Spatial-Temporal Aware Truth Finding in Big Data Social Sensing Applications

This paper presents a spatial-temporal aware analytical framework to solve the truth finding problem in social sensing applications. Social sensing has emerged as a new big data application paradigm of collecting observations about the physical environment from social sensors (e.g., humans) or devices on their behalf. The collected observations may be true or false, and hence are viewed as binary claims. A fundamental challenge in social sensing applications lies in accurately ascertaining the correctness of claims and the reliability of data sources without knowing either of them a priori. This challenge is referred to as truth finding. Significant efforts have been made to address this challenge but two important features were largely missing in the state-of-the-arts solutions: when and where the claims are reported by a source. In this paper, we develop a new spatial-temporal aware truth finding scheme to explicitly incorporate the time information of a claim and location information of a source into a rigorous analytical framework. The new truth finding scheme solves a constraint optimization problem to determine both the source reliability and claim correctness. We evaluated the spatial-temporal aware truth finding scheme through both an extensive simulation study and a real world case study using Twitter data feeds. The evaluation results show that our new scheme outperforms all the compared state-of-the-art baselines and significantly improves the truth finding accuracy in social sensing applications.