Recently, considerable attention has been paid for the many kinds of flexible production processes. Most concerned one is the incremental sheet metal forming (ISMF) process, which does not require any high capacity press machine and a set of dies with specified shape for the product. This process is an innovative approach to Rapid Prototyping and manufacturing of products in batch production. It is performed in room temperature (Cold Forming) and needs CNC machine, tools with hemispherical tip and fixture to clamp the blank. There are two kinds of this process: Single Point Incremental Forming (SPIF) and Two Point Incremental Forming (TPIF). This research focuses on SPIF. Although, this process is widely investigated by several researchers, it requires more practice to make it industrial. Since this process has wide usage in automotive industry, manufacturing of an automotive component (door handle) is consider ed in this research. Experimental tests have been designed and implemented based on the Taguchi design of experiment to investigate the effect of process parameters such as: Axial depth, tool diameter, feed rate and spindle speed on springback and dimensional precision. A reliable statistical analysis is carried out to extract the importance of each process parameter. In recent years, the application of neural network techniques to forming processes has been a research topic for optimizing and predicting process parameters and ISMF process has not been an exception. Neural network can be considered as a black box that a designer can use it in a very simple way without the complete knowledge of the existing relationships between input and output. The approach here addressed is based on the use of a neural network that, starting from a given process condition, supplies an effective prediction of the spring back. Actually, the purpose in this section is using of experimental data to reach a neural network model with the least springback prediction error. One of the main characteristics of a neural network is the large amount of required data during the training and testing phase to obtain a good generalization capability. This could require a very long execution time if a “standard” neural network approach is applied, but it is important that a careful neural network design is required. For this purpose, the large number of neural network configurations must be investigated. In this research, the Taguchi design of experiment is used to design experiments based on network parameters such as: Number of neuron, Initialization and parameters of Learning Rules (Error Back propagation algorithm with momentum constant and Levenberg–Marquardt algorithm) to extract the importance of each network parameter. Also, the network generalization methods (Early Stopping, K-Fold Cross–Validation) are used for improving network performance. Finally, the desired neural network model is identified and its parameters (Weight and Bias matrix) are obtained. Keywords: Incremental Forming, Taguchi Design of Experiment, Spring Back, Artificial Neural Networks.