Scholarlink

<p>With the advancement of smart tourism and the rapid proliferation of digital travel platforms, accurately understanding and predicting tourist behavior has become a cornerstone of intelligent recommendation systems. However, tourist behavior data often exhibits high sparsity, temporal drift, and strong heterogeneity, which pose significant challenges to conventional recommendation algorithms. In this paper, we propose a novel deep learning-based framework, TDTSR (Time-aware Distributed Tourist Stream Recommendation), to address these limitations in real-world smart tourism platforms. TDTSR models user-item interactions as temporally evolving latent processes, incorporating both static and dynamic factors under a continuous-time probabilistic structure. A dual-stream variational inference network, based on gated recurrent units (GRUs), is designed to capture sequential dependencies and temporal preference drift. Furthermore, the model is deployed using a distributed training architecture that integrates TensorFlow, Apache Spark, and Hadoop, enabling scalable learning on large-scale GPU-CPU clusters. Extensive experiments on a public benchmark (MovieLens-25M) and a real-world tourism dataset demonstrate that TDTSR outperforms state-of-the-art methods in both prediction accuracy and training efficiency. The results also reveal the effectiveness of temporal dynamics modeling in capturing user behavioral patterns and enhancing recommendation quality.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Deep learning-driven tourist behavior prediction and intelligent recommendation system for smart tourism platforms

  • 期刊论文

  • 发表日期：

    2025年12月3日

查看全文
  • Lian Tang

摘要

With the advancement of smart tourism and the rapid proliferation of digital travel platforms, accurately understanding and predicting tourist behavior has become a cornerstone of intelligent recommendation systems. However, tourist behavior data often exhibits high sparsity, temporal drift, and strong heterogeneity, which pose significant challenges to conventional recommendation algorithms. In this paper, we propose a novel deep learning-based framework, TDTSR (Time-aware Distributed Tourist Stream Recommendation), to address these limitations in real-world smart tourism platforms. TDTSR models user-item interactions as temporally evolving latent processes, incorporating both static and dynamic factors under a continuous-time probabilistic structure. A dual-stream variational inference network, based on gated recurrent units (GRUs), is designed to capture sequential dependencies and temporal preference drift. Furthermore, the model is deployed using a distributed training architecture that integrates TensorFlow, Apache Spark, and Hadoop, enabling scalable learning on large-scale GPU-CPU clusters. Extensive experiments on a public benchmark (MovieLens-25M) and a real-world tourism dataset demonstrate that TDTSR outperforms state-of-the-art methods in both prediction accuracy and training efficiency. The results also reveal the effectiveness of temporal dynamics modeling in capturing user behavioral patterns and enhancing recommendation quality.