Event Scale Forecasting


Due to the potentially significant benefits for society, forecasting spatio-temporal societal events is currently attracting considerable attention from researchers. Beyond merely predicting the occurrence of future events, practitioners are now looking for information about specific subtypes of future events in order to allocate appropriate amounts and types of resources to manage such events and any associated social risks

More formally, the problem can be formulated as a multi-class classification problem where the input can be a high-dimensional feature vector while the output is an categorical data.

Air Pollution Dataset

The goal is to predict the primay pollutant in different loations.


Download link: [air_pollution]

Data format: See the readme.txt enclosed for details.

Data Source is arranged and processed from the air quality official website in China: https://www.aqistudy.cn/historydata/.

Civil Unerest Dataset

The goal is to predict the event types of civil unrest movements, including empolyee strickes, student protests, farmer boycott, etc.

Download link: [civil_unrest]

Data format: See the readme.txt enclosed for details.

Data Source:

All the civil unrest tweet messages X, label set Y, and keywords are obtained from IARPA OSI project. Please refer to the papers [KDD 14] and [KDD 16] for details. The raw label set can be downloaded here: [Output Raw Data].


To use these datasets, please cite the papers:

Yuyang Gao, Liang Zhao, Lingfei Wu, Yanfang Ye, Hui Xiong, Chaowei Yang. Incomplete Label Multi-task Deep Learning for Spatio-temporal Event Subtype Forecasting.Thirty-third AAAI Conference on Artificial Intelligence (AAAI 2019), (acceptance rate: 16.2%), Hawaii, USA, Feb 2019, to appear.



NSF (sole-PI): III: CAREER: Spatial Network Deep Generative Modeling, Transformation, and Interpretation, $549,656, 2020-2025, National Science Foundation.