Second International Conference on Intelligent Computing & Information Systems (ICICIS)

March 5-7, 2005 Conference Proceedings

Organized by Faculty of Computer & Information Sciences (FCIS)
Ain Shams University and Association for Computing Machinery (ACM)
In co-operation with ACM SIGART (Special Interest Group for Artificial Intelligence) and
ACM SIGMIS (Special Interest Group for Management Information Systems)

This conference held at the Ain Shams University in Cairo, Egypt was attended by Graduate Student Aparna Varde.

Aparna enjoyed meeting with other Phd students who attended the Conference. During the visit to Cairo, she was able to go to the museums and also rode a camel!

Title of Paper Presented:

Data Mining Over Graphical Results of Experiments with Domain Semantics

Abstract:

The results of experiments in science and engineering are often represented graphically, since graphs serve as good visual tools for analysis of the corresponding processes to aid decision-making. Performing a laboratory experiment consumes time and resources. This motivates the need to estimate the results (graphs) that would be obtained from experiements given the input conditions. We propose an approach called "AutoDomainMine" for estimation. This consists of first clustering graphical results of existing experiments, and then using classification to learn the clustering criteria to build a representative pair of input conditions and graph per cluster. The representatives and learnt criteria form domain knowledge useful for estimation. We have found that this strategy provides significantly better estimation compared to similarity search and other methods, since it automates a learning method of scientist in discovering knowledge from experiemental results. Clustering graphs involves issues such as preserving domain semantics, defining similarity measures and reducing dimensionality with minimal loss. AutoDomainMine, with its challenges and evaluation, is discussed here.

Maintained by webmaster@wpi.edu
Last modified: Oct 24, 2006, 17:03 EDT