Quality-Based User Recruitment in Mobile CrowdSensing

Mobile CrowdSensing has played an important role in our daily life. Data quality evaluation and user recruitment are both important problems in CrowdSensing. Recruiting users who will provide high-quality data guarantees the success of a task. In this paper, we will recruit users based on the data quality. First, by exploiting the historical data that users performed on tasks, we use Compressive Sensing(CS) to predict the data quality that a user will achieve on a task which he has never done before. By partitioning the matrix according to the similarity between users, we propose G(grouping) C(Compressive) S(Sensing). Compared with original CS, GCS is more efficient and has higher precision. Then, we use the predicted data quality to guide user recruitment. We consider a general scenario in the real world. For a task, we expect to use as short as possible time to achieve the expected quality. We both consider offline and online scenarios, and we design the greedy approximation algorithms Off-QBUR (Offline Quality-Based User Recruitment) with logarithmic approximation ratio and On-QBUR (Online Quality-Based User Recruitment) algorithm with linear approximation ratio respectively. We use a real-world dataset to evaluate the prediction of the data quality, and the experiment result shows that our method is efficient and can predict data quality with high precision.