2013
DOI: 10.1016/j.ins.2012.10.011
|View full text |Cite
|
Sign up to set email alerts
|

Enhancing directed binary trees for multi-class classification

Abstract: One approach to multi-class classification consists in decomposing the original problem into a collection of binary classification tasks. The outputs of these binary classifiers are combined to produce a single prediction. Winner-takesall, max-wins and tree voting schemes are the most popular methods for this purpose. However, tree schemes can deliver faster predictions because they need to evaluate less binary models. Despite previous conclusions reported in the literature, this paper shows that their perform… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

0
6
0

Year Published

2013
2013
2022
2022

Publication Types

Select...
7

Relationship

1
6

Authors

Journals

citations
Cited by 22 publications
(6 citation statements)
references
References 33 publications
(44 reference statements)
0
6
0
Order By: Relevance
“…Kijsirikul and Ussivakul (2002) optimized for DDAG and proposed adaptive‐directed acyclic graph based on the rules of the tennis game, which enhanced the performance of the DDAG algorithm. After that, many researches show that the structure formed by placing the binary classifier with strong generalization ability on the upper layer of DDAG can achieve higher prediction accuracy (Montañés et al, 2013; Takahashi & Abe, 2003).…”
Section: Literature Reviewmentioning
confidence: 99%
“…Kijsirikul and Ussivakul (2002) optimized for DDAG and proposed adaptive‐directed acyclic graph based on the rules of the tennis game, which enhanced the performance of the DDAG algorithm. After that, many researches show that the structure formed by placing the binary classifier with strong generalization ability on the upper layer of DDAG can achieve higher prediction accuracy (Montañés et al, 2013; Takahashi & Abe, 2003).…”
Section: Literature Reviewmentioning
confidence: 99%
“…Lorena and Carvalho (Lorena and Carvalho 2008a;Lorena and Carvalho 2008b) introduced two algorithms based on minimum spanning trees for automatic generation of hierarchical tree using data collected from multiclass dataset. Montañés et al (2013) proposed two greedy methods by suggesting a measure for generating a directed binary tree. In the first method, two classes with largest distance (based on the proposed measure) are located at nodes of directed binary trees while in the second method, minimum representation of classes in each path is also considered.…”
Section: Literature Reviewmentioning
confidence: 99%
“…On the contrary, if classes are not similar, then ‖w‖ would be smaller and number of elements which do not have any class would decrease ( ∑ n i¼1 ε i →0). Consequently, distance increases (Montañés et al 2013). In order to solve multiclass problems, more than two classes with different setting parameter values are considered.…”
Section: Svm Similarity Measurementioning
confidence: 99%
“…To solve a multi-classification problem, conventional methods, such as ''1-a-1'' and ''1-a-r'' [33], have high training complexity and low classification accuracy. For a multi-classification problem, a multiple classification method that uses a decision tree (DT) is proposed [34].…”
Section: Situation Classificationmentioning
confidence: 99%