Predicting microbial response to anthropogenic environmental disturbances using artificial neural network and multiple linear regression

•Unsupervised learning approach was explored for growth profile prediction of Klebsiella sp.•The model frameworks of ANN and MLR were designed to support predictive microbiology experiments.•The performance criteria for the best prediction were verified by comparative analysis of RMSE and R2.•ANN model presented better results than other state -of-art method. A mathematical model that quantitively describes the combined effect of different environmental variables can be used to predict the growth dynamics of a microorganism. This study evaluates the potential of an artificial neural network (ANN) model to predict the growth of a bacterial strain, Klebsiella sp., during the treatment of diclofenac sodium contaminated wastewaters. Input parameters, temperature, pH, time, agitation and diclofenac sodium concentration were randomly combined to conduct the batch experiments. Experimental data sets obtained were used for the training and optimization of programme learning. The efficiency of the ANN model was demonstrated by comparing it with the multiple linear regression (MLR) model. Root mean squared error (RMSE) and coefficient of determination (R2) were used as model performance parameters. The results obtained depict that the ANN model with RMSE 0.0124 and R2 value 0.926 in the test phase exhibited higher prediction performance. In contrast, low prediction performance was exhibited by the MLR model with RMSE 0.0230 and R2 value of 0.710. The results of this study are expected to guide the development of appropriate operational conditions for bioremediation of wastewater and industrial scale-up of the process. [Display omitted]