图注意力网络融合地理知识图谱的位置预测OA

To address the limitations of existing location prediction methods in geographic semantic fu-sion and users'geographic preference modeling,a graph attention network method for location predic-tion enhanced by a geographic knowledge graph is proposed.A hierarchical geographic knowledge graph with multi-source geographical data is constructed to map users,points of interest,and multi-granularity geographic entities into a unified semantic space.On this basis,a spatial and relation-aware graph attention network together with a spatiotemporal-aware sequential encoder is employed to model users'global geographic preferences and local spatiotemporal transition patterns,respectively.Experiments on the Foursquare-NYC and Foursquare-TKY datasets show that the mean reciprocal rank(MRR)of the proposed method are improved by 19.52%and 4.10%,respectively,compared with those of the strongest baseline.Experimental results demonstrate that the proposed method can effectively improve the accuracy of next location prediction.