Spatio-temporal forecasting in dynamic environments

The research of time series and spatio-temporal forecasting benefits a wide range of applications from weather forecasting to human mobility or traffic prediction. There are several possible project directions on spatio-temporal forecasting.

Project 1: NLPRoC for time series and spatio-temporal data

In this research, we will shape the forecasting problem from a whole new perspective of natural language generation. In the existing methods, the forecasting models take a sequence of numerical values as input and yield numerical values as output. Inspired by the successes of pre-trained language foundation models, in this project, we aim to investigate the adaptation of these models to time series forecasting tasks. This project will be building on top of our initial work in this line of study:

[1] Xue H, Salim FD, Ren Y, Clarke CL. Translating Human Mobility Forecasting through Natural Language Generation. In Proceedings of the Fifteenth ACM International Conference on Web Search and Data Mining (WSDM) 2022 (pp. 1224-1233).

[2] Hao Xue, Bhanu Prakash Voutharoja, and Flora D. Salim. 2022. Leveraging language foundation models for human mobility forecasting. In Proceedings of the 30th International Conference on Advances in Geographic Information Systems (SIGSPATIAL '22). Association for Computing Machinery, New York, NY, USA, Article 90, 1–9. https://doi.org/10.1145/3557915.3561026

Project 2: Spatio-temporal fusion for behaviour/trajectory prediction

The accurate prediction of pedestrian/vehicle trajectories and group behaviours can benefit many applications, ranging from crowd management, smart intersections, to intelligent transportation systems. To achieve a high prediction accuracy, in this project, we intend to explore how to leverage spatio-temporal data collected from multiple data sources. Our aim is to develop a framework that can efficiently and effectively fuse heterogenous spatio-temporal data to conduct the prediction task.  

Shao, W., Jin, Z., Wang, S., Kang, Y., Xiao, X., Menouar, H., Zhang, Z., Zhang, J. and Salim, F., 2022. Long-term Spatio-temporal Forecasting via Dynamic Multiple-Graph Attention, IJCAI 2022.

Project 3: Predict the Unpredictable: Spatio-temporal forecasting during extreme events

While tremendous progress has been made to model high-level spatio-temporal regularities with deep learning, most, if not all of the existing methods cannot be adapted to unprecedented volatility brought by potential societal events or extreme weather conditions. In this project, we aim to explore a robust spatio-temporal forecasting framework that can be well and quickly adapted when extreme events occur (e.g., bushfire, pandemic, flood). This framework will play a significant role in many downstream applications such as disaster management. We intend to investigate related techniques such as zero-shot learning and continual learning for the spatio-temporal forecasting models. This project will be building on top of our initial work in this direction:

Wang, Z., Jiang, R., Xue, H., Salim, F. D., Song, X., & Shibasaki, R. (2022, June). Event-aware multimodal mobility nowcasting. In Proceedings of the AAAI Conference on Artificial Intelligence (Vol. 36, No. 4, pp. 4228-4236).

Scholarship

• $37,684 per annum (2024 rate)

Eligibility

• Domestic applicants only
• PhD only

How to Apply

Email a copy of your CV, transcripts, and publications (if any) to flora.salim@unsw.edu.au.