Research On Multi-Constrain Location Recommendation Model Based On Spatial-Temporal Data

Recent years, location based applications have generated a large amout ofspatial-temporal data, including location data and user data. The user data typically includescheck-in and trajectory data. Many scholars have studied spatial-temporal data based locationrecommendation model with two main directions including check-in based and trajectorybased models. For check-in based model, the user location preference model can be builtdirectly from the semantic information of check-in data while for trajectory based model asemantic enrichment process should be taken before to build the user location preferencemodel. The Spatial-temporal data based location recommendation model usually employscollaborative filtering recommendation algorithm to predict preferences, which considersopinions of other users, but needs to deal with sparsity and cold-start problem. In addition,some scholars studies knowledge based interactive location recommendation model, whichrecommends by constraints inputed by users through interface so it can reflect currentrequirements of users, but it needs to build knowledge preliminarily. Analysing the pros andcons of both models and determining the trade-offs, this article studies an integrated locationrecommendation model including three aspects: considering both spatial and target constraint;combing history and interactive preferences; using both check-in and trajectory data tocompute user location preferences. The main contribution of this article includes two parts asfollow.First, this article proposes an multi-contrain location recommendation model based onspatial-temporal data. Second, this article uses the hierarchical location semantic structure andlocation grid to unify the location semantic attribute of spatial-temporal datas and proposesarea profile and category profile based user location preference model.