Knowledge Management and Scientific Productivity in Engineering Doctoral Education: A Predictive and Non-Linear Analysis

Abstract

Scientific productivity during doctoral education in engineering has become a key indicator of academic quality and institutional positioning; however, its development faces limitations associated with individual, institutional, and technological factors. In this context, the objective of this study was to evaluate knowledge management as a methodology for strengthening scientific productivity among engineering doctoral students, considering human, structural, and relational capitals, as well as technological skills. The research adopted a mixed-methods approach, with a longitudinal, descriptive, correlational, and predictive design, covering the period from 2018 to 2024. The population consisted of 188 doctoral students, complemented by 773 valid survey records. The analysis included descriptive statistics, Spearman’s rho correlation, and machine learning models such as random forest, multilayer perceptron, and metaheuristic additive regression. The results reveal positive and statistically significant associations between knowledge management capitals and scientific productivity, with human capital and technological skills showing the strongest relationships. Moreover, non-linear predictive models demonstrated superior performance, suggesting that the relationship between knowledge management and scientific publication is complex and non-additive. It is concluded that knowledge management constitutes an operational and predictive methodology for strengthening doctoral scientific productivity in engineering education.