Location-based social networks (LBSNs), e.g., Foursquare, Gowalla and Yelp, bridge the physical world with the virtual online world. In an LBSN, users can establish social links with each other and share their experiences of visiting or checking in some specific locations, known as points-of-interest (POIs), e.g., restaurants, stores and museums. LBSNs have accumulated plenty of community-contributed data such as social links between users, check-ins of users on POIs, geographical information and categories of POIs, which reflect the preferences of users to POIs. Recommending users with their preferred POIs has become an important feature for LBSNs, which benefits people to explore new places and businesses to discover potential customers. This thesis aims to recommend personalized POIs for users based on their preferences that are learned from the community-contributed data in LBSNs. To this end, this thesis intensively takes full advantage of the five major characteristics that affect the visiting preferences or check-in behaviors of users to POIs and are implied in the community-contributed data. (1) Geographical influence. The geographical proximity of POIs significantly affects users’ check-in behaviors. Current studies model the geographical influence on all users’ check-in behaviors as a universal way. However, the geographical influence on users’ check-in behaviors should be personalized. Therefore, this thesis models the personalized geographical influence in POI recommendations through estimating a personal geographical distribution of check-in POIs for each user. (2) Social influence. In terms of the fact that users with social links are more likely to share common interests on POIs, this thesis aggregates the check-in frequency of a user’s friends on a POI and models the social check-in frequency as a power-law distribution to employ the social correlations between users in POI recommendations. The proposed method is greatly superior to the social recommendation techniques that simply utilize the social links to derive the similarity of users. (3) Categorical influence. The categories of POIs indicate their usual business activities and nature, and people have different biases on the categories of POIs. Hence, this thesis applies the bias of a user on a POI category to weigh the popularity of a POI in the corresponding category and models the weighed popularity as a power-law distribution to leverage the categorical correlations between POIs for recommendations. The proposed method is much better than the existing techniques that separately utilize the category and popularity information of POIs. (4) Sequential influence. Based on the fact that human movements exhibit sequential patterns, current studies utilize the sequential influence based on the first-order Markov chain that only considers the latest visited POI of a user to estimate the probability of the user visiting a new POI. Nevertheless, in reality such an estimated visiting probability depends on not only the latest visited POI but also the earlier visited POIs of the user. Thus, this thesis investigates the high-order sequential influence by developing an efficient nth-order additive Markov chain in POI recommendations. (5) Temporal influence. Time is a very important factor influencing users’ check-in behaviors. Current studies use the temporal influence to recommend time-aware POIs by dividing users’ check-in POIs into time slots based on their check-in time and learning their preferences to POIs in each time slot separately. Unfortunately, these studies generally suffer from the loss of time information because of dividing a day into time slots and the lack of temporal influence correlations due to modeling users’ preferences to POIs for each time slot separately. Therefore, this thesis correlates the continuous temporal influence at different times for time-aware POI recommendations by estimating a time probability density of a user visiting a new POI. The experimental results on real-world large-scale check-in data sets show that the proposed POI recommendation techniques achieves significantly superior recommendation performance compared to other state-of-the-art POI recommendation methods.