Multivariable power least squares method: Complementary tool for Response Surface Methodology
Year: 2019
Type: article
Abstract: In Response Surface Methodology (RSM), variables are correlated through polynomial functions based on Stone-Weierstrass theorem. However, such formulation inherits four weaknesses: possible misleading approximation, incapability to accurately determine the ranking of factors' dominance, failure to analyse factors in random value and proliferation of guess functions due to Pascal Triangle. Therefore, this article aims to develop an improvised method to rectify and complement the weaknesses of RSM. Multivariable Power Least Squares Method (MPLSM) has been developed to correlate various sets of independent variables with dependent variable in the form of power functions. MPLSM is built upon least squares method, and able to approximate the indices of the variables easily. Two variants of MPLSM are suggested to further ensure the numerical stability : the Normalised MPLSM and Iterative MPLSM. The proposed method is not only substantial in big data analysis and multivariable problems, but also providing an alternative approach in engineering optimisation.
Source: Ain Shams Engineering Journal
Authors: Wah Yen Tey, Kiat Moon Lee, Yutaka Asako, Lit Ken Tan, Norio Arai
Institutions: University of Technology Malaysia, UCSI University
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: gold
Estimated APC paid: $650
OpenAlex ID: W2972033759