2019/02/02 by Giannis Christoforidis, Pavlos Kefalas, Christoforidis, Giannis +5
Computer Science · Engineering · Social Sciences · #FOS: Computer and information sciences #Human Mobility and Location-Based Analysis #Information Retrieval (cs.IR) #Machine Learning (cs.LG) #Recommender Systems and Techniques #Transportation and Mobility Innovations
paper · pdf · doi:10.48550/arxiv.1902.00773
openalex publication_date 2019/02/02 · openalex created_date 2022/07/29 · openalex updated_date 2026/07/28
The rapid growth of users' involvement in Location-Based Social Networks\n(LBSNs) has led to the expeditious growth of the data on a global scale. The\nneed of accessing and retrieving relevant information close to users'\npreferences is an open problem which continuously raises new challenges for\nrecommendation systems. The exploitation of Points-of-Interest (POIs)\nrecommendation by existing models is inadequate due to the sparsity and the\ncold start problems. To overcome these problems many models were proposed in\nthe literature, but most of them ignore important factors such as: geographical\nproximity, social influence, or temporal and preference dynamics, which tackle\ntheir accuracy while personalize their recommendations. In this work, we\ninvestigate these problems and present a unified model that jointly learns\nusers and POI dynamics. Our proposal is termed RELINE (REcommendations with\nmuLtIple Network Embeddings). More specifically, RELINE captures: i) the\nsocial, ii) the geographical, iii) the temporal influence, and iv) the users'\npreference dynamics, by embedding eight relational graphs into one shared\nlatent space. We have evaluated our approach against state-of-the-art methods\nwith three large real-world datasets in terms of accuracy. Additionally, we\nhave examined the effectiveness of our approach against the cold-start problem.\nPerformance evaluation results demonstrate that significant performance\nimprovement is achieved in comparison to existing state-of-the-art methods.\n