Seleksi Peubah menggunakan Algoritme Genetika pada Data Rancangan Faktorial Pecahan Lewat Jenuh Dua Taraf

Authors

  • Ani Safitri Department of Statistics, IPB University, Indonesia
  • Rahma Anisa Department of Statistics, IPB University, Indonesia
  • Bagus Sartono Department of Statistics, IPB University, Indonesia

DOI:

https://doi.org/10.29244/xplore.v10i1.473

Keywords:

genetic algorithms, fractional factorial, supersaturated, stepwise regression, variable selection

Abstract

In certain fields, experiments involve many factors and are constrained by costs. Reducing runs is one of the solutions to reduce experiment costs. But that can cause the number of runs to become less than the number of factors. This case of experimental design also is known as a supersaturated design. The important factors in this design are generally estimated by involving variable selection such as forward selection, stepwise regression, and penalized regression. Genetic algorithm is one of the methods that can be used for variable selection, especially for high dimensional data or supersaturated design. This study aims to use a genetic algorithm for variable selection in the supersaturated design and compare the genetic algorithm results with a stepwise regression which is generally used for a simple design. This study also involved fractional factorial design principles. The result showed that the main factors and interactions of the genetic algorithm and stepwise regression were quite different. But the principle was the same because the variables correlated. The genetic algorithm model had a smaller AIC and BIC and all of the main factors and interactions which had chosen were significant on the 0.1%. Therefore genetic algorithm model was chosen although computation time was much longer than stepwise regression.

Downloads

Published

2021-01-01

How to Cite

Safitri, A., Anisa, R., & Sartono, B. (2021). Seleksi Peubah menggunakan Algoritme Genetika pada Data Rancangan Faktorial Pecahan Lewat Jenuh Dua Taraf. Xplore: Journal of Statistics, 10(1), 55–69. https://doi.org/10.29244/xplore.v10i1.473

Issue

Section

Articles

Most read articles by the same author(s)