Applying active learning to work with time series predictions

This study examines whether active learning activities based on time series forecasting for engineering students yield statistically significant improvements with respect to performance and engagement. In this context, a pair of team-based learning activities were implemented across two consecutive academic years in a STEM-related course. In the first one, students completed two activities involving statistical analysis of generic datasets, whereas in the second one, two activities centered on time series forecasting were introduced. The expectation was that the contextual relevance of the latter would enhance both engagement and performance. With respect to performance, inferential statistical analysis showed a significant improvement in academic results in the year when time series forecasting activities were implemented. Moreover, the observed effect size required a smaller sample size than the actual cohort, reinforcing the robustness of the outcome. On the other hand, with regards to engagement, the ISA engagement scale was carried out in both courses, where a significant improvement was detected as well. Nonetheless, further research with larger samples is recommended to confirm these findings, as the study was limited to one academic program with a modest sample size. Furthermore, broader validation across institutions and disciplines is needed to generalize the results.