<p>A new chaotic circuit-based machine learning algorithm has been designed and implemented. The new algorithm uses a nonlinear resistor chaotic circuit to obtain the dataset. Due to the high sensitivity and nonlinear behavior inherent in chaotic systems, broad and diverse data distributions, which are difficult to obtain using classical methods, can be achieved. The proposed method especially offers a great advantage simulating conditions that are experimentally difficult or costly to obtain in “metallic welding.” The algorithm initially produces gas ratios (i.e., gas mixtures) and corresponding material characteristics and parameters. For training purposes, the material properties of alloys and related welding gases have been measured experimentally. This dataset is then used to train and model the proposed machine learning algorithm. The estimation procedure has been performed on the material parameters (i.e., elongation, yield strength, impact test, and metal hardness) under various gas ratios. Following the training procedure, good estimation performance has been observed across different machine learning regression methods.</p>