Topology-Inspired Algorithms and Visualisations for In-Depth Data Exploration

Abstract

The overarching goal of this dissertation was to improve, simplify, and extend data science algorithms to better support analysts in in-depth data exploration. We have explored this goal through the lens of topology-inspired algorithms and visualisations that support analysts discovering patterns that can solve their problems or answer their questions. As such, the research presented in this dissertation fits with the intelligence augmentation perspective ofcomputing that aims to enhance operators’ capability rather than replace them (Engelbart, 1962). The main contribution of this dissertation is a collection of algorithms developed to address specific problems or tasks related to identifying and interpreting patterns in complex and potentially unfamiliar data. The chapters in Part III expand upon the data science literature by presenting novel algorithms that provide additional functionality, improving existing functionality, or simplifying the use of existing techniques. We have published—or contributed to—open-source implementations of all presented algorithms, making our contributions readily available to data science practitioners.