We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Visualizing multi-dimensional decision boundaries in 2D.
- Authors
Migut, M.; Worring, M.; Veenman, C.
- Abstract
In many applications experts need to make decisions based on the analysis of multi-dimensional data. Various classification models can support the decision making process. To obtain an intuitive understanding of the classification model, interactive visualizations are essential. We argue that this is best done by a series of interactive 2D scatterplots. In this paper, we define a set of characteristics of the multi-dimensional classification model that have to be visually represented in those scatterplots. Our proposed method presents those characteristics in a uniform manner for both linear and non-linear classification methods. We combine a visualization of a Voronoi based representation of multi-dimensional decision boundaries with visualization of the distances of the data elements to these boundaries. To allow the developer of the model to refine the threshold of the classification model and instantly observe the results, we use interactive decision point selection on a performance curve. Finally, we show how the combination of those techniques allows exploration of multi-dimensional decision boundaries in 2D.
- Subjects
CLASSIFICATION algorithms; MULTIDIMENSIONAL databases; VISUALIZATION; DATA mining; IDA (Computer system); VISUAL analytics
- Publication
Data Mining & Knowledge Discovery, 2015, Vol 29, Issue 1, p273
- ISSN
1384-5810
- Publication type
Article
- DOI
10.1007/s10618-013-0342-x