Authors
Simon Thorne, David Ball, & Zoe Lawson
Abstract
In this paper we present experimental data supporting an alternative approach to developing decision support spreadsheets using a Programming by Demonstration (PbD) paradigm. This technique is coined "Example Driven Modelling" and uses example data (attribute classifications) in combination with inductive machine learning to create decision support models as an alternative to spreadsheet programming.
In this experiment we examine whether participants can define attribute classifications ("example-giving") satisfactorily and describe benefits and limitations this method offers through statistical analysis of the experimental results. We then consider the wider implications of this research in traditional programming.
Sample
The user provides examples of the problem. The examples are formatted into a data set and processed by a machine learning algorithm.
The algorithm learns the relationships between the input and output patterns of the example. The learning algorithm can then generalize to new examples in the problem domain.
Publication
2013, International Journal of Human-Computer Interaction, Volume 29, Issue 1, pages 40-53.
Full article
Reducing error in spreadsheets: Example driven modelling versus traditional programming
