Abstract: [Purposes]Region-of-interest(ROI)recommendation is a critical application within location-based social networks.Addressing the limitations of existing methods in adequately expressing diverse inter-regional correlations and mining implicit social relationships,this paper proposes an urban ROI recommendation method based on adaptive anchor graph representation learning.[Methods]Specifically,we leverage dual-view constraints—incorporating geospatial structure and functional semantics—to optimize an adaptive anchor graph.A node-anchor message passing mechanism is then designed to capture complex inter-regional correlations,thereby generating robust regional preference features.Furthermore,we mine implicit social associations based on the overlap of user check-in regions and the similarity of activity times to extract implicit social preference features.Finally,these two types of features are fused to predict user preferences for specific regions.[Findings]Extensive experiments on the Sina Weibo and Foursquare datasets demonstrate that the proposed method achieves significant improvements in terms of recall and normalized discounted cumulative gain (NDCG).[Conclusions]This study enables personalized decision support for LBS applications including smart tourism and business management．

Key words: region-of-interest, recommendation, anchor graph, graph representation learning, location-based social network