This study presents the empirical exploration of food waste recovery throughout the electrostatic separation process. In addition, the paper discusses the potential of artificial neural network (ANN) in predicting the responses. A five-level three-factor Taguchi orthogonal array (OA) design of experiment was employed as an initiative to optimize the prediction process. The electrostatic separation process was modelled using ANN by considering the recovered food waste and misclassified middling product during separation. A multi-layer feed-forward network developed in MATLAB was constructed. It was found that the results from the experiment and predicted model were in very good agreement. To our best knowledge, this is the first report for prediction of food waste separation performance employing ANN and Taguchi design.