A routine that combines Partial Least Squares Structural Equation Modeling (PLS-SEM) with selected Machine Learning (ML) algorithms to leverage both methods' causal prediction and exploration capabilities has been proposed. The triangulation of these two approaches can enhance the predictive accuracy of research models, deepen our understanding of relationships, and help identify new connections, thereby contributing to theory development.

In this study, we illustrate the advantages and challenges of integrating these two methods through a previous study conducted to assess the success of E-learning on student satisfaction and performance based on students' self-assessment. In the previous research, a conceptual model to assess the E-learning success on students' performance and learning achievements was developed and validated, 3S-T model based on DeLone & McLean's information systems model, the Technology Acceptance Model (TAM) and the E-Learning Acceptance Measure (ElAM), with students to different levels and stages of education: Secondary education, baccalaureate and higher education.

In this research, we follow the steps:
1. Use 3S-T model to evaluate the quality of the measurement model and generate latent variable scores.
2. Apply specific ML algorithms to the extracted data to validate existing relationships and discover new (linear) relationships that may extend beyond the initial hypotheses.
3. Assess the theoretical plausibility of alternative models. This approach not only enhances predictive capability but also enriches theoretical insights.

As main conclusions, it can be said that, in our study, we demonstrated the combined use of PLS-SEM and selected ML algorithms, integrating past approaches with new ideas to explore relationships and non-linearities, consolidating new capabilities in PLS-SEM. This study facilitates knowledge transfer between researchers using ML algorithms and PLS-SEM. It will allow academic researchers familiar with PLS-SEM to leverage ML capabilities to enhance their conceptual models.

Keywords: Partial least squares-structural equation modeling (PLS-SEM), Machine learning (ML), Prediction, e-learning Success, Academic performance.