Task Assignment with Worker Churn Prediction in Spatial Crowdsourcing

Abstract

The pervasiveness of GPS-enabled devices and wireless communication technologies flourish the market of Spatial Crowdsourcing (SC), which consists of location-based tasks and requires workers to physically be at specific locations to complete them. In this work, we study the problem of Worker Churn based Task Assignment in SC, where tasks are to be assigned by considering workers' churn. In particular, we aim to achieve the highest total rewards of task assignments based on the worker churn prediction. To solve the problem, we propose a two-phase framework, which consists of a worker churn prediction phase and a task assignment phase. In the first phase, we use an LSTM-based model to extract the latent feelings of workers based on the historical data and then estimate the idle time intervals of workers. In the assignment phase, we design an efficient greedy algorithm and a Kuhn-Munkras (KM)-based algorithm that can achieve the optimal task assignment. Extensive experiments offer insight into the effectiveness and efficiency of the proposed solutions.