Skipping class: Improving human-driven data exploration and querying through instances

Arash Saghafi, Yair Wand, Jeffrey Parsons

Research output: Contribution to journalArticleScientificpeer-review

Abstract

With the growing focus on business analytics and data-driven decision-making, there is a greater need for humans to interact effectively with data. We propose that presenting data to human users in terms of instances and attributes provides a more flexible and usable structure for querying, exploring, and analysing data. Compared to a traditional representation, an instance-based representation does not impose any predefined classification schema over the data when it is presented to users. This paper examines the potential utility of instance-based data through two laboratory experiments – the first focusing on exploration of data for pattern discovery (open-ended tasks) and the second on retrieval of information (closed-ended tasks). In both cases, participants were able to achieve better results in tasks using instance-based data than using class-based representations. Given the growing need for self-service analytics, as well as using information for purposes not anticipated when it was collected, we show that instance-based representations can be an effective way to satisfy the emerging needs of information users.
Original languageEnglish
Number of pages29
JournalEuropean Journal of Information Systems: An official journal of the Operational Research Society
DOIs
Publication statusE-pub ahead of print - Jan 2021

Keywords

  • Database Design
  • attributes
  • classification
  • database queries
  • human-in-the-loop data analytics
  • instance-based data model
  • knowledge discovery
  • non-classified data
  • open information
  • pattern detection

Fingerprint

Dive into the research topics of 'Skipping class: Improving human-driven data exploration and querying through instances'. Together they form a unique fingerprint.

Cite this