This article describes the formulation of a tablet for a specific purpose, primarily using fractional or full factorial designs. The formulation work generated a matrix that was processed by two software packages based on neural networks. When the dataset was divided into smaller subsets, the agreement between the predicted and observed tablet properties of the optimized formulations was reasonable.
