A Directed Acyclic Graph Based Approach to Multi-Class Ensemble Classification

@inproceedings{Alshdaifat2015ADA,
  title={A Directed Acyclic Graph Based Approach to Multi-Class Ensemble Classification},
  author={Esra'a Alshdaifat and Frans Coenen and Keith Dures},
  booktitle={SGAI Conf.},
  year={2015}
}
In this paper a novel, ensemble style, classification architecture is proposed as a solution to the multi-class classification problem. The idea is to use a non-rooted Directed Acyclic Graph (DAG) structure which holds a classifier at each node. The potential advantage offered is that a more accurate and reliable classification can be obtained when the classification process is conducted progressively, starting with groups of class labels that are repeatedly refined into smaller groups until… 

References

SHOWING 1-10 OF 19 REFERENCES

A Multi-path Strategy for Hierarchical Ensemble Classification

TLDR
A multi-path strategy is investigated based on the idea of using Classification Association Rule Miners at individual nodes to determine, at each node, whether one or two paths should be followed.

Hierarchical Single Label Classification: An Alternative Approach

TLDR
Experimental results show that the proposed mechanism can improve classification performance in terms of average accuracy and average AUC in the context of some data sets.

Hierarchical Classification for Solving Multi-class Problems: A New Approach Using Naive Bayesian Classification

TLDR
A hierarchical classification ensemble methodology is proposed as a solution to the multi-class classification problem where the output from a collection of classifiers are combined to produce a better composite global classification.

Multiclass Classification with Filter Trees

We present a new algorithm, filter tree, for reducing (cost-sensitive) k-class classification to binary classification. The filter tree is provably consistent, in the sense that given an optimal

A hierarchical method for multi-class support vector machines

TLDR
This work introduces a framework, which is called Divide-by-2 (DB2), for extending support vector machines (SVM) to multi-class problems and shows that, DB2 is faster than one-against-one and one- against-rest algorithms in terms of testing time, significantly faster than the standard one- Against-Rest algorithms interms of training time, and the cross-validation accuracy ofDB2 is comparable to these two methods.

Hierarchical Fusion of Multiple Classifiers for Hyperspectral Data Analysis

TLDR
This paper introduces a hierarchical technique to recursively decompose a C-class problem into C_1 two-(meta) class problems, and introduces a generalised modular learning framework used to partition a set of classes into two disjoint groups called meta-classes.

Half-Against-Half Multi-class Support Vector Machines

TLDR
Both theoretical estimation and experimental results show that HAH has advantages over OVA and OVO based methods in the evaluation speed as well as the size of the classifier model while maintaining comparable accuracy.

Ensemble Learning

  • Gavin Brown
  • Computer Science
    Encyclopedia of Machine Learning and Data Mining
  • 2017
TLDR
A framework for categorizing ensemble classifiers, that is, machine learning based classifiers that utilize a combination of scoring functions, is provided, and several ensemble techniques are outlined, discussing how each fits into this framework.

Using two-class classifiers for multiclass classification

  • D. TaxR. Duin
  • Computer Science
    Object recognition supported by user interaction for service robots
  • 2002
TLDR
This paper wants to show the possibilities of simple generalizations of the two-class classification, using voting and combinations of approximate posterior probabilities.

Classifier ensembles: Select real-world applications