报告信息：

讲座题目：Machine Learning for Spatial, Timely Food Security Mapping in Conflict Settings（冲突环境下粮食安全空间化、时效性测绘的机器学习方法）

主讲人：郭喆（国际食物政策研究所，高级研究员）

讲座时间：2025年10月16日 15:00-17:00

讲座地点：金码大厦7楼会议室

讲座概要：

Timely, granular mapping of food security is essential for policy and humanitarian action in fragile, conflict-affected settings, yet traditional surveys are costly, infrequent, spatially coarse, and often infeasible. We present a machine-learning approach that maps and monitors food security indicators (e.g., Food Consumption Score), producing near-real-time, spatially resolved estimates at local administrative-unit scales. The model combines multiple data sources—phone-survey data, crowd-sourced datasets, and public available GIS layers to generate operational assessments. In this seminar, we will cover model design and training, data pipelines, and strategies for scaling to other conflict-affected regions. We also discuss key challenges—data representativeness and bias, feature selection, uncertainty quantification, and validation against ground truth—and share lessons for improving predictive performance. We conclude with the broader implications of integrating machine learning with spatial data and survey data to support humanitarian targeting and policy decisions.

主讲人介绍：

郭喆研究员的研究聚焦于大数据分析、机器学习、深度学习与地理空间数据整合，旨在提升与人道主义领域的决策水平，其专长包括将人口健康调查（DHS）、生活水平测量研究（LSMS）等调查数据与空间数据集融合，开发人工智能驱动的粮食安全模型、冲突监测系统及气候风险评估体系，并主导实验数据采集与影响评估工作。同时负责机构数据库的管理、质量管控与成果传播，确保产出具有政策影响力的高质量数据。通过整合高级建模、遥感技术与数字化调查创新方法，郭喆研究员工作致力于为应对全球性挑战提供数据驱动的解决方案。

郭喆 讲座预告.docx