresponding cost. Therefore, the calculation of CER can be expressed as:

[0044]

[0045] Where represents the worker 's perceived quality, represents the cost of allocating the worker to the task .

[0046] Task completion rate: The task completion rate is the ratio of the number of tasks completed within a specific time to the total number of tasks, used to measure the efficiency of task allocation and execution, reflecting the success degree of task completion, and its definition is:

[0047]

[0048] Where represents the number of tasks completed, represents the total number of tasks.

[0049] Embodiment 1

[0050] Referring to Figure 1 , a regional classification method in a crowdsourcing environment with multi-level task allocation in this embodiment,

[0051] Obtain historical data of crowdsourcing workers;

[0052] Using a random forest model to predict the number of workers in each grid after grid division based on the historical data of crowdworkers;

[0053] Based on the predicted number of workers, obtain the density ratio, and divide the task area into low-ratio, medium-ratio, and high-ratio regions. Among them, based on the goal of maximizing cost-effectiveness, adopt a delay strategy in the low-ratio region, a balance strategy in the medium-ratio region, and a collaboration strategy in the high-ratio region;

[0054] For the remaining tasks, adopt cross-grid area recruitment to complete the task allocation.

[0055] Specifically:

[0056] S1. Obtain the historical data of crowdworkers. Record the task participation history of workers through a mobile application or a crowdsourcing platform. The system extracts the geographical location, time information related to the workers, and task-related features (such as weather, traffic, holidays, etc.). Subsequently, based on this data, map the historical activity trajectories of workers to the grids in the task area, and count the number of historical workers in each grid to provide data support for subsequent area division, task allocation, and strategy formulation.

[0057] S2. Use a random forest model to predict the number of workers in each grid after grid division based on the historical data of crowdworkers,

[0058] Among them, first, divide the entire sensing area (the geographical range where tasks are executed and data are collected in the mobile crowdsourcing sensing system. It is usually a specific target area that covers all locations where sensing tasks need to be completed) into grids. Then, use the random forest algorithm to predict the number of workers in each grid. This prediction is based on historical data, including the historical number of workers in each grid, and considers influencing factors such as economic level, traffic conditions, and weather to improve the accuracy of the prediction. Through the comprehensive analysis of these factors, the model can more accurately estimate the distribution of workers in different regions and provide reliable data support for subsequent task allocation strategies. During the prediction process, first perform data preprocessing, process categorical features (weather) and generate lag features (such as the number of workers in the past 7 days, 14 days, etc.) to capture the impact of time series. Subsequently, divide the entire dataset into a training set and a test set, and tune the hyperparameters through grid search and cross-validation to optimize the model performance. Finally, train multiple decision trees and combine the prediction results of each tree to give the final prediction value, reducing overfitting and improving the generalization ability. With the help of the relevant data in the first 39 days of the gMission dataset, predict the number of workers on the 40th day; at the same time, use the relevant data in the first 42 days of the EverySender dataset to predict the data on the 43rd day.

[0059] In the specific implementation process, the attribute set of crowdworkers is , where respectively represent the horizontal and vertical coordinate positions of the worker, represents the level of the worker, that is, the ability of the worker to perceive the task quality. The levels of the workers are divided into three types: senior workers (Grade A), with a perceived quality of 50; intermediate workers (Grade B), with a perceived quality of 30; junior workers (Grade C), with a perceived quality of 20.

[0060] S3. Obtain the density ratio based on the predicted number of workers,

[0061] where, with the number of workers in each grid obtained by prediction and the number of tasks in each grid after the actual task is released, according to the density ratio formula calculate the value . When , it is a low-ratio area, , it is a medium-ratio area, , it is a high-ratio area. Divide the entire perception area into low-ratio, medium-ratio and high-ratio areas.

[0062] In the specific implementation process, the attribute set of the crowdsourcing task is , where respectively represent the horizontal and vertical coordinate positions of the task. represents the urgency of the task, that is, how long it needs to be completed after the task is released. The urgency is divided into three categories: first-level task needs to be completed within half an hour after the task is released, second-level task needs to be completed within one hour, third-level task has no time limit. The ratio of the number of workers in each grid obtained by prediction to the actual number of tasks released is less than 1 for the low-ratio area, between 1 and 1.2 for the medium-ratio area, and the value is greater than 1.2 for the high-ratio area.

