Use of One-factor Design of Experiments (DOE) for Regression Modeling: A Robust Methodology

Abstract

In the present research era, high accuracy methods as a statistical analysis tool are increasing. Therefore, researchers are more focused to produce reliable and accurate results. Hence, the use of data modeling techniques is more focused to meet the needs of the current research trend. On the other hand, Design of Experiment (DOE) is extensively used among various scientific fields; however, its limitations do not allow these study designs for modeling purposes. Therefore, this study was designed to develop a methodology combining statistical methods that can provide to use one-factor DOE study designs for modeling and predictions. The addition of Fuzzy regression and multilayer feedforward (MLFF) neural network along with multiple linear regression would provide more accurate results with high accuracy. Furthermore, the developed methodology was tested on a dataset to test the methodology's performance and results provided that methodology provided regression models through MLR and fuzzy with high accuracy with the testing of the model's predictability through MLFF.