Welcome
The Data Science for Spatial Systems (DSSS, CASA0006) module is a optional element of CASA’s MSc/MRes USS Course and is intended provide an introduction to advanced computational techniques in spatial analysis and spatial data sicence.
As with any computational analysis, spatial data science is built on two pillars: methods and data. In recent years, the rapid development of Artificial Intelligence (AI), particularly deep learning and large language models, has introduced new transformative techniques. These advancements make it possible to analyse large and heterogeneous datasets. For example, we can now combine census data, remote sensing images, street-view imagery, and text to understand complex issues like urban security or urban heat island effect, all on a standard laptop.
While it may seem that we can now model and predict almost anything, this is not the case. AI is not a silver bullet; it has limitations and can produce unreliable or ‘hallucinated’ results. Therefore, it is essential to understand the principles and limitations of these models. We must be critical of the methods we use, question the results, and connect our analysis to real-world applications that can inform policy and practice.
This module provides a solid foundation in geospatial machine learning, with a specific focus on supervised learning techniques and their real-world applications. While this is not a comprehensive overview of all data science, we believe that mastering supervised learning is the essential first step for students to build their skills in spatial data science.
The module is structured progressively across three parts:
- Part 1 (Weeks 1-3): Introduces the fundamentals of machine learning, including the core metrics and workflows of supervised learning.
- Part 2 (Weeks 4-6): Covers two key types of supervised learning: tree-based methods and neural networks.
- Part 3 (Weeks 7-10): Delves into advanced topics and practical challenges, such as model interpretation, feature selection, handling imbalanced data, and an introduction to Machine Learning Operations (MLOps).
Acknowledgements
We really appreciate the support from various people: