In this paper, artificial neural networks (ANNs) have been trained for the prediction of unit shaft friction and unit tip bearing capacity by using two parameters namely soil type and cone penetration resistance value. Thirty five pile loading case histories have been compiled with known soil types and Cone Penetration Test (CPT) results. For all of the 35 piles under the study, the unit skin friction acting around circumferential area of pile has been calculated for every soil layer with which the piles interact. For each pile, the unit end bearing capacity of soil layer in which an individual pile rest has also been calculated. The developed ANNs have been expressed in the form of two sets of equations. The load carrying capacity of all piles under study have been calculated from these equations and compared with load carrying capacity predicted by direct CPT methods. From this study, it has be concluded that the proposed model gives enhanced performance than conventional direct CPT methods. The proposed equations can predict ultimate load carrying capacity of bored piles in axial compression with improved accuracy.