Students¿ Collaborative Learning through Programming Computational Intelligence Algorithms: A Mixed Methods Approach Applied to a Case Study on Energy and Buildings
Academic Article in Scopus
In response to the global demand for sustainable energy systems and the need to equip future professionals with computational intelligence (CI) skills, this study investigates how collaborative programming of CI algorithms can support student learning in the context of building energy consumption prediction. This research contributes to the limited literature on collaborative CI learning by integrating domain-specific energy data and structured team-based programming, demonstrating a viable instructional model for enhancing algorithmic reasoning in sustainability education. A convergent mixed-methods case study was conducted as an exploratory research study with 14 undergraduate students enrolled in a Data Science course at Tecnológico de Monterrey. Students worked in teams to develop and optimize machine learning and CI models (e.g., ANN, LSTM, GPR) for predicting annual energy use in school buildings. Quantitative metrics such as RMSE and R2 (reported to be R2=0.979 for training and R2=0.981 for testing, coupled with low RMSE=20,504 for training and RMSE=19,948 for testing which were the results of one of the students) assessed model performance, while a structured survey captured cognitive and collaborative learning outcomes. Results revealed that the activity improved students¿ understanding of hyperparameter tuning and model evaluation. High levels of engagement were observed, particularly in statistical analysis and data visualization tasks. However, collaborative dynamics yielded mixed perceptions; while many students appreciated teamwork, others encountered challenges in coordination and equitable task division. Qualitative responses emphasized computational constraints and suggested greater support for model integration and real-world relevance. It is concluded that integrating collaborative programming of computational intelligence algorithms into real-world energy prediction tasks effectively supports the development of students¿ data science competencies, while highlighting the need for structured team coordination to ensure balanced participation and learning outcomes.
