Nail polish is a prevalent form of trace evidence in forensic investigations, often linking suspects or victims to crime scenes. Traditional visual inspection methods are insufficient for reliably differentiating nail polish samples due to their complex composition and subtle variations across brands and formulations. This study evaluates the application of machine learning techniques to classify nail polish samples using spectral data from Attenuated Total Reflectance Infrared (ATR-IR) and Raman spectroscopy. A comparative analysis of machine learning algorithms, including Gaussian Mixture Models (GMM), Random Forest, Support Vector Machines (SVM), k-Nearest Neighbors (k-NN), and Logistic Regression, highlights the effectiveness of these approaches. Random Forest achieved the highest classification accuracy (99.95 %), significantly outperforming other algorithms. The GMM model, used for unsupervised clustering, achieved a silhouette score of 0.62, indicating moderate sample separation. ATR-IR spectroscopy is advantageous for detecting polar functional groups, such as carbonyl and hydroxyl groups, providing detailed information on resins and plasticizers in nail polish formulations. In contrast, Raman spectroscopy excels in identifying non-polar bonds and pigments, offering insights into polymeric chains and colorants. The complementary nature of these techniques enhances the discriminatory power of machine learning models, improving forensic classification accuracy. These findings demonstrate machine learning's potential to automate and enhance the forensic analysis of nail polish, offering improved accuracy, reliability, and interpretability in trace evidence classification.