Prediction of electronic parameters of compensated multi-crystalline solar-grade silicon using artificial neural networks