When User Interest Meets Data Quality: A Novel User Filter Scheme for Mobile Crowd Sensing

Mobile crowd sensing has become a promising paradigm for mobile users to collect information. Considering that the task information push is not free and there are many users who are not interested in the current task or provide noisy sensing data, one of the imminent problems is how to recommend high-quality and interested users in real time and steer participators to collect data with adequate budgets. However, it is difficult to predict the data quality and users' interest without the validity of real data. In this paper, we propose a user recommender system where the users' data qualities for sensing tasks are derived from historical statistical data to filter out the non-interested and malicious users in current task. The aim is to recruit a sub-group of participators for efficient crowd sensing, in order to maximize the platform utility. We show that our problem is NP-hard, and model the recruitment process as a sub-modular problem. Finally, an approximation algorithm is designed to guarantee the platform utility and participators' profits. We evaluate our algorithm on simulated data set and the results indicate that the platform utility and data quality improves significantly.