Naive vs Heuristic Methods For Automatic Feature Selection Comparison

Y. Orenes Casanova, A. Rabasa Dolado, J. J. Rodríguez Sala, J. Sánchez Soriano

This study aims to compare a heuristic classifier (provided by WEKA) against a Naive classifier that performs exhaustive searches.
The methodology carried out part of the selection of a set of most relevant attributes of a sample database and on each combination of attributes calculates average accuracies (in several sub-samples) of the classification carried out by a heuristic method (incorporated in a WEKA classifier) and a Naive or exhaustive classifier.

Finally, it is intended to obtain a quantitative comparison of the average accuracies reached with each selection of attributes (for each method), but also a qualitative comparison of which method (Heuristic or Naive) provides better classification results in certain combinations of attributes (that incorporate variables of greater weight or of less weight, on the variable of class).

Keywords: Clasificación, selección de características, heurísticos, métodos exhaustivos

ME-1 Statistics Models
September 3, 2019  6:30 PM
I2L5. Georgina Blanes building

A. Conde Sánchez, A. M. Martínez Rodríguez, M. J. Olmo Jiménez, J. Rodríguez Avi, A. J. Sáez Castillo

N. Capote, I. Montes, S. Montes

B. Pérez-Sánchez, M. Gonzalez Espinosa, M. C. Perea Marco, J. J. López-Espín