Addressing Challenges in Spatial and Temporal Tasks

PhD Thesis Proposal Defence


Title: "Addressing Challenges in Spatial and Temporal Tasks"

by

Miss Jia LI


Abstract:

With the rise of data-collection technologies and tools, we are accumulating 
vast data on human society and the natural world, e.g., financial transactions, 
traffic flow, precipitation, and air quality. The data collected is inherently 
spatial and temporal, as it is associated with the spatial location and the 
time of observation. Effective analysis of such data fosters an enhanced 
understanding of the evolving physical world around us.

In this thesis proposal, we are motivated by real-world applications and 
develop data-driven methodologies for spatial and temporal mining. 
Specifically, this proposal studies two types of tasks, i.e., time series 
anomaly detection and spatial interpolation. The first work aims to discover 
anomalies during each temporal data collection process. We proposed FluxEV, a 
fast and effective unsupervised anomaly detection framework for time-series 
data. By converting data with periodic patterns into a stationary distribution 
and expanding the gap between normal and abnormal values, FluxEV significantly 
improves detection accuracy. Moreover, the Method of Moments is adopted to 
speed up the parameter estimation in the automatic thresholding. Extensive 
experiments show that FluxEV greatly outperforms the state-of-the-art baselines 
on two large public datasets while ensuring high efficiency. In the second 
work, we aim to infer fine-grained spatial information using the observation 
data to address the data sparsity of limited data collection devices. The 
existing spatial interpolation methods rely on some unrealistic pre-settings to 
capture spatial correlations, which limits their performance in real scenarios. 
To tackle this issue, we propose the SSIN, which is a novel data-driven 
self-supervised learning framework for spatial interpolation by mining latent 
spatial patterns from historical observation data. Inspired by the Cloze task, 
we fully consider the characteristics of spatial interpolation and design the 
SpaFormer model based on the Transformer architecture as the core of SSIN. The 
experimental results on real-life raingauge and traffic datasets verify the 
effectiveness and generality of our proposed solution.


Date:                   Friday, 22 March 2024

Time:                   2:00pm - 4:00pm

Venue:                  Room 4472
                        Lifts 25/26

Committee Members:      Prof. Lei Chen (Supervisor)
                        Prof. Xiaofang Zhou (Chairperson)
                        Prof. Ke Yi
                        Prof. Qiong Luo