How different functional regions in urban space are distributed and dynamically changing is determined by how their residents interact with them, which is crucial for urban managers to make urban planning decisions, respond to emergency quickly. Based on these, this paper proposed a novel approach for the probability based labelling individual activities which can be further used to explore the distribution of social land use at base tower station (BTS) level using a combination of multi-source spatiotemporal data, namely, call data and checkin data. We applied an experiment in Shenzhen, China, and the result is compared to Tencent Street View to demonstrate the effectiveness of the proposed approach to infer urban functional regions.