Professor University of Oklahoma University of Oklahoma Norman, Oklahoma, United States
Abstract: Many studies have revealed that participating in adult education provides a wide array of benefits for individuals, organizations, and society. Yet, despite increased research efforts, empirical findings are still inconclusive regarding the contextual factors that decisively contribute to determining working adults’ participation in adult education and training (AET). The purpose of this study is to comprehensively re-examine the determinants of AET participation by employing emerging machine learning (ML) techniques to capture population-level insights into what drives working adults’ participation in AET. The data is drawn from the 2017 U.S. Program for the International Assessment of Adult Competencies, from which we selected 1,283 respondents. Using random forest (RF) and gradient boosting machine (GBM) techniques, we first compared the model performances of these two separate algorithms to identify the optimal ML model that predicts AET participation more accurately. Using prediction accuracy indices (e.g., sensitivity, specificity, and positive/negative predictive values), the results demonstrated that the RF model outperformed the GBM classifier in terms of model accuracy. Additionally, variable importance was evaluated by the mean decrease accuracy measure to examine the important factors associated with participation in formal and non-formal AET, respectively. The results showed that working adults’ participation in AET differed across types of AET, suggesting a determining role of human capital (i.e., skills proficiency) and work/life experiences in formal AET, whereas work-related skills utilization played a significant role in non-formal AET. Based on these findings, we discussed several implications and recommendations for adult education theory, policy, and practice.
