Application of neural network in prediction of acetic acid yield by Acetobacters

• Biotransformation of acetic acid from cheese whey permeate using Acetobacters . • Acetic acid yield prediction through artificial neural network (ANN). • ANNs capture non-linear relationship between governing parameters and yield. • Average R 2 > 0.99 and RMSE ≃ 0.03 signifies high ANN model accuracy. In the present work, artificial neural network (ANN) is applied for the estimation of acetic acid yield for 3 different species of Acetobacters like, Acetobacter pasteurianus (NCIM 2104), Acetobacter aceti (NCIM 2116) and Acetobacter xylinum (NCIM 2526). Though there is open literature mentioning acetic acid and ANN can be found, they hardly describe the usage of ANN in prediction of fermentation based acetic acid production. Indeed, a deep dearth of existing literature is felt in this area to develop a robust ANN model to predict the yield of biologically obtained acetic acid and this work is a step towards bridging that research gap. The performance of the model has been estimated with R 2 (0.992, 0.988 and 0.992, respectively for the mentioned microbial species) and RMSE (0.0287, 0.034 and 0.020, respectively for the same species). The most relevant operating parameters like, temperature, agitator speed, concentrations of supplemented yeast extract and tryptone, have been considered to carry out fermentation on cheese whey permeate containing fermentable lactose (48.5 g L -1 ) to transform into acetic acid. Outcome datasets obtained from rigorous experimental investigations performed on the direct fermentative production of acetic acid are trained in the ANN model to predict the product yield. Such machine-learning methodology encourages reasonably accurate prediction of product generation which is extremely tough to obtain through classical analytical processes.