Classifier-independent feature selection for two-stage feature selection

Lecture notes in computer sciencePublished 1 January 1998Open access

Mineichi Kudo, Jack Sklansky

DOI10.1007/bfb0033278

Source URL

Citations23

SJR quartileQ2

SJR score0.35

SNIP0.55

View PDF

TL;DR

The aim is to remove garbage features and to improve the classification accuracy of all the practical classifiers compared with the situation where all the given features are used.

Abstract

The effectiveness of classifier-independent feature selection is described. The aim is to remove garbage features and to improve the classification accuracy of all the practical classifiers compared with the situation where all the given features are used. Two algorithms of classifier-independent feature selection and two other conventional classifier-specific algorithms are compared on three sets of real data. In addition, two-stage feature selection is proposed.

Keywords

Computer ScienceEngineering

C4.5: Programs for Machine Learning

23,665 Citations1992J. R. Quinlan

A complete guide to the C4.5 system as implemented in C for the UNIX environment, which starts from simple core learning methods and shows how they can be elaborated and extended to deal with typical problems such as missing data and over hitting.

IEEE Transactions on Pattern Analysis and Machine IntelligenceFeature selection: evaluation, application, and small sample performance

2,167 Citations1997Anil K. Jain, Douglas E. Zongker

This work studies the problem of choosing an optimal feature set for land use classification based on SAR satellite images using four different texture models and shows that pooling features derived from different texture Models, followed by a feature selection results in a substantial improvement in the classification accuracy.

Machine intelligence and pattern recognitionComparative study of techniques for large-scale feature selection

301 Citations1994Francesc J. Ferri, Pavel Pudil +2 more

Investigation of the applicability of recently developed techniques based on the classical sequential methods and the Floating search algorithms to high dimensional problems of feature selection to establish whether the properties inferred for these techniques from medium scale experiments extend to dimensionalities of one order of magnitude higher.

Pattern Recognition LettersPattern Recognition in Practice VI

192 Citations1999E. S. Gelsema, Jifke F. Veenland

This special issue contains most of the papers presented at the conference ``Pattern Recognition in Practice VI'' held on Vlieland, The Netherlands, from 2 to 4 June 1999, to stimulate interaction between pattern recognition professionals and scientists from other disciplines who bring pattern recognition techniques into practice.

IEEE Transactions on Pattern Analysis and Machine IntelligenceDivergence based feature selection for multimodal class densities

101 Citations1996Jana Novovičová, Pavel Pudil +1 more

A new feature selection procedure based on the Kullback J-divergence between two class conditional density functions approximated by a finite mixture of parameterized densities of a special type is presented, which simultaneously yields a pseudo-Bayes decision rule.

Pattern RecognitionFeature selection based on the approximation of class densities by finite mixtures of special type

67 Citations1995Pavel Pudil, Jana Novovičová +2 more

A new method of feature selection based on the approximation of class conditional densities by a mixture of parameterized densities of a special type, suitable especially for multimodal data, is presented.

Czech Digital Mathematics Library (Institute of Mathematics CAS)A COMPARATIVE EVALUATION OF MEDIUM-AND LARGE-SCALE FEATURE SELECTORS FOR PATTERN CLASSIFIERS

33 Citations1998Mineichi Kudo, Jack Sklansky

This study proposes a unified way to compare a large variety of algorithms and shows that the sequential floating algorithms promises for up to medium problems and genetic algorithms for medium and large problems.

View PDF

Pattern RecognitionConstruction of class regions by a randomized algorithm: a randomized subclass method

32 Citations1996Mineichi Kudo, S. Yanagi +1 more

A randomized algorithm is proposed for solving the problem of finding hyper-rectangles, sufficiently approximating the true region in each class, which yields a suboptimal solution, but is more efficient than previous methods.

Machine intelligence and pattern recognitionInitializing the EM algorithm for use in Gaussian mixture modelling

29 Citations1994P. McKenzie, M. Alder

A method of initializing the EM algorithm to cluster centers called the Dog Rabbit Strategy is considered and this is compared to the k-means clustering algorithm.

Machine intelligence and pattern recognitionRelative feature importance: A classifier-independent approach to feature selection

29 Citations1994Hilary J. Holz, Murray H. Loew

This work presents an alternative approach that uses a new metric for discriminatory power called relative feature importance (RFI), which ranks features by discriminatory power based on their distributional structure, without parametric assumptions.