“…The input data were normalised, initially selecting for the neural network training the input data reported in an Excel file corresponding to: pH, alkalinity, total solids (TS), total volatile solids (TVS), total nitrogen (N), ammonia nitrogen (N-NH 3 ), Fe, Mn, Zn, Cd, Cu and as output of the network the COD value. Subsequently, taking into account the principal component analysis (PCA) discussed in Grisa et al, (2010) [29], a neural network was designed using as inputs the variables most correlated with the COD value: alkalinity, total volatile solids (TVS), Zn, Cd, Cu.…”