Progressive information technologies resulting from the evolution of location-based services (LBS) have greatly improved people's urban lives. Location-based social networks (LBSNs) provide platforms for users to check in and share their current locations, thoughts, experiences, and reviews of points of interest (POIs) with anyone. This huge amount of heterogeneous data in LBSNs has enabled the development of POI recommendations. It has attracted many efforts in the research community to develop accurate POI recommender systems in various scenarios such as mobile, automotive, and enterprise applications. As our focus is on automotive scenarios and recent modern automotive driver information systems, there is a large volume of data available to the driver, such as digital transmission information, global positioning system (GPS) information and vehicle application information. If all this data is provided raw to the driver, information overload becomes a significant problem. As such, the POI recommendation service is particularly suitable for mobility applications. Say no to plagiarism. Get a tailor-made essay on "Why Violent Video Games Shouldn't Be Banned"? Get an original essayFor example, it can reduce the risk of traffic accidents by avoiding entering long location names when users search for places to go. Therefore, it is not only useful for users to discover new locations easily, but also helps users get relevant POIs without spending much time searching, especially when they are in a new area. In past research, common problems faced in POI recommender systems are cold start and data sparsity. The cold start problem is caused by the limited history of user activities and locations in the system because for a new user or location, the recommendation model does not have enough information to provide useful recommendations. Due to the rapid growth of new users on LBSN, the problem becomes even worse. Similarly, data sparsity is due to the fact that the total data in the recommendation model is not enough for processing and recognizing related users/items. Therefore, some hybrid approaches and new methods that consider different types of recommendation models are needed.