Extending deep learning to frequent pattern and relationship mining in STDM

Determine how to extend deep learning methods—alone or integrated with frequent pattern mining and graphical models—to spatio-temporal data mining tasks beyond prediction and classification, specifically frequent pattern mining and relationship mining.

Background

Most deep learning applications in spatio-temporal data mining focus on tasks such as prediction and classification, where representation learning offers clear advantages. In contrast, tasks like frequent pattern mining and relationship mining typically do not center on learned features, and have seen little to no adoption of deep learning.

Bridging this gap requires principled methods that adapt or combine deep learning with established pattern and relational modeling techniques to address these underexplored spatio-temporal tasks.

References

So it remains an open problem that how deep learning models along or the integration of deep learning models with traditional models such as frequent pattern mining and graphical models can be extented to broader applications to more STDM tasks.

— Deep Learning for Spatio-Temporal Data Mining: A Survey  (1906.04928 - Wang et al., 2019) in Section VI, Open Problems (Broader applications to more STDM tasks)