A new variable selection method based on SVM for analyzing supersaturated designs

Abstract

Supersaturated designs (SSDs) are designs whose factors exceeds run size; thus, there are not enough runs for estimating all the main effects. They are commonly used in screening experiments, where the primary goal is to identify the few, but dominant, active factors, keeping the cost as low as possible. The development of new statistical methods inspired by machine learning algorithms is increasing rapidly, especially nowadays. One of such methods is the support vector machine recursive feature elimination (SVM-RFE), which manages to extract the informative genes in classification problems, while it achieves extremely high performance. In this article, we study a variable selection method for regression problems, called SVR-RFE, to screen active effects in both two-level and mixed-level designs. Simulation studies demonstrate that this procedure is effective enough, especially in terms of statistical power.

Keywords:

• linear models
• recursive feature elimination
• supersaturated designs
• support vector machines
• SVR-RFE

Acknowledgments

We are grateful to the editor and the referees for their constructive comments, which have led to improvements in the present manuscript. The research of the first author (K.D.) was financially supported by a scholarship awarded by the Secretariat of the Research Committee of National Technical.

Additional information

Notes on contributors

Krystallenia Drosou

Krystallenia Drosou received her Bachelor in Applied Mathematical and Physical Sciences in 2013 from the National Technical University of Athens, and her MSc in Mathematical Modelling in Applied Mathematical Sciences in 2015 from the National Technical University of Athens. She is currently a PhD candidate at the National Technical University of Athens. Her research interests are construction and statistical analysis of experimental designs, computer experiments and robust parameter designs, generalized linear models, and model selection criteria.

Christos Koukouvinos

Christos Koukouvinos graduated (PhD 1988) from the University of Thessaloniki. He is currently Professor at the National Technical University of Athens. He is the author of numerous papers in the field of statistics and combinatorics and he served on the editorial board of ten related journals and was a guest editor for three special issues. He was awarded the prestigious Hall Medal of the ICA in 1996. He is a fellow of the ICA and was a member of the Council of ICA for the period 2000-2003. His research interests include statistical experimental and optimal designs, statistical process control, robust parameter design, biostatistics, and combinatorial designs.