Recently, social ties inferring in spatiotemporal data has attracted widespread attentions. Previous studies, which focused on either co-occurrence or context, do not fully exploit the information of spatiotemporal data. In order to better use the spatiotemporal information, in this paper we introduce two novel co-occurrence feature, namely, topic co-occurrence feature and context co-occurrence feature. The former feature is extracted by the topic model on carefully constructed bag-of-words. The latter feature is extracted by natural language processing tools on carefully constructed context sequence, which considers context, co-occurrence and mobility periodicity simultaneously. These two novel co-occurrence feature are both based on time and space perspectives. Then we infer social ties from these multi-view co-occurrence feature (including baseline co-occurrence, topic and context co-occurrence). The experiments demonstrate that the two novel co-occurrence feature contribute to the social tie inferring significantly.
展开▼
