用于空间众包任务匹配的未知工人效用估计方法

Utility Estimation of Unknown Workers for Task Assignment in Spatial Crowdsourcing

为了解决在未知工人效用情况下提高任务完成质量的问题,提出带有效用的最高分数匹配模型。该模型包含两个阶段:阶段一利用多臂老虎机模型计算工人效用值;阶段二利用更改了加分规则后的带有效用的基本方法(U-Basic)、带有效用的最小位置熵方法(U-LLEP)、带有效用的近距离优先方法(U-CDP)进行分配。在MovieLens和Gowalla真实世界数据集上的实验结果表明,所提方法与未使用效用的CDP和LLEP方法相比,一些评价指标有较大提升。

In order to improve the quality of task completion in the absence of worker utilities,a twostage maximum score assignment model with utility was proposed.Stage 1,the worker utility values were calculated using a multi-armed bandit model.Stage 2,they were distributed by using basic method with utility(U-Basic),minimum position entropy method with utility(U-LLEP)and proximity priority method with utility(U-CDP)with changed scoring rules.The experimental results on MovieLens and Gowalla real world data sets showed that some evaluation indexes were improved greatly compared with CDP and LLEP methods without utility.