Location-based social networks (LBSNs) have witnessed a great expansion as an attractive form of social media. LBSNs allow users to “check-in” at geographical locations and share this information with friends. Indeed, with the spatial, temporal and social aspects of user patterns provided by LBSNs data, researchers have a promising opportunity for understanding human mobility dynamics, with the purpose of designing new generation mobile applications, including context-aware advertising and city-wide sensing applications. In this paper, the authors introduce a learning based random walk model (LBRW) combining user interests and “mobility homophily” for location recommendation in LBSNs. These properties are observed from a real-world Location-Based Social Networks (LBSNs) dataset. The authors present experimental evidence that validates LBRW and demonstrates the power of these inferred properties in improving location recommendation performance.