Learning analytics: a study of algorithms for the analysis of educational databases

Alvaro Martinez Navarro, from Universidad Mariana, in Pasto (Colombia), and Pablo Moreno-Ger, director of the Chair IBM-UNIR on Data Science in Education, have published a science article describing the results obtained in an investigation about the analysis techniques applied in educational datasets. The article, published in 2018 in the scientific magazine IJIMAI (International Journal of Interactive Multimedia and Artificial Intelligence), is framed in the field of Learning Analytics, which constitutes a key study area of the Chair.

The algorithms examined throughout the investigation are part of analysis techniques and tools to study the databases generated in the education field. The investigation was launched with the identification of the most relevant clustering algorithms, which were subsequently studied in detail. The second stage, in order to determine the best performance, consisted of the analysis of the algorithms, based on internal validation and stability measurements. The results obtained showed that the algorithms K-means and PAM were the best performers among partition algorithms, and DIANA was selected among hierarchical algorithms.

The investigation intends to contribute to the reinforcement of the theoretical basis of future studies. By focusing on a practical aspect, a performance experimental validation of different clustering techniques is applied to a specific educative dataset. The dataset analyzed during the investigation was extracted from the Computer Engineering students enrolled in the first semester, during the 2010-2016 timeframe, in Universidad Mariana de Pasto.