A Comparison of Hierarchical Biclustering Ensemble Methods

Author(s)

Padilha, Victor A. ; de Carvalho, Andre C. P. L. F. ; IEEE

• Document type: Journal article
• Source: 2017 6TH BRAZILIAN CONFERENCE ON INTELLIGENT SYSTEMS (BRACIS); v. N/A, p. 6-pg., 2017-01-01.

Abstract

Biclustering aims at providing techniques able to detect submatrices following coherent patterns according to a pre-established criterion in a data matrix. These techniques are able to simultaneously cluster both dimensions of a data matrix (objects and features). Several real-world applications can benefit from the use of this paradigm (e.g., gene expression data analysis). However, biclustering is proven to be a NP-hard problem. For such, an alternative to provide more meaningful and robust solutions is the combination of several biclustering results by using ensemble methods. The development of ensemble methods is a relatively new issue in the biclustering literature. Thus, few methods have been proposed and, to the best of the authors' knowledge, no comparative study has been conducted so far. This paper compares two biclustering ensemble methods based on a hierarchical aggregation of previously found biclusters. Experiments are carried out with both on synthetic and real datasets. Despite the results obtained, the authors believe that there is room for further improvement of the methods for real gene expression data scenarios. (AU)

• FAPESP's process: 13/07375-0 - CeMEAI - Center for Mathematical Sciences Applied to Industry
• Grantee: Francisco Louzada Neto
• Support Opportunities: Research Grants - Research, Innovation and Dissemination Centers - RIDC
• FAPESP's process: 17/02975-0 - Ensembles of biclustering results
• Grantee: Victor Alexandre Padilha
• Support Opportunities: Scholarships in Brazil - Doctorate
• FAPESP's process: 16/18615-0 - Advanced machine learning
• Grantee: André Carlos Ponce de Leon Ferreira de Carvalho
• Support Opportunities: Research Grants - Research Partnership for Technological Innovation - PITE