Personalized tourism recommendation model based on temporal multilayer sequential neural network.

Area of Science:

• Computer Science
• Artificial Intelligence
• Tourism Informatics

Background:

• Personalized tourism route recommendations face challenges due to large spatiotemporal data spans.
• Existing systems struggle to integrate historical data, user preferences, and real-time conditions effectively.
• High relevance and accuracy are crucial for advanced personalized travel systems.

Purpose of the Study:

• To propose a novel personalized tourism route recommendation model, the Temporal Multilayer Sequential Neural Network (TMS-Net).
• To address the challenges posed by large temporal and spatial spans in tourist trajectory data.
• To enhance the accuracy and relevance of personalized route recommendations.

Main Methods:

• Developed TMS-Net with adaptive trajectory segmentation to manage spatiotemporal data complexity.
• Integrated a self-attention mechanism with relative positional information to capture path relationships.
• Utilized a multilayer Long Short-Term Memory network for deep temporal dependency modeling in travel routes.
• Trained the model on over six million trajectory data points from Chengdu City (2016-2022).

Main Results:

• Optimal trajectory segmentation interval identified as 0.8 to 1.2 hours.
• Achieved a recommendation accuracy of 88.6%.
• Recorded a Haversine distance error of 1.23, indicating precise point-of-interest identification.
• Demonstrated effective recognition of long-duration travel behaviors.

Conclusions:

• TMS-Net significantly improves personalized tourism route recommendations by accurately identifying points of interest.
• The model offers new methodological insights for personalized travel recommendation systems.
• Adaptive segmentation and deep temporal modeling are key to enhancing recommendation relevance and accuracy.