With the accelerating digital transformation of the cultural and tourism industry, banners, as the core visual vehicle for online traffic generation, must be designed to precisely match user aesthetic preferences with the needs of cultural and tourism scenarios. However, traditional manual design suffers from low efficiency and insufficient personalization. Existing recommendation systems often focus on content theme matching, neglecting the deep correlation between visual style (color, composition, elements) and user profiles. Furthermore, collaborative filtering algorithms limit recommendation accuracy in data-sparse scenarios. This paper’s structure and content: First, a multi-dimensional user profile is constructed (including cultural and tourism preferences, visual aesthetics, and behavioral habits). Second, a hybrid recommendation model is proposed that integrates improved collaborative filtering (introducing a user similarity weighting strategy) with visual style matching (using deep learning to extract banner visual features). Finally, the system’s functional modules (data collection, profile construction, model inference, and result output) are designed and a prototype system is developed. Experiments using banner design data from three cultural and tourism destinations show that activating this factor increases the system’s recommendation accuracy from 78.3% to 85.9% (a 7.6 percentage point increase) in holiday scenarios, and from 83.5% to 85.2% (a 1.7 percentage point increase) in non-holiday scenarios, effectively addressing the challenges of personalized and efficient cultural and tourism banner design.
Meng et al. (Thu,) studied this question.