The primary objective of this study was to investigate the potential value of filtering algorithms in enhancing spectral model performance. 290 mechanical lubricants oils samples from various applications underwent measurement using Fourier transform Raman spectroscopy (FT-Raman). Subsequently, three discrimination models were developed employing C5.0, Bayes discriminant analysis, and support vector machine algorithms. The emphasis was on evaluating the efficacy of four filter algorithms—Fast Fourier Transform (FFT), Discrete Wavelet Transform (DWT), Finite Impulse Response (FIR), and Hilbert Transform (HT)—for enhancing model performance. Among these, the FFT filter-SVM model, DWT filter-SVM model, and FIR (low-pass or band-stop filter)-SVM model emerged as optimal choices for identification. These models demonstrated outstanding performance, achieving 100 % accuracy across all 290 samples. Modelling using FT-Raman and chemometrics offered a simpler, more robust, and valuable approach for identifying mechanical lubrication oils across diverse applications, exploiting the potential of filtering algorithms in forensic spectroscopy.