ICEWS

ICEWS使用混合方法进行不稳定性预测，将异构统计和基于代理的模型结合在集成框架中，总预测精度超过80%。这些模型是从100多个数据源以及250的国际和区域新闻源中近乎实时地提供的。数以百万计的新闻报道是由创新的，浅解析的Jabari技术和BBN的Serif NLP技术处理的。结果显示在交互式，可定制的基于web的门户中，该门户具有时间序列，基于地图的视图和其他视图。模型预测是通过各种可视化来呈现的，这些可视化支持向下钻取，以通过透明度来促进对模型的信任。

ICEWS employs a hybrid approach for instability prediction, combining heterogeneous statistical and agent-based models within an ensemble framework, achieving an overall prediction accuracy exceeding 80%. These models are made available in near-real time, drawing data from over 100 data sources and 250 international and regional news outlets. Millions of news articles are processed by the innovative shallow-parsing Jabari technology and BBN's Serif NLP technology. The analysis results are presented via an interactive, customizable web-based portal equipped with time-series visualizations, map-based views, and other viewing options. Model predictions are visualized through a range of visualizations that support drill-down capabilities, thereby fostering trust in the models through enhanced transparency.

搜集汇总

背景与挑战

背景概述

ICEWS是一个用于不稳定性预测的数据集，采用混合方法结合异构统计和基于代理的模型，预测精度超过80%。它基于100多个数据源和250个新闻源的实时新闻报道，使用Jabari和Serif NLP技术处理，并通过交互式Web门户提供可视化结果。