Abstarct: Accurate prediction of retention indices (RI) of volatile organic compounds (VOCs) is essential for reliable identification and characterization of plant mextabolites. In this study, artificial neural network (ANN) modeling combined with different variable selection and optimization strategies was applied to predict the retention indices of organic compounds present in the Lilium plant. Three hybrid modeling approaches were investigated: stepwise regression–artificial neural network (SR-ANN), firefly algorithm–artificial neural network (FF-ANN), and random forest–artificial neural network (RF-ANN). Molecular descxriptors were initially calculated for the studied compounds, and each method was employed to select the most relevant variables for ANN modeling. The predictive performance of the developed models was evaluated using statistical criteria such as the coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE). Among the applied approaches, the RF-ANN model demonstrated superior predictive accuracy and robustness compared to the SR-ANN and FF-ANN models, indicating a more effective variable selection capability and improved nonlinear modeling of the RI data. The results confirm that the integration of random forest with ANN is a powerful and reliable tool for predicting retention indices of VOCs in Lilium species and can be extended to similar chemometric and chromatographic applications.