Neural network computational techniques are a new alternative to conventional numerical modeling. This paper presents modeling using response surface methodology (RSM). Box and Wilson Central Composite Design (CCD) is used for preparing experiment matrix. The independent parameters in the experiment are speed, feed, force and number of tool passes. These variables are controlled during the burnishing process. The response parameter is micro hardness. Experimental samples are prepared using Single Roller Burnishing Tools (Carbide). Vickers micro hardness tester is used to measure micro hardness. A quadratic mathematical model is developed using RSM. An Artificial neural network (ANN) model is developed using three-layer feed-forward back-propagation. The neural network model is trained using measured values of micro hardness. The different algorithms are used to train the model. Best performance is achieved with correlation coefficient 0.9. This study concludes that an artificial neural network is the best alternative to fit the nonlinear data.