Determining the structure of directed acyclic graphs in multiclass classification using complexity measures of supervised problems

• Grant number: 15/17291-3
• Support Opportunities: Scholarships in Brazil - Master
• Start date: December 01, 2015
• End date: February 28, 2017
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Ana Carolina Lorena
• Grantee: Thaise Marques Quiterio
• Host Institution: Instituto de Ciência e Tecnologia (ICT). Universidade Federal de São Paulo (UNIFESP). Campus São José dos Campos. São José dos Campos , SP, Brazil
• Associated research grant: 12/22608-8 - Use of data complexity measures in the support of supervised machine learning, AP.JP

Abstract

Many practical problems involve distinguishing data from multiple classes. One possible approach to deal with such multiclass classification problems is to decompose them into several binary sub-problems and combine their solutions. The One-versus-One decomposition, which generates a binary subproblem for each pair of classes is one of the most widely used decomposition strategies. Standard binary classification techniques can then be used to induce classifiers for the pairs of classes, while the hierarchical structure of a Directed Acyclic Graph (DAG) can be employed to combine their outputs. As the DAG predictive results in multiclass classification are dependent on how the binary classifiers are arranged within the hierarchy, this work will employ complexity measures of supervised classification problems to determine the structure of DAGs. These measures estimate the complexity of the classification boundary based on indexes derived from data available for learning. The goal is to position simpler binary subproblems in the top of the DAG hierarchy, aiming to minimize error propagation through the hierarchical structure. (AU)

Scientific publications

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

QUITERIO, THAISE M.; LORENA, ANA C.; IEEE. Determining the Structure of Decision Directed Acyclic Graphs for Multiclass Classification Problems. PROCEEDINGS OF 2016 5TH BRAZILIAN CONFERENCE ON INTELLIGENT SYSTEMS (BRACIS 2016), v. N/A, p. 6-pg., 2016-01-01. (12/22608-8, 15/17291-3)

QUITERIO, THAISE M.; LORENA, ANA C.. Using complexity measures to determine the structure of directed acyclic graphs in multiclass classification. APPLIED SOFT COMPUTING, v. 65, p. 428-442, APR 2018. (12/22608-8, 15/17291-3)