Collaborative learning effectiveness strongly depends on forming balanced student groups. Traditional grouping strategies frequently overlook individual behavioral differences, resulting in suboptimal collaboration dynamics. This study compares two computational techniques Multi-Label k-Nearest Neighbors (ML-kNN) and Genetic Algorithms (GA) to optimize student group formation. Based on behavioral data collected from a Social Learning Network (SLN), students were classified into collaborative, diligent, and engaged profiles. Comparative results indicate that ML-kNN provides rapid classification and is ideal for short-term, dynamic learning activities, whereas GA excels at forming cohesive and balanced groups suited to long-term projects. Additionally, new practical evaluation metrics, including student engagement, learning outcomes, and group interaction satisfaction, are proposed to enhance group effectiveness assessment. A structured framework addressing scalability, ethical considerations, and practical integration into existing Learning Management Systems (LMS) is also provided. This study highlights the importance of context-sensitive method selection and suggests potential benefits from combining ML-kNN’s efficiency with GA’s optimization capabilities within a hybrid grouping model to enhance collaborative learning environments.