The present study used handheld Raman and attenuated total reflectance-Fourier transform infrared spectroscopy for detecting drugs spiked in different drinks (matrices). Five key drugs commonly encountered in drug facilitated sexual assault cases were evaluated and included: citalopram; diphenhydramine hydrochloride; flurazepam; gammabutyrolactone; melatonin and valerian. These drugs in alcoholic beverages, non-alcoholic beverages and tea. Samples were then measured post-spiking using infrared and Raman spectrometers at an interval of four weeks and spectra were exported to Matlab 2024a, where machine learning algorithms were applied. Two algorithms were used for classification of drugs and matrices and were correlation in wavelength space and principal component analysis. In addition, partial least square regression was used for predicting the concentration of drugs in different matrices. Classification results showed that principal component analysis yielded the most accurate classification when applied to the infrared spectra of the samples. This was because infrared spectroscopy showed higher sensitivity over Raman in detecting the selected samples. However, both techniques showed the same accuracy and precision in prediction when partial least square regression models were applied. Herein, only two models out of eight showed accurate and precise performance and that was related to the number of samples. Consequently, future work involves applying the classification and regression models to larger sample size.