@unpublished{kurepository51,
     institution = {Keiser University Latin American Campus},
           title = {Interpretable Quadratic Cost Functions for Data-Driven Manufacturing: A Finite-Difference, Calculus-Based Framework Integrating Economic Theory and Explainable Business Analytics},
            note = {Unpublished},
            type = {Discussion Paper},
       publisher = {Keiser University Latin American Campus, San Marcos, Nicaragua},
        keywords = {interpretable mathematical modeling; quadratic cost functions; finite-difference methods; marginal cost analysis; economies of scale; explainable business analytics; Industry 4.0; managerial decision making. Modelado matem{\'a}tico interpretable; funciones de coste cuadr{\'a}ticas; m{\'e}todos de diferencias finitas; an{\'a}lisis de costes marginales; econom{\'i}as de escala; an{\'a}lisis empresarial explicable; Industria 4.0; toma de decisiones gerenciales.},
        abstract = {This study develops a mathematically rigorous, five-stage quadratic cost-modeling framework using seven years of real production data (2015-2021). By integrating finite-difference numerical analysis with differential calculus and microeconomic theory, the deductive-empirical architecture maintains full mathematical transparency over opaque "black-box" machine learning algorithms. The estimated quadratic model captures initial economies of scale through declining marginal costs, and while its predictive accuracy diminishes in later periods due to unmodeled structural shifts, this limitation explicitly serves as a diagnostic tool to highlight model boundaries. Ultimately, the authors contribute a fully replicable, FAIR-aligned pipeline that successfully unifies discrete numerical methods with continuous cost theory to advance explainable business analytics and managerial decision-making.

---------------------
Este estudio desarrolla un marco de modelado de costos cuadr{\'a}tico de cinco etapas, matem{\'a}ticamente riguroso, utilizando datos de producci{\'o}n reales de siete a{\~n}os (2015-2021). Al integrar el an{\'a}lisis num{\'e}rico de diferencias finitas con el c{\'a}lculo diferencial y la teor{\'i}a microecon{\'o}mica, la arquitectura deductivo-emp{\'i}rica mantiene una total transparencia matem{\'a}tica sobre los opacos algoritmos de aprendizaje autom{\'a}tico de "caja negra". El modelo cuadr{\'a}tico estimado captura las econom{\'i}as de escala iniciales mediante la disminuci{\'o}n de los costos marginales, y si bien su precisi{\'o}n predictiva disminuye en per{\'i}odos posteriores debido a cambios estructurales no modelados, esta limitaci{\'o}n sirve expl{\'i}citamente como una herramienta de diagn{\'o}stico para resaltar los l{\'i}mites del modelo. En definitiva, los autores aportan un proceso totalmente replicable y alineado con los principios FAIR que unifica con {\'e}xito los m{\'e}todos num{\'e}ricos discretos con la teor{\'i}a de costos continuos para impulsar el an{\'a}lisis empresarial explicable y la toma de decisiones gerenciales.},
             url = {https://kurepository.keiseruniversity.edu.ni/id/eprint/51/},
          author = {Cruz Treminio, Laura}
}