Mining Semantic Sequential Patterns from Geo-tagged Photos

Social media data associated with geographic location and time information reflect people footprint in real world. Abundance of geo-referenced content represents a massive opportunity to understanding of human geographic mobility behaviors. Most trajectory mining research from geo-enabled social media data focus on spatial geometric features. Integrating trajectory analysis with semantic information can implicate human movement behaviors on semantic levels. In this work, we illustrate a study on mining semantically enriched trajectory patterns using georeferenced content especially using geo-tagged photo data for case study. We first propose a semantic region of interest mining technique to extract reference regions with semantic information. We then present a multi-dimensional sequential pattern mining algorithm to find trajectory patterns with various semantic dimension combinations. We apply our method to real geo-tagged photo data to discover interesting patterns about sequential movement related to multiple semantics. Experimental results show that our method is able to find useful semantic trajectory patterns from geo-tagged content and deal with multi-dimensional semantic trajectories.