Classified regression for systems engineering

Abstract

Systems engineering is critical for the successful realization of today's complex technology-driven systems. Responsibilities range from operations enhancement and efficiency optimization to providing the necessary technical oversight to achieve system compliance, including on-time delivery, within allocated cost, and performance at or exceeding work scope throughout the system's life cycle. Commensurate with these responsibilities is the capability to design, assess, and mitigate the system triad components. The system evaluations are performed according to established standards and practices that use highly detailed models to calculate, predict, and forecast behavior. These approaches, which rely on an average understanding of the system design, often overgeneralize and lack precision. To improve the system engineering resources for system prediction and forecasting, the method of classified regression is presented. Classified regression is a novel machine learning method that leverages the support vector machine architecture and combines regression and clustering. This stochastic methodology organizes and extracts patterns from small, unstructured datasets while demonstrating high-quality, accurate performance. Built on simplicity, the algorithm is a robust, interpretable alternative to conventional prediction algorithms. Its implementation approach avoids the black-box tendency of complex, non-linear, multivariate models, which often overfit and lack transparency. Classified regression has direct applications in healthcare, finance, marketing, and system development and monitoring.