A personalized recommendation framework with user trajectory analysis applied in Location-Based Social Network (LBSN)

There are several existing technologies to tracking down what favors the preferences or increase the satisfaction of a user in a social networking environment. These technologies range from the conventional manual approaches (with high human intervention) to automated approaches (e.g. vision-based, participatory sensing with mobile devices). In this paper, the user's trajectories were recorded with a Location-Based Social Network (LBSN) mobile application namely UniCAT, which provides several smart community services (e.g. information sharing, social networking, e-commerce functionalities) to its users. This paper proposes a personalized recommendation framework, which adopts the generic recommendation process with the integration of KDI (Knowledge-Desire-Intention) model in capturing the user's preferences. The proposed framework is evaluated with the trajectory records from 100 active users over a period of one year by recommending a list of Point-Of-Interests (POIs) during each user's request. The satisfactions of the generated POIs from various selected approaches are benchmarking with the standard information retrieval metrics of precision and recall. From the experimental results, the proposed hybrid approach outperformed other generic recommendation frameworks, and also proves that personalization can further improve user's experience and satisfaction.