VisDRTool

Carlos Manuel Ortega Castillo and Cielo Katherine Basante Villota, Universidad de Narño, San Juan de Pasto - Colombia, 2018

Currently, human analysis capabilities are not enough  imminent in the face of growth of technologies aiming to collect, communicate and store large volumes of information. Typically, such volumes of information are represented in high-dimensional databases, which can not be directly interpreted visually. That said, the dimensionality reduction (DR) has become to be a good alternative. From the input database, the DR techniques extract relevant information represented in a low-dimensional fashion, so that the performance of the subsequent pattern recognition and data mining tasks is improved. In general, the application and interpretation of RD procedures require expert personnel in data analysis, and, therefore, an increase in time and cost is generated for carrying out the subsequent stages of analysis. This fact has motivated the development of interactive strategies allowing for graphically representing outcomes of the analysis of high-dimensional data. Nonetheless, for the academic community, the development of a  - generic, versatile, open source and dedicated - tool for the interactive visualization of data based on DR principles is still considered an open issue.


Figure 1: Block diagram of the tool proposed in this degree work.

In this degree work, the implementation of a versatile tool for the support of databases visual analysis is presented, which enables the user to interactively generate low-dimensional graphic representations. For this purpose, the tool incorporates: i) interaction models - an existing one (chromatic model) and another proposed one (model based on angles) -, ii) a mixture of DR spectral methods represented in kernel-matrices-based approximations, iii) technical traditional visualization (scatter plots and parallel coordinates diagram). Additionally, aimed at generating a dynamic interaction (real time changes), the locally linear landmarks algorithm is implemented to perform the DR procedure at a low-computational cost. It is important to highlight that the entire tool is developed under scalability and modularity settings.




Graphic interface

The graphical interface consists of 4 modules at the top: Load data, interaction methods, visualization techniques and save data. In the lower part we find two panels, in the first one the selected interaction method is shown and in the second one the selected visualization technique is shown.


Figure 2:  View of the interface implemented on NetBeans Software.


Vídeo Tutorial


Publications

Thesis


Angle-Based Model for Interactive Dimensionality Reduction and Data Visualization 
International Workshop on Artificial Intelligence and Pattern Recognition: IWAIPR 2018 
C. K. Basante-Villota, C. M. Ortega-Castillo, D. F. Peña-Unigarro, E. J. Revelo-Fuelagán, J. A. Salazar-Castro, M. Ortega-Bustamante, P. Rosero-Montalvo, L. S. Vega-Escobar, D. H. Peluffo-Ordóñez 
See full paper

Comparative Analysis Between Embedded-Spaces-Based and Kernel-Based Approaches for Interactive Data Representation 
Colombian Conference on Computing - CCC 2018 
C. K. Basante-Villota, C. M. Ortega-Castillo, D. F. Peña-Unigarro, J. E. Revelo-Fuelagán, J. A. Salazar-Castro, D. H. Peluffo-Ordóñez 
See full paper





Personal information

 
Carlos Manuel Ortega Castillo 

Electronic engineer from the University of Nariño in Pasto - Colombia 


 Cielo Katherine Basante Villota

Electronic engineer from the University of Nariño in Pasto - Colombia

E-mail:
 
karlosmaor@hotmail.com


E-mail:
  

cielo.2103@gmail.com




Comments