User Identity Linkage across Location-Based Social Networks with Spatio-Temporal Check-in Patterns

Driven by the tremendous amount of spatio-temporal data obtained from location-based social networks, the implement of cross-domain user linkage, also known as the User Identity Linkage (UIL), brings about abundant promising research and application prospects. User distinctive behavior patterns implicit in the "check-in" spatio-temporal data provide a utility and meritorious way for UIL research. Enlightened by the fundamental weakness of existing algorithms which discretize the spatio-temporal sparse data with continuous nature, we propose a pertinently approach CP-Link to complete UIL task by exploiting user behavior pattern in a continuous way, where the continuous space is divided into irregularly shaped stay regions and a continuous time-based IDTW method is utilized to calculate the similarity. To bridge the gap between the theoretical ideal model and the actual sparse data, we apply user-associated location frequent pattern (LFP) model to supply the sparse deficiency. Ultimately, extensive experiments on real-world datasets are conducted to demonstrate the superiority and stability of our proposed CP-Link, which outperforms state-of-the-art by more than 20% in terms of an AUC increase.