The paper addresses the problem of constructing neural networks for analyzing nonlinear dynamic signals using the support models method. The aim of the work is to ensure the universality of the machine learning pipeline when constructing neural networks using the support models method by developing a generalized machine learning pipeline model based on the use of the “Factory” design pattern to perform the support models superposition procedure. The novelty of the work lies in the use of the “Factory” design pattern to automate the construction of neural networks in the form of a superposition of support models. This increases the reproducibility of the model construction process and significantly reduces the time required to construct target models, while maintaining high accuracy, which is critical for real-world applications. The practical value of the work for the information technology industry lies in the automation of the selection and composition of pre-trained neural networks, which bridges the gap between the support models method for identifying nonlinear dynamic signals and software, and provides an effective solution for overcoming the contradiction between the speed and accuracy of modeling nonlinear dynamic systems.