Combining wavelet filtering with ordinary least squares (wavelet OLS) in multiple linear regression

Abstract

Wavelets are known for their ability to filter out noise from data. Combining wavelet filtering with ordinary least squares Regression leverages this capability, leading to more accurate models by focusing on the essential features of the data and minimizing the impact of random noise. In order to reduce noise in the data when estimating model parameters, combining wavelet filtering with ordinary least squares was proposed in this paper. This method includes the wavelet transform for the dependent variable through one of the family wavelets and uses the threshold method with the threshold rule. The mean squared error can then be used to compare the results of the Combining wavelet filtering with ordinary least squares and classical OLS method, which is simulated using a (MATLAB-R2022b) program created especially for this purpose. The study showed that, compared to the conventional method, the mixed method produces parameter estimations that are more accurate.