Solving real-life optimization problems using conventional methods requires substantial effort and time as the solution lies in the precise quantities associated with the variables which have nonlinear relationships. The variables also have equality and non-equality constraints. The presence of large number of variables and constraints generates an exceptionally wide search space. Classical optimization techniques such as exhaustive search have been found to be inefficient for solving such type of problems. Approximate algorithms based upon the concept of evolution and swarm intelligence have been found to be more efficient in solving high-dimensional problems characterized by the presence of large number of variables and constraints. These algorithms have been found to provide high speedups, easier to implement and provide near optimal solutions by augmenting the current best solution using the principle of randomization. The approximate algorithms are being widely used in multifarious fields of science including computational intelligence, artificial intelligence and soft computing. Many of approximate algorithms are inspired by various phenomena which occur in nature. genetic algorithm (GA), particle swarm optimization (PSO), artificial bee colony (ABC), firefly (FF) algorithm, ant colony optimization (ACO), cuckoo search (CS), gravitational search algorithm (GSA), bacterial foraging (BF), etc. are popular nature-inspired optimization algorithms. These algorithms are being used in the area of civil engineering. Cui and Sheng [1] used genetic algorithm and finite element displacement method to compute reliability index. They found that genetic algorithm relatively took less computation time. Liu et al. [2] used automatic grouping genetic algorithm for optimizing the pile group design. Liu et al. [3] proposed genetic algorithm for determining loadcarrying capacity of composite foundation. Elbeltagia et al. [4] compared PSO with genetic algorithm and other evolutionary techniques and found the PSO algorithm to be more efficient and easy to implement. ANNs have been used to compute the load-carrying capacity of driven piles by a number of researchers [4–۶]. Ismail et al. [7] used ANN for load deformation analysis of pile foundation. The ANN was trained first by PSO and then by back propagation by [8, 9]. Ismail et al. [10] analyzed load deformation of pile foundation by PSO-BP hybrid. The input parameters considered by various researchers for the design of pile foundation are given in Table 1. The input parameters used for the study were effective angle of shearing resistance, effective cohesion intercept, modulus of deformation, Poisson’s ratio, unit weight and depth of soil layer. These input parameters were the same as required by the finite element method for the determination of unit skin friction resistance and unit tip bearing capacity of the pile foundation. The ANNs were developed using the input and target data obtained from the finite element method. These input/output data are free from noise which generally characterizes the field observations due to heterogeneous and anisotropic behavior of the soil. The manual and instrumental errors also add to the noise in the field observations. The ANNs so developed in the study reported a high value of coefficient of determination averaging 0.99996.