A three-layered feedforward backpropagation neural network that is fully connected to the succeeding layer through the connection weights is used for prediction of effective thermal conductivity of metal filled polymer composites. The volume fractions and thermal conductivities of continuous and dispersed phases were used as input parameters and an output in the form of effective thermal conductivity of polymer composites was obtained. The resultant predictions of effective thermal conductivity by the different training functions of artificial neural network agree well with the available experimental data. The different training functions of artificial neural network exhibit the capability to use the artificial neural networks for predictions of effective thermal conductivity of various types of tailored complex materials.