The increase of e-Learning resources such as interactive learning environments, learning management systems or intelligent tutoring systems has created huge repositories of educational data that can be explored. This increase generated the need of integrating machine learning methodologies into the currently existing e-Learning environments. The integration of such procedures focuses on working with a wide range of data analysis algorithms and their various implementations in form of tools or technologies. This paper aims to present a self-contained roadmap for practitioners who want to have basic knowledge about a core set of algorithms and who want to apply them on educational data. The background of this research domain is represented by state-of-the-art data analysis algorithms found in the areas of Machine Learning, Information Retrieval or Data Mining that are adapted to work on educational data. The main goal of the research efforts in the domain of Intelligent Data Analysis on Educational Data is to provide tools that enhance the quality of the on-line educational systems.
TopApplying advanced algorithms on educational data started back in 1995 as in Barnes (2005) or Choquet et al. (2009). At that moment, models that were used in psychometrics literature made a shift towards educational data mining. Right from the beginning EDM domain defined itself as in interdisciplinary field. Theoretical as well as practical aspects from machine learning, statistics, information retrieval and visualization, and computational modeling are usually found in EDM research area.
An important step forward was made by Romero and Ventura (2007) in the attempt to categorize the main directions of the work that needs to be performed in EDM. At that moment there were defined the main types of procedures that may be applied for solving different problems from EDM. The two categories of procedures were defined: web mining (i.e., clustering, classification, association rule mining, and text mining) and visualization. From this perspective it is obvious that a main source of data that was envisioned to be analyzed was represented by the logs (i.e., the activity) of student-computer interaction. On the other hand, IR on educational data tackles with other types of data: databases (i.e., classical IR), documents, images as in works of Stanescu et al. (2010).
Later, Baker (2010) presented a more refined taxonomy of research work in EDM. The main tracks were prediction, clustering, relationship mining, distillation of data for human judgment and discovery with models. Among the subcategories there were set classification, regression, association rule mining. This more refined taxonomy was due to the advances in the fundamental areas of advanced data analysis algorithms.