Year: 2019
Type: dissertation
Abstract: Advancements in data analytics research and the explosion of available data across public and private organisations have drastically improved the potential for deriving value from data.Thus, industry and research alike have long promoted the value of data for most parts of society.However, tapping into this pool of opportunity as an organisation is not necessarily easy since the adoption of data analytics often requires domain expertise and multiple collaborating groups of personnel.Machine learning and data mining techniques are already heavily used by dedicated data analysts, but it remains a challenge to incorporate existing knowledge capital and to communicate analytical findings.In the past decade, visual analytics research has therefore focused on the development of interactive technologies that combine automated data analysis and human knowledge using visualization interfaces.While this research has resulted in multiple important techniques for the analysis of large and complex data, there have been relatively little focus on supporting collaboration.However, experience suggest that collaboration is a central aspect in fulfilling the vision of involving real users in data analytics.Consequently, bridging this gap by augmenting visual analytics methods with technology that supports collaboration holds great potential for simultaneously advancing the adoption of data analytics in real-world domains and produce new research on the use of data analytics.This dissertation bridges the gap by presenting novel visual analytics approaches that enable domain experts to make sense of complex data and by developing interactive data analysis technologies that support aspects of collaboration.These aspects include sharing and re-composition of atomic data analysis components, synchronous interaction on multiple abstraction levels, and integrated data-driven reporting that points towards a new paradigm termed literate analytics.The work is based on collaborations with organisations from domains such as production, healthcare, and public business administration.Thus, working with real-world data and interacting with domain experts have been central components of the research approach.Among others, the contributions include novel methods for the analysis of temporal event sequence data and interactive systems that support collaborative visual analytics.As the types of data and the collaborative challenges that have been addressed transcend the concrete domains presented here, the contributions are also relevant in a more general perspective.Furthermore, the novel systems directly enable future research into different areas of collaboration in data analytics.An example of this is our own work on automatic distribution of visualization interfaces across multiple heterogeneous devices.i Resumé Grundet store fremskridt i forskning indenfor data analytics, samt eksplosionen af tilgængelige data på tværs af offentlige og private organisationer, er mulighederne for at skabe værdi med data øget drastisk.Erhvervslivet, såvel som forskningsmiljøet, har derfor længe promoveret værdien af data for de fleste dele af samfundet.Det er dog ikke ubesværet at udnytte disse muligheder i en organisation, fordi det ofte kræver domæneekspertise og flere samarbejdende personalegrupper at inkorporere dataanalyse som et centralt værktøj.Selvom machine learning og data mining allerede anvendes af dedikerede dataanalytikere, er det fortsat en udfordring at inkorporere eksisterende videnskapital og at kommunikere analytiske resultater.I det sidste årti har forskning indenfor visual analytics derfor fokuseret på at udvikle interaktive teknologier, der kombinerer automatiseret dataanalyse og menneskelig viden ved hjælp af visualiseringsgrænseflader.Selvom denne forskning har resulteret i vigtige teknikker til analyse af store og komplekse data, har der været et relativt lille fokus på at understøtte menneskeligt samarbejde.Erfaringer tyder dog på, at samarbejde er et centralt aspekt for at realisere visionen om at inddrage virkelige brugere i data analytics.Udvidelse af nuværende metoder indenfor visual analytics med interaktive teknologier der muliggør samarbejde, har derfor stort potentiale for at fremme anvendelsen af data analytics i virkelige domæner samt producere ny forskning i brugen af data analytics.Denne afhandling bygger bro mellem visual analytics og de nævnte udfordringer, ved at præsentere nye metoder, der gør det muligt for domæneeksperter at analysere komplekse data, samt ved udviklingen af interaktive teknologier der understøtter forskellige aspekter af samarbejde.Dette inkluderer at videregive og rekombinere atomare analysekomponenter, synkron interaktion på flere abstraktionsniveauer samt integreret datadreven rapportering, der peger imod et nyt paradigme, kaldet literate analytics.Forskningen er baseret på et samarbejde med forskellige organisationer fra domæner såsom produktion, sundhedspleje og offentlig virksomhedsadministration. Arbejdet med faktiske data og interaktionen med domæneeksperter har således været centrale komponenter i den anvendte tilgang.Afhandlingen bidrager blandt andet med nye metoder til analyse af temporale eventsekvenser og interaktive systemer, der understøtter collaborative visual analytics.Da de adresserede datatyper og udfordringer med samarbejde også kan findes i andre domæner, er bidragene således også relevante i et mere generelt perspektiv.Desuden understøtter de nye systemer direkte fremtidig forskning i adskillige områder indenfor samarbejde og avanceret dataanalyse.Et eksempel på dette er vores eget arbejde med automatisk distribution af visualiseringsgrænseflader på tværs af flere heterogene enheder.iii The last three years of my working life have been filled with both ups and downsluckily mostly ups.Which is probably to be expected in any PhD project.This dissertation would therefore not have been possible without the support from a lot of people.A sincere thank you to all the people who helped me throughout in one way or another.First of all, I want to thank my supervisor, Kaj Grønbæk, for creating unique opportunities, his continuous support and advice along the way, and for giving me the freedom to pursue any idea of mine.I also want to thank the co-authors of the publications that have contributed to the making of this dissertation.Thank you Matthias Nielsen for introducing me to the world of visualization when I first started in the group as a student programmer.Thank you Henrik Korsgaard for introducing me to the wicked world of CSCW.Thank you Clemens N. Klokmose and Roman Rädle for introducing me to the world of Webstrates and literate computing.Thank you Niklas Elmqvist for inspiring guidance on by now several papers.And thank you Sriram Karthik Badam and Tom Horak for fruitful collaborations, nerve-wracking paper talks, and multiple long discussions.A huge thanks to Niklas Elmqvist for hosting me on my stay abroad and for supervising me during this time.And also a huge thanks to Clemens N. Klokmose, Sriram Karthik Badram, and Tom Horak.You all made my stay in Maryland a truly great experience
Author: Mathisen, Andreas
Institution: Aarhus University
Language: English
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4412297719
