Prediction of critical micelle concentration of some anionic and cationic surfactants using an artificial neural network

The critical micelle concentration (CMC) of a set of 58 alkylsulfates, alkylsulfonates, alkyltrimethyl ammonium and alkylpyridinium salts were predicted using an artificial neural network (ANN). The multiple linear regression (MLR) technique was used to select the important descriptors that act as inputs for artificial neural network. These descriptors are Balaban index, heat of formation, maximum distance between the atoms in the molecule, Randic index and volume of the molecule. Designed artificial neural network is a fully connected back-propagation network that has a 5-5-1 architecture. The results obtained using neural network were compared with those obtained using MLR technique. Standard error of calibration and standard error of prediction were 0.318 and 0.291, respectively for the MLR model and 0.137 and 0.122, respectively for the ANN model. These values reveal the superiority of artificial neural network over MLR model in prediction of log CIVIC for anionic and cationic surfactants.