Model Building in Visualization Space

Ove Daae Lampe, Helwig Hauser

INPROCEEDINGS, Proceedings of Sigrad 2011 , 2011

• paper
• presentation
• www

Abstract

Researching formal models that explain selected natural phenomena of interest is a central aspect of most scientific work. A tested and confirmed model can be the key to classification, knowledge crystallization, and prediction.With this paper we propose a new approach to rapidly draft, fit and quantify model prototypes in visualization space. We also show that these models can provide important insights and accurate metrics about the original data. Using our technique, which is similar to the statistical concept of de-trending, data that behaves according to the model is de-emphasized, leaving only outliers and potential model flaws for further inspection. Moreover, we provide several techniques to assist the user in the process of prototyping such models. We demonstrate the usability of this approach in the context of the analysis of streaming process data from the Norwegian oil and gas industry, and on weather data, investigating the distribution of temperatures over the course of a year.

Published

Proceedings of Sigrad 2011

• Location: Stockholm, Sweeden
• Project: elad
• URL: http://www.ep.liu.se/ecp_article/index.en.aspx?issue=065;article=007

Media

BibTeX

@inproceedings{lampe11modelbuilding,
  author = {Ove Daae Lampe and Helwig Hauser },
  title = { Model Building in Visualization Space },
  booktitle = {Proceedings of Sigrad 2011 },
  location = {Stockholm, Sweeden},
  year = {2011},
  abstract = {Researching formal models that explain selected natural 
  phenomena of interest is a central aspect of most scientific work. 
  A tested and confirmed model can be the key to classification,
  knowledge crystallization, and prediction.With this paper we propose 
  a new approach to rapidly draft, fit and quantify model prototypes in 
  visualization space. We also show that these models can provide 
  important insights and accurate metrics about the original data. 
  Using our technique, which is similar to the statistical concept of 
  de-trending, data that behaves according to the model is de-emphasized, 
  leaving only outliers and potential model flaws for further inspection. 
  Moreover, we provide several techniques to assist the user in the 
  process of prototyping such models. We demonstrate the usability of 
  this approach in the context of the analysis of streaming process 
  data from the Norwegian oil and gas industry, and on weather data, 
  investigating the distribution of temperatures over the course of a year.},

  url = {http://www.ep.liu.se/ecp_article/index.en.aspx?issue=065;article=007},

}