For micro-channel fabrication using laser micro-machining processing, estimation techniques are normally utilised to develop an approach for the system behaviour evaluation. Design of Experiments (DOE) and the Artificial Neural Networks (ANN) are two methodologies that can be used as estimation techniques. These techniques help in finding a set of laser processing parameters that provides the required micro-channel dimensions and in finding the optimal solutions in terms reducing the product development time, power consumption and of least cost. In this work, an integrated methodology is presented in which the ANN training experiments were obtained by the statistical software DoE to improve the developed models in ANN. A 33 factorial design of experiments (DoE) was used to get the experimental set. Laser power, P; pulse repetition frequency, PRF; and sample translation speed, U were the ANN inputs. The channel width and the produced micro-channel operating cost per metre were the measured responses. Four Artificial Neural Networks (ANNs) models were developed to be applied to internal micro-channels machined in PMMA using a Nd:YVO4 laser. These models were varied in terms of the selection and the quantity of training data set and constructed using a multi-layered, feed-forward structure with a the back-propagation algorithm. The responses were adequately estimated by the ANN models within the set micro-machining parameters limits. Moreover the effect of changing the selection and the quantity of training data on the approximation capability of the developed ANN model was discussed.