Inferring Home Location from User's Photo Collections based on Visual Content and Mobility Patterns

Precise home location detection has been actively studied in the past few years. It is indispensable in the researching fields such as personalized marketing and disease propagation. Since the last few decades, the rapid growth of geotagged multimedia database from online social networks provides a valuable opportunity to predict people's home location from temporal, spatial and visual cues. Among the massive amount of social media data, one important type of data is the geotagged web images from image-sharing websites. In this paper, we developed a reliable photo classifier based on the Convolutional Neutral Networks to classify photos as either home or non-home. We then proposed a novel approach to home location prediction by fusing together the visual content of web images and the spatiotemporal features of people's mobility pattern. Using a linear SVM classifier, we showed that the robust fusion of visual and temporal feature achieves significant accuracy improvement over each of the features alone.