The rapid growth of mobile phone technology and its combination with various technologies like GPS has added location context to social networks and has led to the formation of location-based social networks. In social networking sites, recommender systems are used to recommend points of interest (POIs) to users. Traditional recommender systems, such as film and book recommendations, have a long history. However, due to the existence of the location component and the physical connection of users with the outside world in social networking sites, several special features such as spatial, temporal, and social factors are considered to improve recommendations. Among the specific features of location-based social network data, spatial factor plays an important role in improving recommendations. Because peoplechr('39')s desire to visit places is greatly influenced by the distance between the person and the place. Also, the distribution of POIs in the region changes the pattern of user visits.
In the first part of this study, we discuss challenges which social networking sites may face by comparing location-based recommender systems with traditional recommender systems. In the following, we mention some important contexts and factors in POI recommendation. Spatial factor, social relations, different types of contents, different categories, sequential pattern, and time factor are contexts which are commonly used in POI recommendation. Next, we mention different types of location-based recommender systems: the fused model and the joint model. In the fused model we model user’s preferences and other additional contexts individually and after that, we combine their results with collaborative filtering. In a joint model, all contexts are learned Simultaneously.
In the next part, we discuss methods for extracting spatial context in location-based recommender systems. There are three major ways of modeling spatial data: independent, dependent, and restrictive models.
In independent modeling, we model spatial factor independently without considering the user’s preferences and other contexts. Here we discuss four basic independent models in detail: power law, Gaussian distribution, Kernel Density Estimation, and distance-based models. The power law is a relationship between two quantities in which a relative change in one quantity causes a change in another quantity, and this change is independent of the initial values of the two quantities. This rule is used for modeling spatial data in recommender systems. Changes in many natural quantities around a constant value follow the Gaussian distribution, and this has led to its use to model spatial factors. Kernel density estimation is a non-parametric method for estimating the probability density function of a random variable. To recommend personalized items this method can be very useful because we could model spatial data of every user individually. distance-based methods model spatial factor by considering the distance between users and items or items with each other.
At dependent modeling spatial context is learned with other contexts Simultaneously. For this, we determine four popular methods: matrix factorization, probability-based models, artificial intelligence, and combined models. These methods are general algorithms for recommending items in recommender systems and spatial factor is just one of their components. Restricted models filter recommendations by considering spatial constraints.
At the end of the article, we summarize the various features of the proposed methods and mention their advantages and disadvantages.rapid growth of mobile manufacturing technologies and its combination with various technologies have led to the addition of location dimension to social networks and the formation of location-based social networks. Recommender systems are used on location-based social networks to recommend points of interest to users. Traditional recommender systems such as movies and book recommendations have a long history. However, due to the locational component and physical connection of users with the outside world in location-based social networks, several specific features such as spatial, temporal, and social factors are considered to improve recommendations. Among the specific features of the location-based social network’s data, location factor plays an important role in improving recommendations. Because peoplechr('39')s desire to visit places is largely influenced by the distance between the person and the place. The distribution of attractive places in the area also changes the pattern of user visits. In this study, we first discuss the challenges that location-based social networks face by comparing them with traditional recommender systems. Next, the factors that influence location recommendation in location-based recommender systems are discussed in detail. Finally, a variety of location modeling methods, which is one of the most important factors in recommending attractive locations to users using location-based social network data, are discussed.