Providing recommendations on location-based social networks

• Published in: Journal of ambient intelligence and humanized computing Vol. 7; no. 4; pp. 567 - 578
• Main Authors: Kosmides, Pavlos, Demestichas, Konstantinos, Adamopoulou, Evgenia, Remoundou, Chara, Loumiotis, Ioannis, Theologou, Michael, Anagnostou, Miltiades
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.08.2016; Springer Nature B.V
• Subjects: Algorithms; Artificial Intelligence; Cluster analysis; Clustering; Collaboration; Computational Intelligence; Engineering; Machine learning; Neural networks; Original Research; Recommender systems; Robotics and Automation; Social networks; User Interfaces and Human Computer Interaction; Vector quantization; Probabilistic neural network; Location recommendations; Location-based social networks; Neural networks; Clustering based prediction
• ISSN: 1868-5137; 1868-5145
• DOI: 10.1007/s12652-016-0346-7

During the last decade, in parallel with the rapid growth of mobile communications and devices, location-based social networks have met a tremendous growth with the acceptance of the public being constantly increasing. Users have access to a plethora of venues and points of interest, while they are able to share their visits to various locations along with comments and ratings about their experience (a process which is often referred to as “check-ins”). Location recommendations based on users’ needs have been a subject of interest for many researchers, while location prediction schemes have been developed in order to provide user’s possible future locations. In this paper, we present a novel method for predicting a user’s location based on machine learning techniques. In addition, following the incremental trend towards data accumulation in social networks, we introduce a clustering based prediction method in order to enhance the recommender system. For the prediction process we propose a probabilistic neural network and confirm its superior performance against two other types of neural networks, while for the clustering process we use a K-means clustering algorithm. The dataset we used was based on input from a well-known location-based social network. Prediction results can be used in order to make appropriate suggestions for venues or points of interests to users, based on their interests and social connections.