A neural-network-based intelligent system for time-series prediction problems in product development

We investigate the application of artificial neural networks (ANNs) as an intelligent agent to time-series prediction problems in pharmaceutical product development. The objective of the study is to use the Elman recurrent network as a development tool for predicting in-vitro dissolution profiles of matrix controlled release theophylline pellets-one of the widely used drugs in the treatment of asthma patients. Instead of estimating parameters of certain mathematical models that fit the profiles, a different approach has been applied by using ANN as a model for predicting the whole profile directly. Performance of the network was assessed by evaluating the similarity factor, f₂, in accordance with the recommendation of the United State Food and Drug Administration. In addition, reliability of the f₂ values was justified by calculating confidence bounds using the bootstrap method. The promising results obtained reveal the potential of the Elman network as a decision support tool to overcome problems in pharmaceutical product formulation