What is it about?
Do you want to know how the neighbors' correlation will change over time? Well, this research proposes DAMR to represent the dynamic spatial correlation (the correlation between weather monitoring station A and station B may change over time). Furthermore, the multivariate time series imputation task is conducted by stacking the graph learning layer with DAMR. This research can also be used in prediction task.
Featured Image
Photo by Aldrin Rachman Pradana on Unsplash
Why is it important?
Multivariate time series prediction/imputation is vital for weather forecasting, traffic flow prediction, etc. The challenge here lies in the modeling difficulty of dynamic spatial correlations. Our DAMR structure can handle imputation tasks by considering different patterns of spatial correlations, which will be really close to the actual scenario.
Perspectives
It will be enjoyable to see the application of DAMR in many industrial areas. I am looking forward to seeing many opportunities and cooperation.
Dr Xiaobin Ren
University of Auckland
Read the Original
This page is a summary of: DAMR: Dynamic Adjacency Matrix Representation Learning for Multivariate Time Series Imputation, Proceedings of the ACM on Management of Data, June 2023, ACM (Association for Computing Machinery),
DOI: 10.1145/3589333.
You can read the full text:
Contributors
The following have contributed to this page
