Transient Group Recommendation for Shared Environment using Check-in and Tag Data

Group Recommendation Systems (GRSs) are crucial in facilitating decision-making processes for shared playlists. However, recommending appropriate items is challenging when it comes to dynamic and transient groups such as restaurant customers. Thus, we proposed a hybrid GRS designed to address these difficulties. By combining collaborative filtering and content-based recommendations enriched with tag information, the proposed system used historical and current check-in data to predict future group composition and tailor recommendations accordingly. The system generated top-N music playlists for every period of the week and provided diverse and novel recommendations prepared before customers’ arrival. We enhanced recommendation performance by extending the traditional user-item matrix with tags and combining memory-based and model-based computation. The proposed method was tested with two real-world datasets. The result showed the improvement of the performance over popularity-based item selection, offering a promising approach for transient group recommendation in a shared environment.