Forecasting Bilateral Refugee Flows with High-dimensional Data and Machine Learning Techniques*
收藏ICPSR2024-01-01 更新2026-04-16 收录
下载链接：
https://www.openicpsr.org/openicpsr/project/198322/version/V1/view?path=/openicpsr/198322/fcr:versions/V1/CODE/main_replication_file_revision_JoEG_2.R&type=file
下载链接
链接失效反馈官方服务：
资源简介：
We develop monthly refugee flow forecasting models for 157 origin countries tothe EU27, using machine learning and high-dimensional data, including digital tracedata from Google Trends. Comparing different models and forecasting horizons andvalidating them out-of-sample, we find that an ensemble forecast combining RandomForest and Extreme Gradient Boosting algorithms consistently outperformsfor forecast horizons between 3 to 12 months. For large refugee flow corridors, thisholds in a parsimonious model exclusively based on Google Trends variables, whichhas the advantage of close-to-real-time availability. We provide practical recommendationsabout how our approach can enable ahead-of-period refugee forecastingapplications.
提供机构：
Kiel Institute for the World Economy & Kiel University; UAB & BSE; Kiel Institute for the World Economy, Kiel University, and Institute for the Study of Labor; UAB, BSE, MOVE
创建时间：
2024-01-01
