📄 中文摘要
失踪儿童调查的前72小时对于成功找回至关重要。然而，执法机构常常面临数据碎片化、结构不清晰以及缺乏动态地理空间预测工具的问题。Guardian系统提供了一种端到端的决策支持系统，旨在支持失踪儿童调查和早期搜索规划。该系统将异构、非结构化的案件文档转换为符合模式的时空表示，结合地理编码和交通背景丰富案件信息，并提供覆盖0-72小时的概率搜索产品。该研究详细描述了Guardian系统的三层预测组件，其中第一层为马尔可夫链，构成了系统的基础。
可解释的基于马尔可夫的时空风险模型用于失踪儿童搜索规划，结合强化学习和基于大语言模型的质量保证
📄 English Summary
Interpretable Markov-Based Spatiotemporal Risk Surfaces for Missing-Child Search Planning with Reinforcement Learning and LLM-Based Quality Assurance
The first 72 hours of a missing-child investigation are crucial for successful recovery. However, law enforcement agencies often encounter fragmented and unstructured data, along with a lack of dynamic geospatial predictive tools. The Guardian system offers an end-to-end decision-support framework for missing-child investigations and early search planning. It transforms heterogeneous, unstructured case documents into a schema-aligned spatiotemporal representation, enriches cases with geocoding and transportation context, and delivers probabilistic search products for the critical 0-72 hour window. A detailed description of the system's three-layer predictive component is provided, with the first layer based on a Markov chain, forming the foundation of the system.