Thin layer drying characteristics of dill leaves under fixed, semi-fluidized, and fluidized bed conditions were studied at air temperatures of 30, 40, 50, and 60°C. In order to find a suitable drying curve, 12 thin layer-drying models were fitted to the experimental data of the moisture ratio. Among the applied mathematical models, the Midilli et al. model was the best for drying behavior prediction in thin layer drying of dill leaves. To obtain the optimum network for drying of dill leaves, various numbers of multilayer feed-forward neural networks were made and tested with different numbers of hidden layers and neurons. The best neural network feed-forward back-propagation topology for the prediction of drying of dill leaves (moisture ratio and drying rate) was the 3-45-2 structure with the training algorithm trainlm and threshold functions logsig and purelin. The coefficient of determination for this topology for training, validation, and testing patterns was 0.9998, 0.9981, and 0.9990, respectively. Effective moisture diffusivity of dill leaves during the drying process in different bed types was found to be in the range from 7.10 10^-12 to 1.62 10^-10 m² s^-1. Also, the values of activation energy were determined to be between 75.435 and 80.118 kJ mol^-1.