[0063] (1) In the process of task allocation in the low-ratio area, the scheme described in this example adopts the binary integer programming method, and based on the perceived quality of the worker and the distance between tasks, selects the most cost-effective worker to maximize the cost-benefit ratio in each grid of the low-ratio area. This process can be expressed by the following formula:

[0064] (3)

[0065] The constraint conditions are: (4)

[0066] (5)

[0067] where equals 1, representing that the task is assigned to the worker When it is equal to 0, it means that the allocation is not successful. The distance cost calculation for recruiting workers in this process is expressed as:

[0068] (6)

[0069] Where represents the Haversine distance between the worker and the task.

[0070] (2) In the process of task allocation in the medium ratio area, this example adopts a multi-round packaged task allocation strategy. First, the tasks in the area are initially packaged, and then they are allocated according to the emergency priority of the task packages. After the first round of packaged allocation is completed, for the task packages that fail to be successfully allocated, they are unpacked and repackaged to improve the matching rate and the flexibility of task allocation. Throughout the process, the goal is always to maximize the cost-benefit ratio to ensure the optimal efficiency and benefit of resource allocation.

[0071] Among them, the constraints include:

[0072] Constraints on the first-round task packaging rules: The packaging group follows the combination strategy of "primary task - secondary task" or "secondary task - tertiary task". There can only be one primary task and one secondary task in each packaging group, but there can be multiple tertiary tasks.

[0073] Constraints on the second-round task packaging rules: In the second round, the allocation is carried out according to the strategy of "tertiary task - tertiary task". The packaging group is allowed to contain multiple tertiary tasks.

[0074] Distance constraint between packaged tasks: The distance between any two tasks within the group shall not exceed .

[0075] In the specific implementation, according to factors such as the quantity and density of the data in the dataset, the value range is set between 5 kilometers and 8 kilometers. It can be understood that the value of can be adjusted according to actual needs.

[0076] (3) In the process of task assignment in the high ratio region, in this example, the greedy algorithm is used to select a combination of workers with higher perceived quality for each task. The system will consider all possible combinations of workers (such as A-C, A-B, B-C, and C-C), and calculate the total recruitment cost and perceived quality of each combination. Based on the greedy algorithm, the system preferentially selects the combination with the lowest total cost and the highest perceived quality. The specific steps are as follows: The system first arranges all possible combinations of workers for each task, and calculates the total cost and perceived quality of each combination one by one. When the total cost of a certain combination is lower or the perceived quality is higher under the same cost, the system sets this combination as the current optimal choice. As the assignment is completed, the workers who have been assigned to a specific task will not be reassigned to other tasks. During this process, the system calculates the recruitment cost distance of workers through the following formula:

[0077] (7)

[0078] S4. After completing the task assignment in the low ratio, medium ratio, and high ratio regions, this example uses a cross-grid region partitioning method for the remaining unassigned tasks. Specifically, the system will search for the set of available workers in adjacent grid regions to form an optional set of workers for each task, as Figure 2 shown. On this basis, the system further applies the linear programming method to find the global optimal solution by maximizing the cost-benefit ratio. This method comprehensively considers the perceived quality and recruitment cost of workers to ensure the effective assignment of the remaining tasks on the premise of minimizing costs and maximizing quality. In this way, even in different grid regions, the system can reasonably utilize resources and achieve global optimization across regions. Specifically, after completing the task assignment in the low ratio, medium ratio, and high ratio regions, analyze the grid where each unassigned task is located, and detect all its directly adjacent grid regions. These adjacent grid regions are regarded as potential candidate resources for further expanding the range of optional workers for the task. According to the urgency of the task, the system screens eligible workers in the adjacent grids within the reachable distance of the workers to generate an optional set of workers for each task. After determining the optional set of workers for each task, the system uses the linear programming method to maximize the cost-benefit ratio of the objective function while satisfying the corresponding constraints to achieve global optimal assignment. After the linear programming solution is completed, the system assigns an optimal worker to each unassigned task to ensure that the task assignment result not only meets the quality requirements but also minimizes the resource cost.

