A new approach that integrates partial least squares (PLS) and windowing weighted mean filter (WWMF), a modification of windowing mean filter (WMF), is introduced for the analysis of six concentrations of ethanol aqueous solutions measured by THz frequency multiplication chain (TFMC) and vector network analyzer (VNA) instrument. The proposed WWMF was used to denoise and extract features of the measured signals. In WWMF, the linear correlation coefficients between concentrations of ethanol aqueous solutions and signal amplitudes are served as weighted coefficients to preform average on the signals. Partial least squares (PLS) model is subsequently established to obtain the prediction of concentrations. Final results suggest WWMF-PLS exhibit some preponderance comparing with WMF-PLS pursuant to the values of root mean square error. Furthermore, this investigation provides another potential approach for analysis different concentrations of aqueous solutions.