A neural network based methodology for the modelling of a sequencing batch reactor (SBR) for producing Polyhydroxybutyrate (PHB) with Mixed Microbial Cultures (MMC) is proposed. The advantages of applying MMC for more effective production of PHB have already been documented and mechanistic models were developed, however, the lack of good understanding and the ability to describe phenomena involved in the complex nature of the bioprocess led to unsuccessful release of reliable and accurate mechanistic models. In order to perform successful process control and optimisation, empirical models developed from process operational data should be capitalised. Bootstrap aggregated neural networks are used in this study to enhance model accuracy and reliability. In the case of PHB production through SBR using MMC, the two feeding substrates of acetate and ammonia were found to play dominant roles in PHB production trajectory and different process operation regimes exist depending on the concentrations of these substrates. This paper proposes a method for the classification of such operation regimes and building neural network models corresponding to these regimes using bootstrap aggregated neural networks.