[0079] During the process of linear programming, by establishing a mathematical model, the goal is set to maximize the cost-benefit ratio, and it is solved in combination with relevant constraints. Specifically, the objective function uses the cost-benefit ratio of each task-worker combination as the weight to select the optimal allocation plan to achieve overall optimization. In the model: ① Decision variables are used to represent whether a task is completed by a certain worker; ② The objective function is defined as the maximization of the sum of the cost-benefit ratios of the selected combinations; ③ The constraints include that each task can only be assigned to one worker, each worker can only complete one task, and the distance between the task and the worker must be within the specified range. By introducing perceived quality and distance data, a linear programming tool (such as PuLP) will generate a standardized solution problem and apply a solver to calculate the global optimal solution of the optimal allocation of tasks and workers, thereby ensuring the highest perceived quality coverage at the lowest cost.

[0080] In the specific implementation, it is assumed that the walking speed of the worker is 60 km / h. According to the urgency of the task, the system sets the maximum distance range of reachable workers for different levels of tasks. For level-1 tasks, the maximum distance of reachable workers is 30 km; for level-2 tasks, the maximum distance range is extended to 60 km; and for level-3 tasks, the maximum distance range of reachable workers is set to 100 - 160 km.

[0081] To verify the effectiveness of the solution described in this embodiment, the following experimental verification was carried out in this embodiment:

[0082] The simulation experiment is as follows:

[0083] Using four groups of real datasets, experiments were conducted on the task allocation system on the EverySender and gMission datasets respectively: among them, the regional ratio algorithm represents the method described in this embodiment, the random allocation algorithm mentioned in "A heterogeneous mobile crowd sensing system for urban public safety", the lowest cost algorithm mentioned in "Participant quantity-aware online task allocation in mobile crowd sensing", the random utility value algorithm mentioned in "Online dependent task assignment in preference aware spatial crowdsourcing", and the two-stage online algorithm and two-stage greedy algorithm mentioned in "Two-sided online micro-task assignment in spatial crowdsourcing" represent the comparative task allocation algorithms.

[0084] In the experiment, some data from the EverySender and gMission datasets were extracted for evaluation. The practicability and coverage rate of the algorithm were evaluated by continuously increasing the labor cost and observing the task completion rate and the overall perceived quality of the tasks. Figure 3 in (a) to Figure 3 In (b) shows the comparison of the overall perceived quality after the matching of the crowdsourcing workers and the tasks at different costs. Figure 4 in (a) to Figure 4 In (b) shows the comparison of the task completion rates under different distribution methods at different costs. It can be seen from the figure that the method proposed in this implementation has a better distribution effect than other online algorithms.

[0085] Embodiment 2

[0086] This embodiment provides a regional classification system in a crowdsourcing environment with multi-level task allocation, including:

[0087] A data acquisition module, configured to acquire the historical data of the crowdsourcing workers;

[0088] A prediction module, configured to use a random forest model to predict the number of workers in each grid after grid division according to the historical data of the crowdsourcing workers;

[0089] A strategy module, configured to obtain a density ratio based on the predicted number of workers, and divide the task area into low-ratio, medium-ratio, and high-ratio areas, where based on the goal of maximizing cost-benefit, a delay strategy is adopted in the low-ratio area, a balance strategy is adopted in the medium-ratio area, and a collaboration strategy is adopted in the high-ratio area;

[0090] An allocation module, configured to recruit across grid regions for the remaining tasks to complete the task allocation.

[0091] A computer-readable storage medium, in which multiple instructions are stored, and the instructions are adapted to be loaded and executed by a processor of a terminal device for the regional classification method in a crowdsourcing environment with multi-level task allocation as described above.

[0092] A terminal device, including a processor and a computer-readable storage medium, the processor is used to implement each instruction; the computer-readable storage medium is used to store multiple instructions, and the instructions are adapted to be loaded and executed by the processor for the regional classification method in a crowdsourcing environment with multi-level task allocation as described above.

