CN-122022345-A - Fair space crowdsourcing task recommendation method based on competition balance

CN122022345ACN 122022345 ACN122022345 ACN 122022345ACN-122022345-A

Abstract

The application discloses a fair space crowdsourcing task recommendation method based on competition balance, relates to the technical field of space crowdsourcing task recommendation, and particularly relates to a fair space crowdsourcing task recommendation method based on competition balance, which comprises a task recommendation, worker selection and task distribution three-stage closed-loop flow, and a multi-stage probability recommendation, supply and demand flow balance and fairness perception distribution mechanism are integrated. The application can obviously improve the task allocation success rate and the opportunity fairness among workers while ensuring the overall expected profit of the platform.

Inventors

ZHENG KAI
ZHAO YAN
CHEN JINWEN
MIAO HAO

Assignees

电子科技大学长三角研究院(衢州)

Dates

Publication Date: 20260512
Application Date: 20260202

Claims (8)

1. A fair space crowdsourcing task recommending method based on competition balance is characterized by comprising the following steps: A task recommending step of determining and recommending a recommended task set from an effective task set of each worker in a plurality of workers, wherein the determining of the recommended task set comprises the steps of constructing initial worker-task weights based on the effect which can be generated by the workers executing the task and the probability of the workers selecting the task; A worker selecting step of receiving an intention task subset autonomously selected by each worker from a recommended task set recommended by the worker; And the task allocation step is to solve the conflict based on a fairness-aware allocation algorithm to execute final task allocation when a plurality of workers select the same task to generate conflict, wherein the fairness-aware allocation algorithm dynamically adjusts allocation weights according to the historical allocation success rate of each worker, and workers with low historical allocation success rates are given higher allocation weights.
2. The fair space crowdsourcing task recommending method based on the competition balance according to claim 1, wherein the task recommending step further comprises a supply and demand balancing sub-step of: Identifying a first type of area with the number of workers being greater than the number of tasks and a second type of area with the number of tasks being greater than the number of workers; Constructing a flow network pointing from the first type area to the second type area, and solving a minimum cost maximum flow problem to determine a transfer scheme of a worker from the first type area to the second type area; Determining a corresponding recommendation preference area for each worker based on the transfer scheme; and when the recommended task set is determined, improving the tendency that the task in the recommended preference area corresponding to the worker is recommended to the worker.
3. The fair space crowdsourcing task recommendation method based on competition balance of claim 1 wherein the initial worker-task weight The calculation is as follows: Wherein, the Representing workers Completing a task The profit to be generated is that, Is a worker Is used for the speed of the (c) in the (c), Is the distance from the worker's location to the task's origin, Is the distance from the start of the task to the destination, Is a worker Selecting tasks Is a probability of (2).
4. The fair space crowdsourcing task recommending method based on the competition balance according to claim 3, wherein the worker is Selecting tasks Probability of (2) By exponential distribution Modeling is performed in which Is a parameter.
5. The fair space crowdsourcing task recommending method based on competition balance as set forth in claim 1, wherein in the multi-stage iterative process, the first step Worker-task weights for multiple iterations The calculation is as follows: Wherein, the Representing workers The probability of rejecting all of the previously recommended tasks, Indicating no worker selection task Is a function of the probability of (1), Weights for the initial worker-task.
6. The fair space crowdsourcing task recommending method based on the competition balance according to claim 1 or 5, wherein the optimal matching is realized through a Hungary algorithm.
7. The fair space crowdsourcing task recommending method based on the competition balance as set forth in claim 1, wherein the worker is Historical allocation success rate The calculation formula of (2) is as follows: Wherein, the Is a worker The number of times a task was successfully assigned, Is a worker The number of tasks but not assigned is selected.
8. The method for recommending fair space crowdsourcing tasks based on competition balance according to claim 7, wherein the dynamic allocation weights used in the fair sensing allocation algorithm The calculation formula of (2) is as follows: Wherein, the For the initial worker-task weight, Success rates are assigned to the average history of all workers.

Description

Fair space crowdsourcing task recommendation method based on competition balance Technical Field The invention relates to the technical field of space crowdsourcing and intelligent task recommendation, in particular to a fair space crowdsourcing task recommendation method based on competition balance. Background With the widespread popularity of wireless edge devices (such as smartphones), space crowdsourcing has rapidly evolved in the field of urban services and data management. Typical application scenarios include network about platform (e.g., drop out), instant delivery service (e.g., beauty group delivery), etc., with the core mechanism being to distribute tasks with geographic location attributes (e.g., pick up passengers, deliver meals or packages) to a wide range of individual crowd-sourced workers. To improve quality of service and task completion rate, platforms typically need to intelligently recommend task sets to workers that match their location, capabilities, and preferences. However, if the recommended result deviates seriously from the actual intention of the worker, the recommended result may cause a negative response or even exit from the order, thereby significantly reducing the overall task completion efficiency and user experience. The current mainstream space crowdsourcing task recommendation method mainly focuses on optimizing preference utility, geographic coverage rate or diversity of a task recommendation set, but generally ignores a key reality factor, namely vigorous competition among workers after task recommendation. In actual operation, the platform often adopts a 'push-select' strategy, namely the same task is simultaneously pushed to a plurality of adjacent workers, and only one person finally receives a bill. Although the mechanism is helpful for improving the matching efficiency, frequent falling selection is easy to cause frustration of workers, long-term accumulation can cause loss of high-value workers, and further the ecological stability and profitability of the platform are weakened. The prior art reveals several key drawbacks in this context. First, most studies fail to model efficiently and maximize the overall profit of the system, since they tend to focus only on static indicators (e.g., coverage) of the recommendation phase, ignoring the high uncertainty introduced by subsequent worker autonomous selection actions, making it difficult to accurately evaluate and optimize the actual profit. Second, it is very challenging to increase the dispensing success rate (i.e., the ratio of the number of times a worker is successfully dispensed with tasks to the number of times he selects tasks) while maintaining a high profit. Excessive recommendation can exacerbate internal competition, lower the allocation success rate and impair worker satisfaction, while insufficient recommendation can result in unmanned task response and affect platform operation efficiency. Therefore, how to balance the task completion by stimulating moderate competition and the fair sense by avoiding excessive competition is a core problem to be solved. Third, while some studies have attempted to introduce fairness considerations, they have typically only used inter-worker profit differences as fairness metrics, ignoring the more essential fairness dimension in the task recommendation scenario-i.e., the uniformity of opportunity acquisition versus allocation success rate. The narrow fairness view is difficult to truly reflect the perception of the fairness of the system by workers, and internal contradictions caused by uneven resource contention cannot be effectively relieved. In summary, the existing space crowdsourcing task recommendation method has obvious defects in terms of considering platform profit, allocation success rate and multidimensional fairness, and a novel recommendation method capable of cooperatively optimizing competition balance, opportunity fairness and economic benefits is needed. Disclosure of Invention Aiming at the technical problems of unfair task allocation, low task completion rate, impaired platform profit and the like caused by ignoring competition relations among workers commonly existing in the existing space crowdsourcing task recommendation system, the invention provides a fair space crowdsourcing task recommendation method based on competition balance. The prior art generally focuses only on preference matching, geographic coverage or diversity optimization between tasks and workers, and does not adequately model competing conflicts and their negative impact on allocation success rate and worker satisfaction after tasks are simultaneously selected by multiple workers. In addition, conventional fairness mechanisms often use profit differences as the only measure, and fail to effectively reflect the balance of task acquisition opportunities. Therefore, the invention constructs a closed-loop task recommending and distributing framework integrating three mechanisms of