Year: 2026
Type: article
Abstract: ABSTRACT Point of interest (POI) recommendation services have gained widespread adoption in various consumer fields. However, existing POI recommendation methods often neglect the intrinsic logic of consumer behaviors and multimodal data fusion analysis. To fill this research gap, this paper proposes a multimodal fusion‐Based POI recommendation method with consumer equilibrium guidance. This method considers different data modalities such as graphs, time series, images, text, and so forth, and draws inspiration from consumer equilibrium and diminishing marginal utility theories to simulate the consumer purchasing decision process. Specifically, it is structured into five modules: demand analysis, information fusion, plan optimization, decision recommendation, and feedback. The core idea is to extract consumer behavioral features and POI geographical features through multimodal deep learning. And then these features jointly form the utility of POIs for users. By applying the theories of diminishing marginal utility and consumer equilibrium, we maximize the total utility of POI selection, generating Top‐ K recommendations. We also account for privacy concerns by employing differential privacy technique to protect users' historical behavior data. Experimental results on real‐world datasets demonstrate that our method outperforms state‐of‐the‐art methods, showing its potential and prospects in the POI recommendation domain.
DOI: 10.1002/cpe.70970
Authors: Chonghuan Xu, Zhaohui Wang, Kaidi Zhao
Institutions: Zhejiang Gongshang University, Fudan University
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
Award: G2988228719
OpenAlex ID: W7214447450