[0093] The above are all the preferred embodiments of the present invention. The protection scope of the present invention is not limited by this. Therefore, all equivalent changes made according to the structure, shape, and principle of the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for region classification in a crowdsourcing environment with multi-level task allocation, characterized in that: include: Obtain historical data of crowdsourcing workers; The random forest model is used to predict the number of workers in each grid after grid division based on the historical data of crowdsourcing workers; Based on the predicted number of workers, the density ratio is obtained, and the task area is divided into low ratio, medium ratio and high ratio areas. Based on the cost-effectiveness maximization goal, a delay strategy is adopted in the low ratio area, a balance strategy is adopted in the medium ratio area, and a collaborative strategy is adopted in the high ratio area. For the remaining tasks, cross-grid area recruitment is adopted to complete the task allocation; The strategy of adopting collaboration in the high ratio area includes selecting a worker combination for each task through a greedy algorithm, wherein all possible worker combinations for each task are first arranged, and the total cost and perceived quality of each combination are calculated one by one. When a combination has the lowest total cost or the highest perceived quality at the same cost, the system sets the combination as the current optimal choice. With the completion of the allocation, workers who have been assigned to a specific task will not be repeatedly assigned to other tasks, and the recruitment cost distance of the workers is calculated by the following formula: in, Representing the workers and mission points The Haversine distance between them is expressed as: in: in, is the radius of the earth, and the Haversine distance is used in the relevant distance calculations in the present invention; The delay strategy in the low ratio area includes adopting a binary integer programming method to select the most cost-effective worker based on the perceived quality of the worker and the distance between the tasks to maximize the cost-effectiveness ratio in each grid of the low ratio area, which is expressed as: ， in Indicates workers The perceived quality of Indicates that workers Assign to task The cost of The value is 1 or 0. When the value is 1, it means that the task Assign to workers , when the value is 0, it means that the allocation is not successful; The balanced strategy adopted in the medium ratio area includes adopting a multi-round task allocation method, firstly packaging the tasks in the area, and then allocating them according to the emergency priority of the task packages. After the first round of packaging and allocation is completed, the task packages that failed to be successfully allocated are unpacked and repacked to improve the matching rate and the flexibility of task allocation; The density ratio is obtained based on the predicted number of workers, including the ratio of the number of workers in each grid obtained according to the prediction to the number of tasks in each grid after the actual task is released, as the density ratio, wherein the density ratio is calculated by using the predicted number of workers in each grid and the actual number of tasks in each grid after the task is released. Calculate the value ,when It is a low ratio area. When is the middle ratio area, It is a high ratio area, where represents the predicted number of workers for a grid block, Represents the actual number of tasks within a grid block.

2. The method for regional classification in a crowdsourcing environment with multi-level task allocation according to claim 1 is characterized in that: The random forest model is used to predict the number of workers in each grid after grid division based on the historical data of crowdsourcing workers, including grid division of the entire perception area and using the random forest algorithm to predict the number of workers in each grid, where the attribute set of crowdsourcing workers is ,in Represent the horizontal and vertical coordinate positions of the workers, Represents the perceived quality ability level of crowdsourcing workers.

3. The method for regional classification in a crowdsourcing environment with multi-level task allocation according to claim 2 is characterized in that: The method of allocating the remaining tasks by recruiting across grid areas includes analyzing the grid where each unassigned task is located, detecting all directly adjacent grid areas, screening qualified workers in adjacent grids according to the urgency of the task, generating a set of optional workers for each task, and achieving global optimal allocation while maximizing the cost-effectiveness ratio of the objective function through a linear programming method.

4. A system for regional classification in a crowdsourcing environment with multi-level task allocation, which implements the method for regional classification in a crowdsourcing environment with multi-level task allocation as claimed in claim 1, characterized in that: include: The data acquisition module is configured to acquire historical data of crowdsourcing workers; The prediction module is configured to use a random forest model to predict the number of workers in each grid after grid division based on the historical data of crowdsourcing workers; The strategy module is configured to obtain a density ratio based on the predicted number of workers, divide the task area into low ratio, medium ratio and high ratio areas, wherein based on the cost-effectiveness maximization goal, a delay strategy is adopted in the low ratio area, a balance strategy is adopted in the medium ratio area, and a collaborative strategy is adopted in the high ratio area; The allocation module is configured to adopt cross-grid area recruitment for the remaining tasks to complete the task allocation.

5. A computer-readable storage medium storing a plurality of instructions, characterized in that: The instructions are suitable for being loaded by a processor of a terminal device and executing the method according to claim 1 .

6. A terminal device, comprising a processor and a computer-readable storage medium, wherein the processor is used to implement each instruction; and the computer-readable storage medium is used to store multiple instructions, characterized in that: The instructions are suitable for being loaded by a processor and executing the method as claimed in claim 1 .

Images