A general-purpose tourism intent understanding and route planning based on multimodal spatio-temporal pre-training

Accurately understanding tourists' travel intentions and planning routes has become a key issue in smart tourism.Traditional methods rely on a single data modality and are unable to cope with multi-dimensional user demands in complex temporal and spatial contexts.Therefore, this paper first explicitly models the multi-modal spatio-temporal relationships by constructing timestamp-aligned heterogeneous graphs, and uses graph convolution networks for semantic interaction.A dynamic weight mechanism based on the Pearson correlation coefficient is introduced to optimise the fusion process of cross-modal features, ultimately achieving the precise identification of travel intentions.Finally, guided by intent understanding results, a multi-objective tourism route planning method is designed.Combining heuristic search and evolutionary strategies, an optimal and balanced route plan is generated.Experimental results show that the recognition accuracy of the proposed method reaches 93.51%, and the reverse generation distance of the route planning is only 1.9, demonstrating strong generalisation ability and practical value.