Combining User Inherent and Contextual Preferences for Online Recommendation in Location-Based Services

Abstract

AbstractWith the development of mobile internet and smart devices, location-based services (LBS) have developed rapidly and attracted more and more users. The availability of a large amount of user interaction data makes it possible to provide more personalized and accurate recommendation services. However, in mobile scenarios, multiple influencing factors such as the diversity of user preferences, the variability of user behavior, and the dynamics of spatiotemporal contexts bring great challenges to recommendation services. To accurately capture the preferences of mobile users in dynamic contexts, we propose an Inherent and Contextual Preference-aware Attention Network (ICPAN) for online recommendation in location-based services. Our ICPAN consists of an inherent-preference mining module with self-attention layers, a contextual-preference perception module with improved IR\(^2\)-tree-based index structures, and an online recommendation module. The inherent-preference and contextual-preference models are trained based on global historical behavior data and instantly selected context-sensitive data, respectively. And then the online recommendation module uses attention aggregation to couple the two preference representations to generate the final recommendation result. Extensive experiments are conducted on three real datasets, and the experimental results show that the proposed ICPAN outperforms existing state-of-the-art methods.KeywordsLocation-based serviceRecommender systemSelf-attentionIR\(^2\)-tree