A Comparative Analysis of Feature Selection Methods for Ensembles with Different Combination Methods

Santana, Laura Emmanuella Alves dos Santos; Oliveira, Diogo F. de; Canuto, Anne M. P.; Souto, Michael Vandesteen Silva

doi:10.1109/ijcnn.2007.4371032

Cited by 12 publications

(7 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It finds its applications in various areas not limited to signal processing, statistics, classification techniques suggested by Zheng et al (2004) and clustering techniques, machine learning as discussed by Canuto et al (2006). Of late in machine learning, feature selection with ensembles described by Santana et al (2007), Caragea et al (2003) is gaining high attention. Feature selection methods reduce the training time significantly.…”

Section: Related Workmentioning

confidence: 99%

“…In many of the previous works, the researchers used single classifiers for the pattern recognition problem. Santana et al (2007) had used six different feature selections over six different methods to carry out their experiment. They had used ensembles of classifiers such as those provided in the literature of Ho (1998), Rodriguez et al (2006) and Tumer and Oza (2003).…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Empirical study of feature selection methods over classification algorithms

Bhalaji¹,

Kumar²,

Selvaraj³

2018

IJISTA

View full text Add to dashboard Cite

Feature selection methods are deployed in machine-learning algorithms for reducing the redundancy in the dataset and to increase the clarity in the system models without loss of much information. The objective of this paper is to investigate the performance of feature selection methods when they are exposed to different datasets and different classification algorithms. In this paper, we have investigated standard parameters such as accuracy, precision and recall over two feature selection algorithms namely Chi-Square feature selection and Boruta feature selection algorithms. Observations of the experiments conducted using R studio resulted around 5-6% increased performance in above said parameters when they were exposed to Boruta feature selection algorithm. The experiment was done on two different datasets with different set of features and we have used the following five standard classification algorithms-Naive Bayes, decision tree, support vector machines (SVM), random forest and gradient boosting.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Empirical study of feature selection methods over classification algorithms

Bhalaji¹,

Kumar²,

Selvaraj³

2018

IJISTA

View full text Add to dashboard Cite

show abstract

“…There are several feature selection methods that can be used for ensembles, which can be broadly divided into two main approaches, which are: filter and wrapper. In the filter approach, as it can be found in [6], [7], [8], no need for a classification method to be used during the feature selection process. In other words, the feature selection process is independent from the classification method.…”

Section: Feature Selection In Ensemblesmentioning

confidence: 99%

“…For the ensemble systems using feature selection (original and proposed), two criteria were used as basis (first ranking) for the ranking of the attributes, Spearman Correlation and Variance. The idea of using Spearman Correlation is to have a rank-based correlation measure, while the use of variance is because it is a criterion that does not need the class label vector to calculate its ranking [8].…”

Section: Experimental Workmentioning

confidence: 99%

A Filter-Based Feature Selection Method Applied in Different Structures of Classifier Ensembles

Canuto¹,

Vale²,

Santos³

et al. 2016

Anais Do 9. Congresso Brasileiro De Redes Neurais

View full text Add to dashboard Cite

-Diversity is considered as one of the main prerequisites for an efficient use of ensemble systems. One way of increasing diversity is through the use of feature selection methods in ensemble systems. In this paper, a class-based feature selection method for ensemble systems is proposed. The proposed method is inserted into the filter approach of feature selection methods and it chooses only the attributes that are important only for a specific class. An analysis of the performance of the proposed method is also investigated in this paper and it shows that the proposed method has outperformed the standard feature selection method.

show abstract

“…At present, according to the statistics, aiming at the problem of solving linear programming, the scholars has proposed more than one thousand kinds of algorithm! Although they belong to three different algorithms faction [3][4][5][6][7][8][9][10][11] --simplex method, ellipse method and interior point method, there is a common characteristic --geometric meaning is not obvious, that is to say, the existing algorithm does not make full use of the geometrical property of the linear programming model.…”

Section: Introductionmentioning

confidence: 99%

A Reduced-Order Correction Algorithm for the Linear Programming Problem by Non-negative Constraints with the Most-Obtuse Angle

Liu¹,

Huang²

2013

JNW

View full text Add to dashboard Cite

Based on the linear and geometric-plane features of linear programming, and the consistent quantity of null-elements for all the non-degenerate base-solutions of some linear programming problem, with the help of the new platform, constructed by the conceptions including characteristic matrix of linear programming problem, negative pivoting iteration transformation and so on, a new solution method, named as reduced-order correction algorithm by nonnegative constraints with the most-obtuse angle, is proposed in this paper. At the same time, a -regularization method is given to avoid basis-circulating problem, which maybe take place for the reason of the degenerate phenomenon, in the process of solving linear programming. Comparing with the Simplex Method, the instance shows that it is more highly efficient for us to apply this new method to solving the linear programming proble

show abstract

A Comparative Analysis of Feature Selection Methods for Ensembles with Different Combination Methods

Cited by 12 publications

References 14 publications

Empirical study of feature selection methods over classification algorithms

Empirical study of feature selection methods over classification algorithms

A Filter-Based Feature Selection Method Applied in Different Structures of Classifier Ensembles

A Reduced-Order Correction Algorithm for the Linear Programming Problem by Non-negative Constraints with the Most-Obtuse Angle

Contact Info

Product

Resources

About