报告主题：Predicting Restaurant Survival in a Dynamic Catering Ecosystem: A Theory-Informed Multisource, Multimodal Interactive Fusion Approach

报  告 人：李恒云

报告时间：2026年10月20日（周二）下午4:00-5:00

报告地址：计算机大楼A318

报告摘要：

Restaurant closures are common, highlighting the need for reliable survival prediction. However, prior studies have paid limited attention to spatial interdependence among restaurants, multimodal customer feedback, and heterogeneous review attention. Using restaurant data from Hong Kong, this study proposes a theory-informed multisource and multimodal interactive fusion network (MMIFN) that integrates business supply, market demand, and macro-regional factors. The model uses dynamic graph neural networks with spatiotemporal attention to capture evolving geographical relationships among restaurants, applies cross-attention to fuse textual and visual review information, and incorporates customer- and macro-attention mechanisms to account for review salience and contextual moderation. Empirical results show that MMIFN outperforms existing models. Additional analyses reveal the importance of dynamic spatial relationships, bidirectional multimodal fusion, heterogeneous consumer attention, and geographical variation in restaurant survivability. The findings offer a new analytical framework for improving survival prediction and provide practical recommendations for enhancing the resilience and sustainable development of Hong Kong’s restaurant industry.

报告人简介：

李恒云，香港理工大学酒店及旅游业管理学院长聘副教授。主要研究方向为平台经济、中小企业生存与风险分析、人工智能与可解释人工智能、社交媒体及社会网络分析等。研究成果发表于旅游管理、工商管理及人工智能领域的国际顶级期刊和会议，Google Scholar 学术引用近8000次，H指数44，连续入选2024—2026年斯坦福大学与Elsevier发布的全球前2%顶尖科学家榜单。作为项目负责人或共同负责人累计获得近1600万港元科研资助，包括中国国家自然科学基金及香港研究资助局等科研项目。担任 Tourism Management、Annals of Tourism Research、Journal of Travel Research 等7部国际权威期刊编委，并受邀担任多个国际期刊客座编辑。曾获香港理工大学杰出青年研究学者奖、优秀早期职业研究者奖及英国萨里大学高级研究院学者奖等荣誉，研究成果曾获国际学术会议奖项或入围相关奖项，并被世界卫生组织、BBC等国际机构和媒体关注与报道。